freebsd-dev/sys/cam/cam_xpt.c

5612 lines
143 KiB
C
Raw Normal View History

/*-
* Implementation of the Common Access Method Transport (XPT) layer.
*
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 1997, 1998, 1999 Justin T. Gibbs.
* Copyright (c) 1997, 1998, 1999 Kenneth D. Merry.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions, and the following disclaimer,
* without modification, immediately at the beginning of the file.
* 2. The name of the author may not be used to endorse or promote products
* derived from this software without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE FOR
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
2003-06-10 17:50:20 +00:00
#include "opt_printf.h"
2003-06-10 17:50:20 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include <sys/param.h>
#include <sys/bio.h>
#include <sys/bus.h>
#include <sys/systm.h>
#include <sys/types.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
#include <sys/time.h>
#include <sys/conf.h>
#include <sys/fcntl.h>
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
#include <sys/proc.h>
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
#include <sys/sbuf.h>
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
#include <sys/smp.h>
#include <sys/taskqueue.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/sysctl.h>
#include <sys/kthread.h>
#include <cam/cam.h>
#include <cam/cam_ccb.h>
#include <cam/cam_iosched.h>
#include <cam/cam_periph.h>
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
#include <cam/cam_queue.h>
#include <cam/cam_sim.h>
#include <cam/cam_xpt.h>
#include <cam/cam_xpt_sim.h>
#include <cam/cam_xpt_periph.h>
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
#include <cam/cam_xpt_internal.h>
#include <cam/cam_debug.h>
#include <cam/cam_compat.h>
#include <cam/scsi/scsi_all.h>
#include <cam/scsi/scsi_message.h>
#include <cam/scsi/scsi_pass.h>
#include <machine/stdarg.h> /* for xpt_print below */
#include "opt_cam.h"
/* Wild guess based on not wanting to grow the stack too much */
#define XPT_PRINT_MAXLEN 512
#ifdef PRINTF_BUFR_SIZE
#define XPT_PRINT_LEN PRINTF_BUFR_SIZE
#else
#define XPT_PRINT_LEN 128
#endif
_Static_assert(XPT_PRINT_LEN <= XPT_PRINT_MAXLEN, "XPT_PRINT_LEN is too large");
/*
* This is the maximum number of high powered commands (e.g. start unit)
* that can be outstanding at a particular time.
*/
#ifndef CAM_MAX_HIGHPOWER
#define CAM_MAX_HIGHPOWER 4
#endif
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
/* Datastructures internal to the xpt layer */
MALLOC_DEFINE(M_CAMXPT, "CAM XPT", "CAM XPT buffers");
MALLOC_DEFINE(M_CAMDEV, "CAM DEV", "CAM devices");
MALLOC_DEFINE(M_CAMCCB, "CAM CCB", "CAM CCBs");
MALLOC_DEFINE(M_CAMPATH, "CAM path", "CAM paths");
struct xpt_softc {
uint32_t xpt_generation;
/* number of high powered commands that can go through right now */
struct mtx xpt_highpower_lock;
STAILQ_HEAD(highpowerlist, cam_ed) highpowerq;
int num_highpower;
/* queue for handling async rescan requests. */
TAILQ_HEAD(, ccb_hdr) ccb_scanq;
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
int buses_to_config;
int buses_config_done;
int announce_nosbuf;
/*
* Registered buses
*
* N.B., "busses" is an archaic spelling of "buses". In new code
* "buses" is preferred.
*/
TAILQ_HEAD(,cam_eb) xpt_busses;
u_int bus_generation;
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
int boot_delay;
struct callout boot_callout;
struct task boot_task;
struct root_hold_token xpt_rootmount;
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
struct mtx xpt_topo_lock;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct taskqueue *xpt_taskq;
};
typedef enum {
DM_RET_COPY = 0x01,
DM_RET_FLAG_MASK = 0x0f,
DM_RET_NONE = 0x00,
DM_RET_STOP = 0x10,
DM_RET_DESCEND = 0x20,
DM_RET_ERROR = 0x30,
DM_RET_ACTION_MASK = 0xf0
} dev_match_ret;
typedef enum {
XPT_DEPTH_BUS,
XPT_DEPTH_TARGET,
XPT_DEPTH_DEVICE,
XPT_DEPTH_PERIPH
} xpt_traverse_depth;
struct xpt_traverse_config {
xpt_traverse_depth depth;
void *tr_func;
void *tr_arg;
};
typedef int xpt_busfunc_t (struct cam_eb *bus, void *arg);
typedef int xpt_targetfunc_t (struct cam_et *target, void *arg);
typedef int xpt_devicefunc_t (struct cam_ed *device, void *arg);
typedef int xpt_periphfunc_t (struct cam_periph *periph, void *arg);
typedef int xpt_pdrvfunc_t (struct periph_driver **pdrv, void *arg);
/* Transport layer configuration information */
static struct xpt_softc xsoftc;
MTX_SYSINIT(xpt_topo_init, &xsoftc.xpt_topo_lock, "XPT topology lock", MTX_DEF);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
SYSCTL_INT(_kern_cam, OID_AUTO, boot_delay, CTLFLAG_RDTUN,
&xsoftc.boot_delay, 0, "Bus registration wait time");
SYSCTL_UINT(_kern_cam, OID_AUTO, xpt_generation, CTLFLAG_RD,
&xsoftc.xpt_generation, 0, "CAM peripheral generation count");
SYSCTL_INT(_kern_cam, OID_AUTO, announce_nosbuf, CTLFLAG_RWTUN,
&xsoftc.announce_nosbuf, 0, "Don't use sbuf for announcements");
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_doneq {
struct mtx_padalign cam_doneq_mtx;
STAILQ_HEAD(, ccb_hdr) cam_doneq;
int cam_doneq_sleep;
};
static struct cam_doneq cam_doneqs[MAXCPU];
static int cam_num_doneqs;
static struct proc *cam_proc;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
SYSCTL_INT(_kern_cam, OID_AUTO, num_doneqs, CTLFLAG_RDTUN,
&cam_num_doneqs, 0, "Number of completion queues/threads");
struct cam_periph *xpt_periph;
static periph_init_t xpt_periph_init;
static struct periph_driver xpt_driver =
{
xpt_periph_init, "xpt",
TAILQ_HEAD_INITIALIZER(xpt_driver.units), /* generation */ 0,
CAM_PERIPH_DRV_EARLY
};
PERIPHDRIVER_DECLARE(xpt, xpt_driver);
static d_open_t xptopen;
static d_close_t xptclose;
static d_ioctl_t xptioctl;
static d_ioctl_t xptdoioctl;
static struct cdevsw xpt_cdevsw = {
.d_version = D_VERSION,
.d_flags = 0,
.d_open = xptopen,
.d_close = xptclose,
.d_ioctl = xptioctl,
.d_name = "xpt",
};
/* Storage for debugging datastructures */
struct cam_path *cam_dpath;
u_int32_t __read_mostly cam_dflags = CAM_DEBUG_FLAGS;
SYSCTL_UINT(_kern_cam, OID_AUTO, dflags, CTLFLAG_RWTUN,
&cam_dflags, 0, "Enabled debug flags");
u_int32_t cam_debug_delay = CAM_DEBUG_DELAY;
SYSCTL_UINT(_kern_cam, OID_AUTO, debug_delay, CTLFLAG_RWTUN,
&cam_debug_delay, 0, "Delay in us after each debug message");
/* Our boot-time initialization hook */
static int cam_module_event_handler(module_t, int /*modeventtype_t*/, void *);
static moduledata_t cam_moduledata = {
"cam",
cam_module_event_handler,
NULL
};
static int xpt_init(void *);
DECLARE_MODULE(cam, cam_moduledata, SI_SUB_CONFIGURE, SI_ORDER_SECOND);
MODULE_VERSION(cam, 1);
static void xpt_async_bcast(struct async_list *async_head,
u_int32_t async_code,
struct cam_path *path,
void *async_arg);
static path_id_t xptnextfreepathid(void);
static path_id_t xptpathid(const char *sim_name, int sim_unit, int sim_bus);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static union ccb *xpt_get_ccb(struct cam_periph *periph);
static union ccb *xpt_get_ccb_nowait(struct cam_periph *periph);
static void xpt_run_allocq(struct cam_periph *periph, int sleep);
static void xpt_run_allocq_task(void *context, int pending);
static void xpt_run_devq(struct cam_devq *devq);
static callout_func_t xpt_release_devq_timeout;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static void xpt_acquire_bus(struct cam_eb *bus);
static void xpt_release_bus(struct cam_eb *bus);
static uint32_t xpt_freeze_devq_device(struct cam_ed *dev, u_int count);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static int xpt_release_devq_device(struct cam_ed *dev, u_int count,
int run_queue);
static struct cam_et*
xpt_alloc_target(struct cam_eb *bus, target_id_t target_id);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static void xpt_acquire_target(struct cam_et *target);
static void xpt_release_target(struct cam_et *target);
static struct cam_eb*
xpt_find_bus(path_id_t path_id);
static struct cam_et*
xpt_find_target(struct cam_eb *bus, target_id_t target_id);
static struct cam_ed*
xpt_find_device(struct cam_et *target, lun_id_t lun_id);
static void xpt_config(void *arg);
static void xpt_hold_boot_locked(void);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static int xpt_schedule_dev(struct camq *queue, cam_pinfo *dev_pinfo,
u_int32_t new_priority);
static xpt_devicefunc_t xptpassannouncefunc;
static void xptaction(struct cam_sim *sim, union ccb *work_ccb);
static void xptpoll(struct cam_sim *sim);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static void camisr_runqueue(void);
static void xpt_done_process(struct ccb_hdr *ccb_h);
static void xpt_done_td(void *);
static dev_match_ret xptbusmatch(struct dev_match_pattern *patterns,
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
u_int num_patterns, struct cam_eb *bus);
static dev_match_ret xptdevicematch(struct dev_match_pattern *patterns,
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
u_int num_patterns,
struct cam_ed *device);
static dev_match_ret xptperiphmatch(struct dev_match_pattern *patterns,
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
u_int num_patterns,
struct cam_periph *periph);
static xpt_busfunc_t xptedtbusfunc;
static xpt_targetfunc_t xptedttargetfunc;
static xpt_devicefunc_t xptedtdevicefunc;
static xpt_periphfunc_t xptedtperiphfunc;
static xpt_pdrvfunc_t xptplistpdrvfunc;
static xpt_periphfunc_t xptplistperiphfunc;
static int xptedtmatch(struct ccb_dev_match *cdm);
static int xptperiphlistmatch(struct ccb_dev_match *cdm);
static int xptbustraverse(struct cam_eb *start_bus,
xpt_busfunc_t *tr_func, void *arg);
static int xpttargettraverse(struct cam_eb *bus,
struct cam_et *start_target,
xpt_targetfunc_t *tr_func, void *arg);
static int xptdevicetraverse(struct cam_et *target,
struct cam_ed *start_device,
xpt_devicefunc_t *tr_func, void *arg);
static int xptperiphtraverse(struct cam_ed *device,
struct cam_periph *start_periph,
xpt_periphfunc_t *tr_func, void *arg);
static int xptpdrvtraverse(struct periph_driver **start_pdrv,
xpt_pdrvfunc_t *tr_func, void *arg);
static int xptpdperiphtraverse(struct periph_driver **pdrv,
struct cam_periph *start_periph,
xpt_periphfunc_t *tr_func,
void *arg);
static xpt_busfunc_t xptdefbusfunc;
static xpt_targetfunc_t xptdeftargetfunc;
static xpt_devicefunc_t xptdefdevicefunc;
static xpt_periphfunc_t xptdefperiphfunc;
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
static void xpt_finishconfig_task(void *context, int pending);
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
static void xpt_dev_async_default(u_int32_t async_code,
struct cam_eb *bus,
struct cam_et *target,
struct cam_ed *device,
void *async_arg);
static struct cam_ed * xpt_alloc_device_default(struct cam_eb *bus,
struct cam_et *target,
lun_id_t lun_id);
static xpt_devicefunc_t xptsetasyncfunc;
static xpt_busfunc_t xptsetasyncbusfunc;
static cam_status xptregister(struct cam_periph *periph,
void *arg);
static __inline int
xpt_schedule_devq(struct cam_devq *devq, struct cam_ed *dev)
{
int retval;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_assert(&devq->send_mtx, MA_OWNED);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
if ((dev->ccbq.queue.entries > 0) &&
(dev->ccbq.dev_openings > 0) &&
(dev->ccbq.queue.qfrozen_cnt == 0)) {
/*
* The priority of a device waiting for controller
* resources is that of the highest priority CCB
* enqueued.
*/
retval =
xpt_schedule_dev(&devq->send_queue,
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
&dev->devq_entry,
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
CAMQ_GET_PRIO(&dev->ccbq.queue));
} else {
retval = 0;
}
return (retval);
}
static __inline int
device_is_queued(struct cam_ed *device)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
return (device->devq_entry.index != CAM_UNQUEUED_INDEX);
}
static void
xpt_periph_init(void)
{
make_dev(&xpt_cdevsw, 0, UID_ROOT, GID_OPERATOR, 0600, "xpt0");
}
static int
xptopen(struct cdev *dev, int flags, int fmt, struct thread *td)
{
/*
* Only allow read-write access.
*/
if (((flags & FWRITE) == 0) || ((flags & FREAD) == 0))
return(EPERM);
/*
* We don't allow nonblocking access.
*/
if ((flags & O_NONBLOCK) != 0) {
printf("%s: can't do nonblocking access\n", devtoname(dev));
return(ENODEV);
}
return(0);
}
static int
xptclose(struct cdev *dev, int flag, int fmt, struct thread *td)
{
return(0);
}
/*
* Don't automatically grab the xpt softc lock here even though this is going
* through the xpt device. The xpt device is really just a back door for
* accessing other devices and SIMs, so the right thing to do is to grab
* the appropriate SIM lock once the bus/SIM is located.
*/
static int
xptioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flag, struct thread *td)
{
int error;
if ((error = xptdoioctl(dev, cmd, addr, flag, td)) == ENOTTY) {
error = cam_compat_ioctl(dev, cmd, addr, flag, td, xptdoioctl);
}
return (error);
}
static int
xptdoioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flag, struct thread *td)
{
int error;
error = 0;
switch(cmd) {
/*
* For the transport layer CAMIOCOMMAND ioctl, we really only want
* to accept CCB types that don't quite make sense to send through a
* passthrough driver. XPT_PATH_INQ is an exception to this, as stated
* in the CAM spec.
*/
case CAMIOCOMMAND: {
union ccb *ccb;
union ccb *inccb;
struct cam_eb *bus;
inccb = (union ccb *)addr;
#if defined(BUF_TRACKING) || defined(FULL_BUF_TRACKING)
if (inccb->ccb_h.func_code == XPT_SCSI_IO)
inccb->csio.bio = NULL;
#endif
if (inccb->ccb_h.flags & CAM_UNLOCKED)
return (EINVAL);
bus = xpt_find_bus(inccb->ccb_h.path_id);
if (bus == NULL)
return (EINVAL);
switch (inccb->ccb_h.func_code) {
case XPT_SCAN_BUS:
case XPT_RESET_BUS:
if (inccb->ccb_h.target_id != CAM_TARGET_WILDCARD ||
inccb->ccb_h.target_lun != CAM_LUN_WILDCARD) {
xpt_release_bus(bus);
return (EINVAL);
}
break;
case XPT_SCAN_TGT:
if (inccb->ccb_h.target_id == CAM_TARGET_WILDCARD ||
inccb->ccb_h.target_lun != CAM_LUN_WILDCARD) {
xpt_release_bus(bus);
return (EINVAL);
}
break;
default:
break;
}
switch(inccb->ccb_h.func_code) {
case XPT_SCAN_BUS:
case XPT_RESET_BUS:
case XPT_PATH_INQ:
case XPT_ENG_INQ:
case XPT_SCAN_LUN:
case XPT_SCAN_TGT:
ccb = xpt_alloc_ccb();
/*
* Create a path using the bus, target, and lun the
* user passed in.
*/
if (xpt_create_path(&ccb->ccb_h.path, NULL,
inccb->ccb_h.path_id,
inccb->ccb_h.target_id,
inccb->ccb_h.target_lun) !=
CAM_REQ_CMP){
error = EINVAL;
xpt_free_ccb(ccb);
break;
}
/* Ensure all of our fields are correct */
xpt_setup_ccb(&ccb->ccb_h, ccb->ccb_h.path,
inccb->ccb_h.pinfo.priority);
xpt_merge_ccb(ccb, inccb);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_path_lock(ccb->ccb_h.path);
cam_periph_runccb(ccb, NULL, 0, 0, NULL);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_path_unlock(ccb->ccb_h.path);
bcopy(ccb, inccb, sizeof(union ccb));
xpt_free_path(ccb->ccb_h.path);
xpt_free_ccb(ccb);
break;
case XPT_DEBUG: {
union ccb ccb;
/*
* This is an immediate CCB, so it's okay to
* allocate it on the stack.
*/
/*
* Create a path using the bus, target, and lun the
* user passed in.
*/
if (xpt_create_path(&ccb.ccb_h.path, NULL,
inccb->ccb_h.path_id,
inccb->ccb_h.target_id,
inccb->ccb_h.target_lun) !=
CAM_REQ_CMP){
error = EINVAL;
break;
}
/* Ensure all of our fields are correct */
xpt_setup_ccb(&ccb.ccb_h, ccb.ccb_h.path,
inccb->ccb_h.pinfo.priority);
xpt_merge_ccb(&ccb, inccb);
xpt_action(&ccb);
bcopy(&ccb, inccb, sizeof(union ccb));
xpt_free_path(ccb.ccb_h.path);
break;
}
case XPT_DEV_MATCH: {
struct cam_periph_map_info mapinfo;
struct cam_path *old_path;
/*
* We can't deal with physical addresses for this
* type of transaction.
*/
if ((inccb->ccb_h.flags & CAM_DATA_MASK) !=
CAM_DATA_VADDR) {
error = EINVAL;
break;
}
/*
* Save this in case the caller had it set to
* something in particular.
*/
old_path = inccb->ccb_h.path;
/*
* We really don't need a path for the matching
* code. The path is needed because of the
* debugging statements in xpt_action(). They
* assume that the CCB has a valid path.
*/
inccb->ccb_h.path = xpt_periph->path;
bzero(&mapinfo, sizeof(mapinfo));
/*
* Map the pattern and match buffers into kernel
* virtual address space.
*/
error = cam_periph_mapmem(inccb, &mapinfo, MAXPHYS);
if (error) {
inccb->ccb_h.path = old_path;
break;
}
/*
* This is an immediate CCB, we can send it on directly.
*/
xpt_action(inccb);
/*
* Map the buffers back into user space.
*/
cam_periph_unmapmem(inccb, &mapinfo);
inccb->ccb_h.path = old_path;
error = 0;
break;
}
default:
error = ENOTSUP;
break;
}
xpt_release_bus(bus);
break;
}
/*
* This is the getpassthru ioctl. It takes a XPT_GDEVLIST ccb as input,
* with the periphal driver name and unit name filled in. The other
* fields don't really matter as input. The passthrough driver name
* ("pass"), and unit number are passed back in the ccb. The current
* device generation number, and the index into the device peripheral
* driver list, and the status are also passed back. Note that
* since we do everything in one pass, unlike the XPT_GDEVLIST ccb,
* we never return a status of CAM_GDEVLIST_LIST_CHANGED. It is
* (or rather should be) impossible for the device peripheral driver
* list to change since we look at the whole thing in one pass, and
* we do it with lock protection.
2008-01-02 01:45:31 +00:00
*
*/
case CAMGETPASSTHRU: {
union ccb *ccb;
struct cam_periph *periph;
struct periph_driver **p_drv;
char *name;
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
u_int unit;
int base_periph_found;
ccb = (union ccb *)addr;
unit = ccb->cgdl.unit_number;
name = ccb->cgdl.periph_name;
base_periph_found = 0;
#if defined(BUF_TRACKING) || defined(FULL_BUF_TRACKING)
if (ccb->ccb_h.func_code == XPT_SCSI_IO)
ccb->csio.bio = NULL;
#endif
/*
* Sanity check -- make sure we don't get a null peripheral
* driver name.
*/
if (*ccb->cgdl.periph_name == '\0') {
error = EINVAL;
break;
}
/* Keep the list from changing while we traverse it */
xpt_lock_buses();
/* first find our driver in the list of drivers */
for (p_drv = periph_drivers; *p_drv != NULL; p_drv++)
if (strcmp((*p_drv)->driver_name, name) == 0)
break;
if (*p_drv == NULL) {
xpt_unlock_buses();
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
ccb->cgdl.status = CAM_GDEVLIST_ERROR;
*ccb->cgdl.periph_name = '\0';
ccb->cgdl.unit_number = 0;
error = ENOENT;
break;
2008-01-02 01:45:31 +00:00
}
/*
* Run through every peripheral instance of this driver
* and check to see whether it matches the unit passed
* in by the user. If it does, get out of the loops and
* find the passthrough driver associated with that
* peripheral driver.
*/
for (periph = TAILQ_FIRST(&(*p_drv)->units); periph != NULL;
periph = TAILQ_NEXT(periph, unit_links)) {
if (periph->unit_number == unit)
break;
}
/*
* If we found the peripheral driver that the user passed
* in, go through all of the peripheral drivers for that
* particular device and look for a passthrough driver.
*/
if (periph != NULL) {
struct cam_ed *device;
int i;
base_periph_found = 1;
device = periph->path->device;
for (i = 0, periph = SLIST_FIRST(&device->periphs);
periph != NULL;
periph = SLIST_NEXT(periph, periph_links), i++) {
/*
* Check to see whether we have a
2008-01-02 01:45:31 +00:00
* passthrough device or not.
*/
if (strcmp(periph->periph_name, "pass") == 0) {
/*
* Fill in the getdevlist fields.
*/
strlcpy(ccb->cgdl.periph_name,
periph->periph_name,
sizeof(ccb->cgdl.periph_name));
ccb->cgdl.unit_number =
periph->unit_number;
if (SLIST_NEXT(periph, periph_links))
ccb->cgdl.status =
CAM_GDEVLIST_MORE_DEVS;
else
ccb->cgdl.status =
CAM_GDEVLIST_LAST_DEVICE;
ccb->cgdl.generation =
device->generation;
ccb->cgdl.index = i;
/*
* Fill in some CCB header fields
* that the user may want.
*/
ccb->ccb_h.path_id =
periph->path->bus->path_id;
ccb->ccb_h.target_id =
periph->path->target->target_id;
ccb->ccb_h.target_lun =
periph->path->device->lun_id;
ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
}
}
/*
* If the periph is null here, one of two things has
* happened. The first possibility is that we couldn't
* find the unit number of the particular peripheral driver
* that the user is asking about. e.g. the user asks for
* the passthrough driver for "da11". We find the list of
* "da" peripherals all right, but there is no unit 11.
* The other possibility is that we went through the list
* of peripheral drivers attached to the device structure,
* but didn't find one with the name "pass". Either way,
* we return ENOENT, since we couldn't find something.
*/
if (periph == NULL) {
ccb->ccb_h.status = CAM_REQ_CMP_ERR;
ccb->cgdl.status = CAM_GDEVLIST_ERROR;
*ccb->cgdl.periph_name = '\0';
ccb->cgdl.unit_number = 0;
error = ENOENT;
/*
* It is unfortunate that this is even necessary,
* but there are many, many clueless users out there.
* If this is true, the user is looking for the
* passthrough driver, but doesn't have one in his
* kernel.
*/
if (base_periph_found == 1) {
printf("xptioctl: pass driver is not in the "
"kernel\n");
printf("xptioctl: put \"device pass\" in "
"your kernel config file\n");
}
}
xpt_unlock_buses();
break;
}
default:
error = ENOTTY;
break;
}
return(error);
}
static int
cam_module_event_handler(module_t mod, int what, void *arg)
{
int error;
switch (what) {
case MOD_LOAD:
if ((error = xpt_init(NULL)) != 0)
return (error);
break;
case MOD_UNLOAD:
return EBUSY;
default:
return EOPNOTSUPP;
}
return 0;
}
static struct xpt_proto *
xpt_proto_find(cam_proto proto)
{
struct xpt_proto **pp;
SET_FOREACH(pp, cam_xpt_proto_set) {
if ((*pp)->proto == proto)
return *pp;
}
return NULL;
}
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
static void
xpt_rescan_done(struct cam_periph *periph, union ccb *done_ccb)
{
if (done_ccb->ccb_h.ppriv_ptr1 == NULL) {
xpt_free_path(done_ccb->ccb_h.path);
xpt_free_ccb(done_ccb);
} else {
done_ccb->ccb_h.cbfcnp = done_ccb->ccb_h.ppriv_ptr1;
(*done_ccb->ccb_h.cbfcnp)(periph, done_ccb);
}
xpt_release_boot();
}
/* thread to handle bus rescans */
static void
xpt_scanner_thread(void *dummy)
{
union ccb *ccb;
struct mtx *mtx;
struct cam_ed *device;
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_lock_buses();
for (;;) {
if (TAILQ_EMPTY(&xsoftc.ccb_scanq))
msleep(&xsoftc.ccb_scanq, &xsoftc.xpt_topo_lock, PRIBIO,
"-", 0);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
if ((ccb = (union ccb *)TAILQ_FIRST(&xsoftc.ccb_scanq)) != NULL) {
TAILQ_REMOVE(&xsoftc.ccb_scanq, &ccb->ccb_h, sim_links.tqe);
xpt_unlock_buses();
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
/*
* We need to lock the device's mutex which we use as
* the path mutex. We can't do it directly because the
* cam_path in the ccb may wind up going away because
* the path lock may be dropped and the path retired in
* the completion callback. We do this directly to keep
* the reference counts in cam_path sane. We also have
* to copy the device pointer because ccb_h.path may
* be freed in the callback.
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
*/
mtx = xpt_path_mtx(ccb->ccb_h.path);
device = ccb->ccb_h.path->device;
xpt_acquire_device(device);
mtx_lock(mtx);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_action(ccb);
mtx_unlock(mtx);
xpt_release_device(device);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_lock_buses();
}
}
}
void
xpt_rescan(union ccb *ccb)
{
struct ccb_hdr *hdr;
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
/* Prepare request */
if (ccb->ccb_h.path->target->target_id == CAM_TARGET_WILDCARD &&
ccb->ccb_h.path->device->lun_id == CAM_LUN_WILDCARD)
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
ccb->ccb_h.func_code = XPT_SCAN_BUS;
else if (ccb->ccb_h.path->target->target_id != CAM_TARGET_WILDCARD &&
ccb->ccb_h.path->device->lun_id == CAM_LUN_WILDCARD)
ccb->ccb_h.func_code = XPT_SCAN_TGT;
else if (ccb->ccb_h.path->target->target_id != CAM_TARGET_WILDCARD &&
ccb->ccb_h.path->device->lun_id != CAM_LUN_WILDCARD)
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
ccb->ccb_h.func_code = XPT_SCAN_LUN;
else {
xpt_print(ccb->ccb_h.path, "illegal scan path\n");
xpt_free_path(ccb->ccb_h.path);
xpt_free_ccb(ccb);
return;
}
CAM_DEBUG(ccb->ccb_h.path, CAM_DEBUG_TRACE,
("xpt_rescan: func %#x %s\n", ccb->ccb_h.func_code,
xpt_action_name(ccb->ccb_h.func_code)));
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
ccb->ccb_h.ppriv_ptr1 = ccb->ccb_h.cbfcnp;
ccb->ccb_h.cbfcnp = xpt_rescan_done;
xpt_setup_ccb(&ccb->ccb_h, ccb->ccb_h.path, CAM_PRIORITY_XPT);
/* Don't make duplicate entries for the same paths. */
xpt_lock_buses();
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
if (ccb->ccb_h.ppriv_ptr1 == NULL) {
TAILQ_FOREACH(hdr, &xsoftc.ccb_scanq, sim_links.tqe) {
if (xpt_path_comp(hdr->path, ccb->ccb_h.path) == 0) {
wakeup(&xsoftc.ccb_scanq);
xpt_unlock_buses();
xpt_print(ccb->ccb_h.path, "rescan already queued\n");
xpt_free_path(ccb->ccb_h.path);
xpt_free_ccb(ccb);
return;
}
}
}
TAILQ_INSERT_TAIL(&xsoftc.ccb_scanq, &ccb->ccb_h, sim_links.tqe);
xpt_hold_boot_locked();
wakeup(&xsoftc.ccb_scanq);
xpt_unlock_buses();
}
/* Functions accessed by the peripheral drivers */
static int
xpt_init(void *dummy)
{
struct cam_sim *xpt_sim;
struct cam_path *path;
struct cam_devq *devq;
cam_status status;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
int error, i;
TAILQ_INIT(&xsoftc.xpt_busses);
TAILQ_INIT(&xsoftc.ccb_scanq);
STAILQ_INIT(&xsoftc.highpowerq);
xsoftc.num_highpower = CAM_MAX_HIGHPOWER;
mtx_init(&xsoftc.xpt_highpower_lock, "XPT highpower lock", NULL, MTX_DEF);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xsoftc.xpt_taskq = taskqueue_create("CAM XPT task", M_WAITOK,
taskqueue_thread_enqueue, /*context*/&xsoftc.xpt_taskq);
#ifdef CAM_BOOT_DELAY
/*
* Override this value at compile time to assist our users
* who don't use loader to boot a kernel.
*/
xsoftc.boot_delay = CAM_BOOT_DELAY;
#endif
/*
* The xpt layer is, itself, the equivalent of a SIM.
* Allow 16 ccbs in the ccb pool for it. This should
* give decent parallelism when we probe buses and
* perform other XPT functions.
*/
devq = cam_simq_alloc(16);
xpt_sim = cam_sim_alloc(xptaction,
xptpoll,
"xpt",
/*softc*/NULL,
/*unit*/0,
/*mtx*/NULL,
/*max_dev_transactions*/0,
/*max_tagged_dev_transactions*/0,
devq);
if (xpt_sim == NULL)
return (ENOMEM);
if ((status = xpt_bus_register(xpt_sim, NULL, 0)) != CAM_SUCCESS) {
2006-12-21 23:20:51 +00:00
printf("xpt_init: xpt_bus_register failed with status %#x,"
" failing attach\n", status);
return (EINVAL);
}
/*
* Looking at the XPT from the SIM layer, the XPT is
* the equivalent of a peripheral driver. Allocate
* a peripheral driver entry for us.
*/
if ((status = xpt_create_path(&path, NULL, CAM_XPT_PATH_ID,
CAM_TARGET_WILDCARD,
CAM_LUN_WILDCARD)) != CAM_REQ_CMP) {
printf("xpt_init: xpt_create_path failed with status %#x,"
" failing attach\n", status);
return (EINVAL);
}
xpt_path_lock(path);
Fix a problem with the way we handled device invalidation when attaching to a device failed. In theory, the same steps that happen when we get an AC_LOST_DEVICE async notification should have been taken when a driver fails to attach. In practice, that wasn't the case. This only affected the da, cd and ch drivers, but the fix affects all peripheral drivers. There were several possible problems: - In the da driver, we didn't remove the peripheral's softc from the da driver's linked list of softcs. Once the peripheral and softc got removed, we'd get a kernel panic the next time the timeout routine called dasendorderedtag(). - In the da, cd and possibly ch drivers, we didn't remove the peripheral's devstat structure from the devstat queue. Once the peripheral and softc were removed, this could cause a panic if anyone tried to access device statistics. (one component of the linked list wouldn't exist anymore) - In the cd driver, we didn't take the peripheral off the changer run queue if it was scheduled to run. In practice, it's highly unlikely, and maybe impossible that the peripheral would have been on the changer run queue at that stage of the probe process. The fix is: - Add a new peripheral callback function (the "oninvalidate" function) that is called the first time cam_periph_invalidate() is called for a peripheral. - Create new foooninvalidate() routines for each peripheral driver. This routine is always called at splsoftcam(), and contains all the stuff that used to be in the AC_LOST_DEVICE case of the async callback handler. - Move the devstat cleanup call to the destructor/cleanup routines, since some of the drivers do I/O in their close routines. - Make sure that when we're flushing the buffer queue, we traverse it at splbio(). - Add a check for the invalid flag in the pt driver's open routine. Reviewed by: gibbs
1998-10-22 22:16:56 +00:00
cam_periph_alloc(xptregister, NULL, NULL, NULL, "xpt", CAM_PERIPH_BIO,
path, NULL, 0, xpt_sim);
xpt_path_unlock(path);
xpt_free_path(path);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (cam_num_doneqs < 1)
cam_num_doneqs = 1 + mp_ncpus / 6;
else if (cam_num_doneqs > MAXCPU)
cam_num_doneqs = MAXCPU;
for (i = 0; i < cam_num_doneqs; i++) {
mtx_init(&cam_doneqs[i].cam_doneq_mtx, "CAM doneq", NULL,
MTX_DEF);
STAILQ_INIT(&cam_doneqs[i].cam_doneq);
error = kproc_kthread_add(xpt_done_td, &cam_doneqs[i],
&cam_proc, NULL, 0, 0, "cam", "doneq%d", i);
if (error != 0) {
cam_num_doneqs = i;
break;
}
}
if (cam_num_doneqs < 1) {
printf("xpt_init: Cannot init completion queues "
"- failing attach\n");
return (ENOMEM);
}
/*
* Register a callback for when interrupts are enabled.
*/
config_intrhook_oneshot(xpt_config, NULL);
return (0);
}
static cam_status
xptregister(struct cam_periph *periph, void *arg)
{
struct cam_sim *xpt_sim;
if (periph == NULL) {
printf("xptregister: periph was NULL!!\n");
return(CAM_REQ_CMP_ERR);
}
xpt_sim = (struct cam_sim *)arg;
xpt_sim->softc = periph;
xpt_periph = periph;
periph->softc = NULL;
return(CAM_REQ_CMP);
}
int32_t
xpt_add_periph(struct cam_periph *periph)
{
struct cam_ed *device;
int32_t status;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
TASK_INIT(&periph->periph_run_task, 0, xpt_run_allocq_task, periph);
device = periph->path->device;
status = CAM_REQ_CMP;
if (device != NULL) {
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&device->target->bus->eb_mtx);
device->generation++;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
SLIST_INSERT_HEAD(&device->periphs, periph, periph_links);
mtx_unlock(&device->target->bus->eb_mtx);
atomic_add_32(&xsoftc.xpt_generation, 1);
}
return (status);
}
void
xpt_remove_periph(struct cam_periph *periph)
{
struct cam_ed *device;
device = periph->path->device;
if (device != NULL) {
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&device->target->bus->eb_mtx);
device->generation++;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
SLIST_REMOVE(&device->periphs, periph, cam_periph, periph_links);
mtx_unlock(&device->target->bus->eb_mtx);
atomic_add_32(&xsoftc.xpt_generation, 1);
}
}
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
void
xpt_announce_periph(struct cam_periph *periph, char *announce_string)
{
struct cam_path *path = periph->path;
struct xpt_proto *proto;
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
cam_periph_assert(periph, MA_OWNED);
periph->flags |= CAM_PERIPH_ANNOUNCED;
printf("%s%d at %s%d bus %d scbus%d target %d lun %jx\n",
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
periph->periph_name, periph->unit_number,
path->bus->sim->sim_name,
path->bus->sim->unit_number,
path->bus->sim->bus_id,
path->bus->path_id,
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
path->target->target_id,
(uintmax_t)path->device->lun_id);
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
printf("%s%d: ", periph->periph_name, periph->unit_number);
proto = xpt_proto_find(path->device->protocol);
if (proto)
proto->ops->announce(path->device);
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
else
printf("%s%d: Unknown protocol device %d\n",
periph->periph_name, periph->unit_number,
path->device->protocol);
if (path->device->serial_num_len > 0) {
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
/* Don't wrap the screen - print only the first 60 chars */
printf("%s%d: Serial Number %.60s\n", periph->periph_name,
periph->unit_number, path->device->serial_num);
}
/* Announce transport details. */
path->bus->xport->ops->announce(periph);
/* Announce command queueing. */
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
if (path->device->inq_flags & SID_CmdQue
|| path->device->flags & CAM_DEV_TAG_AFTER_COUNT) {
2010-02-26 10:33:48 +00:00
printf("%s%d: Command Queueing enabled\n",
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
periph->periph_name, periph->unit_number);
}
/* Announce caller's details if they've passed in. */
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
if (announce_string != NULL)
printf("%s%d: %s\n", periph->periph_name,
periph->unit_number, announce_string);
}
void
xpt_announce_periph_sbuf(struct cam_periph *periph, struct sbuf *sb,
char *announce_string)
{
struct cam_path *path = periph->path;
struct xpt_proto *proto;
cam_periph_assert(periph, MA_OWNED);
periph->flags |= CAM_PERIPH_ANNOUNCED;
/* Fall back to the non-sbuf method if necessary */
if (xsoftc.announce_nosbuf != 0) {
xpt_announce_periph(periph, announce_string);
return;
}
proto = xpt_proto_find(path->device->protocol);
if (((proto != NULL) && (proto->ops->announce_sbuf == NULL)) ||
(path->bus->xport->ops->announce_sbuf == NULL)) {
xpt_announce_periph(periph, announce_string);
return;
}
sbuf_printf(sb, "%s%d at %s%d bus %d scbus%d target %d lun %jx\n",
periph->periph_name, periph->unit_number,
path->bus->sim->sim_name,
path->bus->sim->unit_number,
path->bus->sim->bus_id,
path->bus->path_id,
path->target->target_id,
(uintmax_t)path->device->lun_id);
sbuf_printf(sb, "%s%d: ", periph->periph_name, periph->unit_number);
if (proto)
proto->ops->announce_sbuf(path->device, sb);
else
sbuf_printf(sb, "%s%d: Unknown protocol device %d\n",
periph->periph_name, periph->unit_number,
path->device->protocol);
if (path->device->serial_num_len > 0) {
/* Don't wrap the screen - print only the first 60 chars */
sbuf_printf(sb, "%s%d: Serial Number %.60s\n",
periph->periph_name, periph->unit_number,
path->device->serial_num);
}
/* Announce transport details. */
path->bus->xport->ops->announce_sbuf(periph, sb);
/* Announce command queueing. */
if (path->device->inq_flags & SID_CmdQue
|| path->device->flags & CAM_DEV_TAG_AFTER_COUNT) {
sbuf_printf(sb, "%s%d: Command Queueing enabled\n",
periph->periph_name, periph->unit_number);
}
/* Announce caller's details if they've passed in. */
if (announce_string != NULL)
sbuf_printf(sb, "%s%d: %s\n", periph->periph_name,
periph->unit_number, announce_string);
}
void
xpt_announce_quirks(struct cam_periph *periph, int quirks, char *bit_string)
{
if (quirks != 0) {
printf("%s%d: quirks=0x%b\n", periph->periph_name,
periph->unit_number, quirks, bit_string);
}
}
void
xpt_announce_quirks_sbuf(struct cam_periph *periph, struct sbuf *sb,
int quirks, char *bit_string)
{
if (xsoftc.announce_nosbuf != 0) {
xpt_announce_quirks(periph, quirks, bit_string);
return;
}
if (quirks != 0) {
sbuf_printf(sb, "%s%d: quirks=0x%b\n", periph->periph_name,
periph->unit_number, quirks, bit_string);
}
}
void
xpt_denounce_periph(struct cam_periph *periph)
{
struct cam_path *path = periph->path;
struct xpt_proto *proto;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
cam_periph_assert(periph, MA_OWNED);
printf("%s%d at %s%d bus %d scbus%d target %d lun %jx\n",
periph->periph_name, periph->unit_number,
path->bus->sim->sim_name,
path->bus->sim->unit_number,
path->bus->sim->bus_id,
path->bus->path_id,
path->target->target_id,
(uintmax_t)path->device->lun_id);
printf("%s%d: ", periph->periph_name, periph->unit_number);
proto = xpt_proto_find(path->device->protocol);
if (proto)
proto->ops->denounce(path->device);
else
printf("%s%d: Unknown protocol device %d\n",
periph->periph_name, periph->unit_number,
path->device->protocol);
if (path->device->serial_num_len > 0)
printf(" s/n %.60s", path->device->serial_num);
printf(" detached\n");
}
void
xpt_denounce_periph_sbuf(struct cam_periph *periph, struct sbuf *sb)
{
struct cam_path *path = periph->path;
struct xpt_proto *proto;
cam_periph_assert(periph, MA_OWNED);
/* Fall back to the non-sbuf method if necessary */
if (xsoftc.announce_nosbuf != 0) {
xpt_denounce_periph(periph);
return;
}
proto = xpt_proto_find(path->device->protocol);
if ((proto != NULL) && (proto->ops->denounce_sbuf == NULL)) {
xpt_denounce_periph(periph);
return;
}
sbuf_printf(sb, "%s%d at %s%d bus %d scbus%d target %d lun %jx\n",
periph->periph_name, periph->unit_number,
path->bus->sim->sim_name,
path->bus->sim->unit_number,
path->bus->sim->bus_id,
path->bus->path_id,
path->target->target_id,
(uintmax_t)path->device->lun_id);
sbuf_printf(sb, "%s%d: ", periph->periph_name, periph->unit_number);
if (proto)
proto->ops->denounce_sbuf(path->device, sb);
else
sbuf_printf(sb, "%s%d: Unknown protocol device %d\n",
periph->periph_name, periph->unit_number,
path->device->protocol);
if (path->device->serial_num_len > 0)
sbuf_printf(sb, " s/n %.60s", path->device->serial_num);
sbuf_printf(sb, " detached\n");
}
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
int
xpt_getattr(char *buf, size_t len, const char *attr, struct cam_path *path)
{
int ret = -1, l, o;
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
struct ccb_dev_advinfo cdai;
struct scsi_vpd_device_id *did;
struct scsi_vpd_id_descriptor *idd;
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_path_assert(path, MA_OWNED);
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
memset(&cdai, 0, sizeof(cdai));
xpt_setup_ccb(&cdai.ccb_h, path, CAM_PRIORITY_NORMAL);
cdai.ccb_h.func_code = XPT_DEV_ADVINFO;
cdai.flags = CDAI_FLAG_NONE;
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
cdai.bufsiz = len;
cdai.buf = buf;
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
if (!strcmp(attr, "GEOM::ident"))
cdai.buftype = CDAI_TYPE_SERIAL_NUM;
else if (!strcmp(attr, "GEOM::physpath"))
cdai.buftype = CDAI_TYPE_PHYS_PATH;
else if (strcmp(attr, "GEOM::lunid") == 0 ||
strcmp(attr, "GEOM::lunname") == 0) {
cdai.buftype = CDAI_TYPE_SCSI_DEVID;
cdai.bufsiz = CAM_SCSI_DEVID_MAXLEN;
cdai.buf = malloc(cdai.bufsiz, M_CAMXPT, M_NOWAIT);
if (cdai.buf == NULL) {
ret = ENOMEM;
goto out;
}
} else
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
goto out;
xpt_action((union ccb *)&cdai); /* can only be synchronous */
if ((cdai.ccb_h.status & CAM_DEV_QFRZN) != 0)
cam_release_devq(cdai.ccb_h.path, 0, 0, 0, FALSE);
if (cdai.provsiz == 0)
goto out;
switch(cdai.buftype) {
case CDAI_TYPE_SCSI_DEVID:
did = (struct scsi_vpd_device_id *)cdai.buf;
if (strcmp(attr, "GEOM::lunid") == 0) {
idd = scsi_get_devid(did, cdai.provsiz,
scsi_devid_is_lun_naa);
if (idd == NULL)
idd = scsi_get_devid(did, cdai.provsiz,
scsi_devid_is_lun_eui64);
if (idd == NULL)
idd = scsi_get_devid(did, cdai.provsiz,
scsi_devid_is_lun_uuid);
if (idd == NULL)
idd = scsi_get_devid(did, cdai.provsiz,
scsi_devid_is_lun_md5);
} else
idd = NULL;
if (idd == NULL)
idd = scsi_get_devid(did, cdai.provsiz,
scsi_devid_is_lun_t10);
if (idd == NULL)
idd = scsi_get_devid(did, cdai.provsiz,
scsi_devid_is_lun_name);
if (idd == NULL)
break;
ret = 0;
if ((idd->proto_codeset & SVPD_ID_CODESET_MASK) ==
SVPD_ID_CODESET_ASCII) {
if (idd->length < len) {
for (l = 0; l < idd->length; l++)
buf[l] = idd->identifier[l] ?
idd->identifier[l] : ' ';
buf[l] = 0;
} else
ret = EFAULT;
break;
}
if ((idd->proto_codeset & SVPD_ID_CODESET_MASK) ==
SVPD_ID_CODESET_UTF8) {
l = strnlen(idd->identifier, idd->length);
if (l < len) {
bcopy(idd->identifier, buf, l);
buf[l] = 0;
} else
ret = EFAULT;
break;
}
if ((idd->id_type & SVPD_ID_TYPE_MASK) ==
SVPD_ID_TYPE_UUID && idd->identifier[0] == 0x10) {
if ((idd->length - 2) * 2 + 4 >= len) {
ret = EFAULT;
break;
}
for (l = 2, o = 0; l < idd->length; l++) {
if (l == 6 || l == 8 || l == 10 || l == 12)
o += sprintf(buf + o, "-");
o += sprintf(buf + o, "%02x",
idd->identifier[l]);
}
break;
}
if (idd->length * 2 < len) {
for (l = 0; l < idd->length; l++)
sprintf(buf + l * 2, "%02x",
idd->identifier[l]);
} else
ret = EFAULT;
break;
default:
if (cdai.provsiz < len) {
cdai.buf[cdai.provsiz] = 0;
ret = 0;
} else
ret = EFAULT;
break;
}
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
out:
if ((char *)cdai.buf != buf)
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
free(cdai.buf, M_CAMXPT);
return ret;
}
static dev_match_ret
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
xptbusmatch(struct dev_match_pattern *patterns, u_int num_patterns,
struct cam_eb *bus)
{
dev_match_ret retval;
u_int i;
retval = DM_RET_NONE;
/*
* If we aren't given something to match against, that's an error.
*/
if (bus == NULL)
return(DM_RET_ERROR);
/*
* If there are no match entries, then this bus matches no
* matter what.
*/
if ((patterns == NULL) || (num_patterns == 0))
return(DM_RET_DESCEND | DM_RET_COPY);
for (i = 0; i < num_patterns; i++) {
struct bus_match_pattern *cur_pattern;
/*
* If the pattern in question isn't for a bus node, we
* aren't interested. However, we do indicate to the
* calling routine that we should continue descending the
* tree, since the user wants to match against lower-level
* EDT elements.
*/
if (patterns[i].type != DEV_MATCH_BUS) {
if ((retval & DM_RET_ACTION_MASK) == DM_RET_NONE)
retval |= DM_RET_DESCEND;
continue;
}
cur_pattern = &patterns[i].pattern.bus_pattern;
/*
* If they want to match any bus node, we give them any
* device node.
*/
if (cur_pattern->flags == BUS_MATCH_ANY) {
/* set the copy flag */
retval |= DM_RET_COPY;
/*
* If we've already decided on an action, go ahead
* and return.
*/
if ((retval & DM_RET_ACTION_MASK) != DM_RET_NONE)
return(retval);
}
/*
* Not sure why someone would do this...
*/
if (cur_pattern->flags == BUS_MATCH_NONE)
continue;
if (((cur_pattern->flags & BUS_MATCH_PATH) != 0)
&& (cur_pattern->path_id != bus->path_id))
continue;
if (((cur_pattern->flags & BUS_MATCH_BUS_ID) != 0)
&& (cur_pattern->bus_id != bus->sim->bus_id))
continue;
if (((cur_pattern->flags & BUS_MATCH_UNIT) != 0)
&& (cur_pattern->unit_number != bus->sim->unit_number))
continue;
if (((cur_pattern->flags & BUS_MATCH_NAME) != 0)
&& (strncmp(cur_pattern->dev_name, bus->sim->sim_name,
DEV_IDLEN) != 0))
continue;
/*
2008-01-02 01:45:31 +00:00
* If we get to this point, the user definitely wants
* information on this bus. So tell the caller to copy the
* data out.
*/
retval |= DM_RET_COPY;
/*
* If the return action has been set to descend, then we
* know that we've already seen a non-bus matching
* expression, therefore we need to further descend the tree.
* This won't change by continuing around the loop, so we
* go ahead and return. If we haven't seen a non-bus
* matching expression, we keep going around the loop until
* we exhaust the matching expressions. We'll set the stop
* flag once we fall out of the loop.
*/
if ((retval & DM_RET_ACTION_MASK) == DM_RET_DESCEND)
return(retval);
}
/*
* If the return action hasn't been set to descend yet, that means
* we haven't seen anything other than bus matching patterns. So
* tell the caller to stop descending the tree -- the user doesn't
* want to match against lower level tree elements.
*/
if ((retval & DM_RET_ACTION_MASK) == DM_RET_NONE)
retval |= DM_RET_STOP;
return(retval);
}
static dev_match_ret
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
xptdevicematch(struct dev_match_pattern *patterns, u_int num_patterns,
struct cam_ed *device)
{
dev_match_ret retval;
u_int i;
retval = DM_RET_NONE;
/*
* If we aren't given something to match against, that's an error.
*/
if (device == NULL)
return(DM_RET_ERROR);
/*
* If there are no match entries, then this device matches no
* matter what.
*/
if ((patterns == NULL) || (num_patterns == 0))
return(DM_RET_DESCEND | DM_RET_COPY);
for (i = 0; i < num_patterns; i++) {
struct device_match_pattern *cur_pattern;
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
struct scsi_vpd_device_id *device_id_page;
/*
* If the pattern in question isn't for a device node, we
* aren't interested.
*/
if (patterns[i].type != DEV_MATCH_DEVICE) {
if ((patterns[i].type == DEV_MATCH_PERIPH)
&& ((retval & DM_RET_ACTION_MASK) == DM_RET_NONE))
retval |= DM_RET_DESCEND;
continue;
}
cur_pattern = &patterns[i].pattern.device_pattern;
/* Error out if mutually exclusive options are specified. */
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
if ((cur_pattern->flags & (DEV_MATCH_INQUIRY|DEV_MATCH_DEVID))
== (DEV_MATCH_INQUIRY|DEV_MATCH_DEVID))
return(DM_RET_ERROR);
/*
* If they want to match any device node, we give them any
* device node.
*/
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
if (cur_pattern->flags == DEV_MATCH_ANY)
goto copy_dev_node;
/*
* Not sure why someone would do this...
*/
if (cur_pattern->flags == DEV_MATCH_NONE)
continue;
if (((cur_pattern->flags & DEV_MATCH_PATH) != 0)
&& (cur_pattern->path_id != device->target->bus->path_id))
continue;
if (((cur_pattern->flags & DEV_MATCH_TARGET) != 0)
&& (cur_pattern->target_id != device->target->target_id))
continue;
if (((cur_pattern->flags & DEV_MATCH_LUN) != 0)
&& (cur_pattern->target_lun != device->lun_id))
continue;
if (((cur_pattern->flags & DEV_MATCH_INQUIRY) != 0)
&& (cam_quirkmatch((caddr_t)&device->inq_data,
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
(caddr_t)&cur_pattern->data.inq_pat,
1, sizeof(cur_pattern->data.inq_pat),
scsi_static_inquiry_match) == NULL))
continue;
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
device_id_page = (struct scsi_vpd_device_id *)device->device_id;
if (((cur_pattern->flags & DEV_MATCH_DEVID) != 0)
&& (device->device_id_len < SVPD_DEVICE_ID_HDR_LEN
|| scsi_devid_match((uint8_t *)device_id_page->desc_list,
device->device_id_len
- SVPD_DEVICE_ID_HDR_LEN,
cur_pattern->data.devid_pat.id,
cur_pattern->data.devid_pat.id_len) != 0))
continue;
copy_dev_node:
/*
2008-01-02 01:45:31 +00:00
* If we get to this point, the user definitely wants
* information on this device. So tell the caller to copy
* the data out.
*/
retval |= DM_RET_COPY;
/*
* If the return action has been set to descend, then we
* know that we've already seen a peripheral matching
* expression, therefore we need to further descend the tree.
* This won't change by continuing around the loop, so we
* go ahead and return. If we haven't seen a peripheral
* matching expression, we keep going around the loop until
* we exhaust the matching expressions. We'll set the stop
* flag once we fall out of the loop.
*/
if ((retval & DM_RET_ACTION_MASK) == DM_RET_DESCEND)
return(retval);
}
/*
* If the return action hasn't been set to descend yet, that means
* we haven't seen any peripheral matching patterns. So tell the
* caller to stop descending the tree -- the user doesn't want to
* match against lower level tree elements.
*/
if ((retval & DM_RET_ACTION_MASK) == DM_RET_NONE)
retval |= DM_RET_STOP;
return(retval);
}
/*
* Match a single peripheral against any number of match patterns.
*/
static dev_match_ret
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
xptperiphmatch(struct dev_match_pattern *patterns, u_int num_patterns,
struct cam_periph *periph)
{
dev_match_ret retval;
u_int i;
/*
* If we aren't given something to match against, that's an error.
*/
if (periph == NULL)
return(DM_RET_ERROR);
/*
* If there are no match entries, then this peripheral matches no
* matter what.
*/
if ((patterns == NULL) || (num_patterns == 0))
return(DM_RET_STOP | DM_RET_COPY);
/*
* There aren't any nodes below a peripheral node, so there's no
* reason to descend the tree any further.
*/
retval = DM_RET_STOP;
for (i = 0; i < num_patterns; i++) {
struct periph_match_pattern *cur_pattern;
/*
* If the pattern in question isn't for a peripheral, we
* aren't interested.
*/
if (patterns[i].type != DEV_MATCH_PERIPH)
continue;
cur_pattern = &patterns[i].pattern.periph_pattern;
/*
* If they want to match on anything, then we will do so.
*/
if (cur_pattern->flags == PERIPH_MATCH_ANY) {
/* set the copy flag */
retval |= DM_RET_COPY;
/*
* We've already set the return action to stop,
* since there are no nodes below peripherals in
* the tree.
*/
return(retval);
}
/*
* Not sure why someone would do this...
*/
if (cur_pattern->flags == PERIPH_MATCH_NONE)
continue;
if (((cur_pattern->flags & PERIPH_MATCH_PATH) != 0)
&& (cur_pattern->path_id != periph->path->bus->path_id))
continue;
/*
* For the target and lun id's, we have to make sure the
* target and lun pointers aren't NULL. The xpt peripheral
* has a wildcard target and device.
*/
if (((cur_pattern->flags & PERIPH_MATCH_TARGET) != 0)
&& ((periph->path->target == NULL)
||(cur_pattern->target_id != periph->path->target->target_id)))
continue;
if (((cur_pattern->flags & PERIPH_MATCH_LUN) != 0)
&& ((periph->path->device == NULL)
|| (cur_pattern->target_lun != periph->path->device->lun_id)))
continue;
if (((cur_pattern->flags & PERIPH_MATCH_UNIT) != 0)
&& (cur_pattern->unit_number != periph->unit_number))
continue;
if (((cur_pattern->flags & PERIPH_MATCH_NAME) != 0)
&& (strncmp(cur_pattern->periph_name, periph->periph_name,
DEV_IDLEN) != 0))
continue;
/*
2008-01-02 01:45:31 +00:00
* If we get to this point, the user definitely wants
* information on this peripheral. So tell the caller to
* copy the data out.
*/
retval |= DM_RET_COPY;
/*
* The return action has already been set to stop, since
* peripherals don't have any nodes below them in the EDT.
*/
return(retval);
}
/*
* If we get to this point, the peripheral that was passed in
* doesn't match any of the patterns.
*/
return(retval);
}
static int
xptedtbusfunc(struct cam_eb *bus, void *arg)
{
struct ccb_dev_match *cdm;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_et *target;
dev_match_ret retval;
cdm = (struct ccb_dev_match *)arg;
/*
* If our position is for something deeper in the tree, that means
* that we've already seen this node. So, we keep going down.
*/
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
&& (cdm->pos.cookie.bus == bus)
&& (cdm->pos.position_type & CAM_DEV_POS_TARGET)
&& (cdm->pos.cookie.target != NULL))
retval = DM_RET_DESCEND;
else
retval = xptbusmatch(cdm->patterns, cdm->num_patterns, bus);
/*
* If we got an error, bail out of the search.
*/
if ((retval & DM_RET_ACTION_MASK) == DM_RET_ERROR) {
cdm->status = CAM_DEV_MATCH_ERROR;
return(0);
}
/*
* If the copy flag is set, copy this bus out.
*/
if (retval & DM_RET_COPY) {
int spaceleft, j;
spaceleft = cdm->match_buf_len - (cdm->num_matches *
sizeof(struct dev_match_result));
/*
* If we don't have enough space to put in another
* match result, save our position and tell the
* user there are more devices to check.
*/
if (spaceleft < sizeof(struct dev_match_result)) {
bzero(&cdm->pos, sizeof(cdm->pos));
2008-01-02 01:45:31 +00:00
cdm->pos.position_type =
CAM_DEV_POS_EDT | CAM_DEV_POS_BUS;
cdm->pos.cookie.bus = bus;
cdm->pos.generations[CAM_BUS_GENERATION]=
xsoftc.bus_generation;
cdm->status = CAM_DEV_MATCH_MORE;
return(0);
}
j = cdm->num_matches;
cdm->num_matches++;
cdm->matches[j].type = DEV_MATCH_BUS;
cdm->matches[j].result.bus_result.path_id = bus->path_id;
cdm->matches[j].result.bus_result.bus_id = bus->sim->bus_id;
cdm->matches[j].result.bus_result.unit_number =
bus->sim->unit_number;
strlcpy(cdm->matches[j].result.bus_result.dev_name,
bus->sim->sim_name,
sizeof(cdm->matches[j].result.bus_result.dev_name));
}
/*
* If the user is only interested in buses, there's no
* reason to descend to the next level in the tree.
*/
if ((retval & DM_RET_ACTION_MASK) == DM_RET_STOP)
return(1);
/*
* If there is a target generation recorded, check it to
* make sure the target list hasn't changed.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&bus->eb_mtx);
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
&& (cdm->pos.cookie.bus == bus)
&& (cdm->pos.position_type & CAM_DEV_POS_TARGET)
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
&& (cdm->pos.cookie.target != NULL)) {
if ((cdm->pos.generations[CAM_TARGET_GENERATION] !=
bus->generation)) {
mtx_unlock(&bus->eb_mtx);
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return (0);
}
target = (struct cam_et *)cdm->pos.cookie.target;
target->refcount++;
} else
target = NULL;
mtx_unlock(&bus->eb_mtx);
return (xpttargettraverse(bus, target, xptedttargetfunc, arg));
}
static int
xptedttargetfunc(struct cam_et *target, void *arg)
{
struct ccb_dev_match *cdm;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_eb *bus;
struct cam_ed *device;
cdm = (struct ccb_dev_match *)arg;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
bus = target->bus;
/*
* If there is a device list generation recorded, check it to
* make sure the device list hasn't changed.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&bus->eb_mtx);
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
&& (cdm->pos.cookie.bus == bus)
&& (cdm->pos.position_type & CAM_DEV_POS_TARGET)
&& (cdm->pos.cookie.target == target)
&& (cdm->pos.position_type & CAM_DEV_POS_DEVICE)
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
&& (cdm->pos.cookie.device != NULL)) {
if (cdm->pos.generations[CAM_DEV_GENERATION] !=
target->generation) {
mtx_unlock(&bus->eb_mtx);
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return(0);
}
device = (struct cam_ed *)cdm->pos.cookie.device;
device->refcount++;
} else
device = NULL;
mtx_unlock(&bus->eb_mtx);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
return (xptdevicetraverse(target, device, xptedtdevicefunc, arg));
}
static int
xptedtdevicefunc(struct cam_ed *device, void *arg)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_eb *bus;
struct cam_periph *periph;
struct ccb_dev_match *cdm;
dev_match_ret retval;
cdm = (struct ccb_dev_match *)arg;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
bus = device->target->bus;
/*
* If our position is for something deeper in the tree, that means
* that we've already seen this node. So, we keep going down.
*/
if ((cdm->pos.position_type & CAM_DEV_POS_DEVICE)
&& (cdm->pos.cookie.device == device)
&& (cdm->pos.position_type & CAM_DEV_POS_PERIPH)
&& (cdm->pos.cookie.periph != NULL))
retval = DM_RET_DESCEND;
else
retval = xptdevicematch(cdm->patterns, cdm->num_patterns,
device);
if ((retval & DM_RET_ACTION_MASK) == DM_RET_ERROR) {
cdm->status = CAM_DEV_MATCH_ERROR;
return(0);
}
/*
* If the copy flag is set, copy this device out.
*/
if (retval & DM_RET_COPY) {
int spaceleft, j;
spaceleft = cdm->match_buf_len - (cdm->num_matches *
sizeof(struct dev_match_result));
/*
* If we don't have enough space to put in another
* match result, save our position and tell the
* user there are more devices to check.
*/
if (spaceleft < sizeof(struct dev_match_result)) {
bzero(&cdm->pos, sizeof(cdm->pos));
2008-01-02 01:45:31 +00:00
cdm->pos.position_type =
CAM_DEV_POS_EDT | CAM_DEV_POS_BUS |
CAM_DEV_POS_TARGET | CAM_DEV_POS_DEVICE;
cdm->pos.cookie.bus = device->target->bus;
cdm->pos.generations[CAM_BUS_GENERATION]=
xsoftc.bus_generation;
cdm->pos.cookie.target = device->target;
cdm->pos.generations[CAM_TARGET_GENERATION] =
device->target->bus->generation;
cdm->pos.cookie.device = device;
2008-01-02 01:45:31 +00:00
cdm->pos.generations[CAM_DEV_GENERATION] =
device->target->generation;
cdm->status = CAM_DEV_MATCH_MORE;
return(0);
}
j = cdm->num_matches;
cdm->num_matches++;
cdm->matches[j].type = DEV_MATCH_DEVICE;
cdm->matches[j].result.device_result.path_id =
device->target->bus->path_id;
cdm->matches[j].result.device_result.target_id =
device->target->target_id;
cdm->matches[j].result.device_result.target_lun =
device->lun_id;
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
cdm->matches[j].result.device_result.protocol =
device->protocol;
bcopy(&device->inq_data,
&cdm->matches[j].result.device_result.inq_data,
sizeof(struct scsi_inquiry_data));
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
bcopy(&device->ident_data,
&cdm->matches[j].result.device_result.ident_data,
sizeof(struct ata_params));
Add a number of interrelated CAM feature enhancements and bug fixes. NOTE: These changes will require recompilation of any userland applications, like cdrecord, xmcd, etc., that use the CAM passthrough interface. A make world is recommended. camcontrol.[c8]: - We now support two new commands, "tags" and "negotiate". - The tags commands allows users to view the number of tagged openings for a device as well as a number of other related parameters, and it allows users to set tagged openings for a device. - The negotiate command allows users to enable and disable disconnection and tagged queueing, set sync rates, offsets and bus width. Note that not all of those features are available for all controllers. Only the adv, ahc, and ncr drivers fully support all of the features at this point. Some cards do not allow the setting of sync rates, offsets and the like, and some of the drivers don't have any facilities to do so. Some drivers, like the adw driver, only support enabling or disabling sync negotiation, but do not support setting sync rates. - new description in the camcontrol man page of how to format a disk - cleanup of the camcontrol inquiry command - add support in the 'devlist' command for skipping unconfigured devices if -v was not specified on the command line. - make use of the new base_transfer_speed in the path inquiry CCB. - fix CCB bzero cases cam_xpt.c, cam_sim.[ch], cam_ccb.h: - new flags on many CCB function codes to designate whether they're non-immediate, use a user-supplied CCB, and can only be passed from userland programs via the xpt device. Use these flags in the transport layer and pass driver to categorize CCBs. - new flag in the transport layer device matching code for device nodes that indicates whether a device is unconfigured - bump the CAM version from 0x10 to 0x11 - Change the CAM ioctls to use the version as their group code, so we can force users to recompile code even when the CCB size doesn't change. - add + fill in a new value in the path inquiry CCB, base_transfer_speed. Remove a corresponding field from the cam_sim structure, and add code to every SIM to set this field to the proper value. - Fix the set transfer settings code in the transport layer. scsi_cd.c: - make some variables volatile instead of just casting them in various places - fix a race condition in the changer code - attach unless we get a "logical unit not supported" error. This should fix all of the cases where people have devices that return weird errors when they don't have media in the drive. scsi_da.c: - attach unless we get a "logical unit not supported" error scsi_pass.c: - for immediate CCBs, just malloc a CCB to send the user request in. This gets rid of the 'held' count problem in camcontrol tags. scsi_pass.h: - change the CAM ioctls to use the CAM version as their group code. adv driver: - Allow changing the sync rate and offset separately. adw driver - Allow changing the sync rate and offset separately. aha driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. ahc driver: - Allow setting offset and sync rate separately bt driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. NCR driver: - Fix the ultra/ultra 2 negotiation bug - allow setting both the sync rate and offset separately Other HBA drivers: - Put code in to set the base_transfer_speed field for XPT_GET_TRAN_SETTINGS CCBs. Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
/* Let the user know whether this device is unconfigured */
if (device->flags & CAM_DEV_UNCONFIGURED)
cdm->matches[j].result.device_result.flags =
DEV_RESULT_UNCONFIGURED;
else
cdm->matches[j].result.device_result.flags =
DEV_RESULT_NOFLAG;
}
/*
* If the user isn't interested in peripherals, don't descend
* the tree any further.
*/
if ((retval & DM_RET_ACTION_MASK) == DM_RET_STOP)
return(1);
/*
* If there is a peripheral list generation recorded, make sure
* it hasn't changed.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_lock_buses();
mtx_lock(&bus->eb_mtx);
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
&& (cdm->pos.cookie.bus == bus)
&& (cdm->pos.position_type & CAM_DEV_POS_TARGET)
&& (cdm->pos.cookie.target == device->target)
&& (cdm->pos.position_type & CAM_DEV_POS_DEVICE)
&& (cdm->pos.cookie.device == device)
&& (cdm->pos.position_type & CAM_DEV_POS_PERIPH)
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
&& (cdm->pos.cookie.periph != NULL)) {
if (cdm->pos.generations[CAM_PERIPH_GENERATION] !=
device->generation) {
mtx_unlock(&bus->eb_mtx);
xpt_unlock_buses();
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return(0);
}
periph = (struct cam_periph *)cdm->pos.cookie.periph;
periph->refcount++;
} else
periph = NULL;
mtx_unlock(&bus->eb_mtx);
xpt_unlock_buses();
return (xptperiphtraverse(device, periph, xptedtperiphfunc, arg));
}
static int
xptedtperiphfunc(struct cam_periph *periph, void *arg)
{
struct ccb_dev_match *cdm;
dev_match_ret retval;
cdm = (struct ccb_dev_match *)arg;
retval = xptperiphmatch(cdm->patterns, cdm->num_patterns, periph);
if ((retval & DM_RET_ACTION_MASK) == DM_RET_ERROR) {
cdm->status = CAM_DEV_MATCH_ERROR;
return(0);
}
/*
* If the copy flag is set, copy this peripheral out.
*/
if (retval & DM_RET_COPY) {
int spaceleft, j;
size_t l;
spaceleft = cdm->match_buf_len - (cdm->num_matches *
sizeof(struct dev_match_result));
/*
* If we don't have enough space to put in another
* match result, save our position and tell the
* user there are more devices to check.
*/
if (spaceleft < sizeof(struct dev_match_result)) {
bzero(&cdm->pos, sizeof(cdm->pos));
2008-01-02 01:45:31 +00:00
cdm->pos.position_type =
CAM_DEV_POS_EDT | CAM_DEV_POS_BUS |
CAM_DEV_POS_TARGET | CAM_DEV_POS_DEVICE |
CAM_DEV_POS_PERIPH;
cdm->pos.cookie.bus = periph->path->bus;
cdm->pos.generations[CAM_BUS_GENERATION]=
xsoftc.bus_generation;
cdm->pos.cookie.target = periph->path->target;
cdm->pos.generations[CAM_TARGET_GENERATION] =
periph->path->bus->generation;
cdm->pos.cookie.device = periph->path->device;
2008-01-02 01:45:31 +00:00
cdm->pos.generations[CAM_DEV_GENERATION] =
periph->path->target->generation;
cdm->pos.cookie.periph = periph;
cdm->pos.generations[CAM_PERIPH_GENERATION] =
periph->path->device->generation;
cdm->status = CAM_DEV_MATCH_MORE;
return(0);
}
j = cdm->num_matches;
cdm->num_matches++;
cdm->matches[j].type = DEV_MATCH_PERIPH;
cdm->matches[j].result.periph_result.path_id =
periph->path->bus->path_id;
cdm->matches[j].result.periph_result.target_id =
periph->path->target->target_id;
cdm->matches[j].result.periph_result.target_lun =
periph->path->device->lun_id;
cdm->matches[j].result.periph_result.unit_number =
periph->unit_number;
l = sizeof(cdm->matches[j].result.periph_result.periph_name);
strlcpy(cdm->matches[j].result.periph_result.periph_name,
periph->periph_name, l);
}
return(1);
}
static int
xptedtmatch(struct ccb_dev_match *cdm)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_eb *bus;
int ret;
cdm->num_matches = 0;
/*
* Check the bus list generation. If it has changed, the user
* needs to reset everything and start over.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_lock_buses();
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
&& (cdm->pos.cookie.bus != NULL)) {
if (cdm->pos.generations[CAM_BUS_GENERATION] !=
xsoftc.bus_generation) {
xpt_unlock_buses();
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return(0);
}
bus = (struct cam_eb *)cdm->pos.cookie.bus;
bus->refcount++;
} else
bus = NULL;
xpt_unlock_buses();
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
ret = xptbustraverse(bus, xptedtbusfunc, cdm);
/*
* If we get back 0, that means that we had to stop before fully
* traversing the EDT. It also means that one of the subroutines
* has set the status field to the proper value. If we get back 1,
* we've fully traversed the EDT and copied out any matching entries.
*/
if (ret == 1)
cdm->status = CAM_DEV_MATCH_LAST;
return(ret);
}
static int
xptplistpdrvfunc(struct periph_driver **pdrv, void *arg)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_periph *periph;
struct ccb_dev_match *cdm;
cdm = (struct ccb_dev_match *)arg;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_lock_buses();
if ((cdm->pos.position_type & CAM_DEV_POS_PDPTR)
&& (cdm->pos.cookie.pdrv == pdrv)
&& (cdm->pos.position_type & CAM_DEV_POS_PERIPH)
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
&& (cdm->pos.cookie.periph != NULL)) {
if (cdm->pos.generations[CAM_PERIPH_GENERATION] !=
(*pdrv)->generation) {
xpt_unlock_buses();
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return(0);
}
periph = (struct cam_periph *)cdm->pos.cookie.periph;
periph->refcount++;
} else
periph = NULL;
xpt_unlock_buses();
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
return (xptpdperiphtraverse(pdrv, periph, xptplistperiphfunc, arg));
}
static int
xptplistperiphfunc(struct cam_periph *periph, void *arg)
{
struct ccb_dev_match *cdm;
dev_match_ret retval;
cdm = (struct ccb_dev_match *)arg;
retval = xptperiphmatch(cdm->patterns, cdm->num_patterns, periph);
if ((retval & DM_RET_ACTION_MASK) == DM_RET_ERROR) {
cdm->status = CAM_DEV_MATCH_ERROR;
return(0);
}
/*
* If the copy flag is set, copy this peripheral out.
*/
if (retval & DM_RET_COPY) {
int spaceleft, j;
size_t l;
spaceleft = cdm->match_buf_len - (cdm->num_matches *
sizeof(struct dev_match_result));
/*
* If we don't have enough space to put in another
* match result, save our position and tell the
* user there are more devices to check.
*/
if (spaceleft < sizeof(struct dev_match_result)) {
struct periph_driver **pdrv;
pdrv = NULL;
bzero(&cdm->pos, sizeof(cdm->pos));
2008-01-02 01:45:31 +00:00
cdm->pos.position_type =
CAM_DEV_POS_PDRV | CAM_DEV_POS_PDPTR |
CAM_DEV_POS_PERIPH;
/*
* This may look a bit non-sensical, but it is
* actually quite logical. There are very few
* peripheral drivers, and bloating every peripheral
* structure with a pointer back to its parent
* peripheral driver linker set entry would cost
* more in the long run than doing this quick lookup.
*/
for (pdrv = periph_drivers; *pdrv != NULL; pdrv++) {
if (strcmp((*pdrv)->driver_name,
periph->periph_name) == 0)
break;
}
if (*pdrv == NULL) {
cdm->status = CAM_DEV_MATCH_ERROR;
return(0);
}
cdm->pos.cookie.pdrv = pdrv;
/*
* The periph generation slot does double duty, as
* does the periph pointer slot. They are used for
* both edt and pdrv lookups and positioning.
*/
cdm->pos.cookie.periph = periph;
cdm->pos.generations[CAM_PERIPH_GENERATION] =
(*pdrv)->generation;
cdm->status = CAM_DEV_MATCH_MORE;
return(0);
}
j = cdm->num_matches;
cdm->num_matches++;
cdm->matches[j].type = DEV_MATCH_PERIPH;
cdm->matches[j].result.periph_result.path_id =
periph->path->bus->path_id;
/*
* The transport layer peripheral doesn't have a target or
* lun.
*/
if (periph->path->target)
cdm->matches[j].result.periph_result.target_id =
periph->path->target->target_id;
else
cdm->matches[j].result.periph_result.target_id =
CAM_TARGET_WILDCARD;
if (periph->path->device)
cdm->matches[j].result.periph_result.target_lun =
periph->path->device->lun_id;
else
cdm->matches[j].result.periph_result.target_lun =
CAM_LUN_WILDCARD;
cdm->matches[j].result.periph_result.unit_number =
periph->unit_number;
l = sizeof(cdm->matches[j].result.periph_result.periph_name);
strlcpy(cdm->matches[j].result.periph_result.periph_name,
periph->periph_name, l);
}
return(1);
}
static int
xptperiphlistmatch(struct ccb_dev_match *cdm)
{
int ret;
cdm->num_matches = 0;
/*
* At this point in the edt traversal function, we check the bus
* list generation to make sure that no buses have been added or
* removed since the user last sent a XPT_DEV_MATCH ccb through.
* For the peripheral driver list traversal function, however, we
* don't have to worry about new peripheral driver types coming or
* going; they're in a linker set, and therefore can't change
* without a recompile.
*/
if ((cdm->pos.position_type & CAM_DEV_POS_PDPTR)
&& (cdm->pos.cookie.pdrv != NULL))
ret = xptpdrvtraverse(
(struct periph_driver **)cdm->pos.cookie.pdrv,
xptplistpdrvfunc, cdm);
else
ret = xptpdrvtraverse(NULL, xptplistpdrvfunc, cdm);
/*
* If we get back 0, that means that we had to stop before fully
* traversing the peripheral driver tree. It also means that one of
* the subroutines has set the status field to the proper value. If
* we get back 1, we've fully traversed the EDT and copied out any
* matching entries.
*/
if (ret == 1)
cdm->status = CAM_DEV_MATCH_LAST;
return(ret);
}
static int
xptbustraverse(struct cam_eb *start_bus, xpt_busfunc_t *tr_func, void *arg)
{
struct cam_eb *bus, *next_bus;
int retval;
retval = 1;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (start_bus)
bus = start_bus;
else {
xpt_lock_buses();
bus = TAILQ_FIRST(&xsoftc.xpt_busses);
if (bus == NULL) {
xpt_unlock_buses();
return (retval);
}
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
bus->refcount++;
xpt_unlock_buses();
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
for (; bus != NULL; bus = next_bus) {
retval = tr_func(bus, arg);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (retval == 0) {
xpt_release_bus(bus);
break;
}
xpt_lock_buses();
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
next_bus = TAILQ_NEXT(bus, links);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (next_bus)
next_bus->refcount++;
xpt_unlock_buses();
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
xpt_release_bus(bus);
}
return(retval);
}
static int
xpttargettraverse(struct cam_eb *bus, struct cam_et *start_target,
xpt_targetfunc_t *tr_func, void *arg)
{
struct cam_et *target, *next_target;
int retval;
retval = 1;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (start_target)
target = start_target;
else {
mtx_lock(&bus->eb_mtx);
target = TAILQ_FIRST(&bus->et_entries);
if (target == NULL) {
mtx_unlock(&bus->eb_mtx);
return (retval);
}
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
target->refcount++;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&bus->eb_mtx);
}
for (; target != NULL; target = next_target) {
retval = tr_func(target, arg);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (retval == 0) {
xpt_release_target(target);
break;
}
mtx_lock(&bus->eb_mtx);
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
next_target = TAILQ_NEXT(target, links);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (next_target)
next_target->refcount++;
mtx_unlock(&bus->eb_mtx);
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
xpt_release_target(target);
}
return(retval);
}
static int
xptdevicetraverse(struct cam_et *target, struct cam_ed *start_device,
xpt_devicefunc_t *tr_func, void *arg)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_eb *bus;
struct cam_ed *device, *next_device;
int retval;
retval = 1;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
bus = target->bus;
if (start_device)
device = start_device;
else {
mtx_lock(&bus->eb_mtx);
device = TAILQ_FIRST(&target->ed_entries);
if (device == NULL) {
mtx_unlock(&bus->eb_mtx);
return (retval);
}
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
device->refcount++;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&bus->eb_mtx);
}
for (; device != NULL; device = next_device) {
mtx_lock(&device->device_mtx);
retval = tr_func(device, arg);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&device->device_mtx);
if (retval == 0) {
xpt_release_device(device);
break;
}
mtx_lock(&bus->eb_mtx);
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
next_device = TAILQ_NEXT(device, links);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (next_device)
next_device->refcount++;
mtx_unlock(&bus->eb_mtx);
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
xpt_release_device(device);
}
return(retval);
}
static int
xptperiphtraverse(struct cam_ed *device, struct cam_periph *start_periph,
xpt_periphfunc_t *tr_func, void *arg)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_eb *bus;
struct cam_periph *periph, *next_periph;
int retval;
retval = 1;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
bus = device->target->bus;
if (start_periph)
periph = start_periph;
else {
xpt_lock_buses();
mtx_lock(&bus->eb_mtx);
periph = SLIST_FIRST(&device->periphs);
while (periph != NULL && (periph->flags & CAM_PERIPH_FREE) != 0)
periph = SLIST_NEXT(periph, periph_links);
if (periph == NULL) {
mtx_unlock(&bus->eb_mtx);
xpt_unlock_buses();
return (retval);
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
}
periph->refcount++;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&bus->eb_mtx);
xpt_unlock_buses();
}
for (; periph != NULL; periph = next_periph) {
retval = tr_func(periph, arg);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (retval == 0) {
cam_periph_release_locked(periph);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
break;
}
xpt_lock_buses();
mtx_lock(&bus->eb_mtx);
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
next_periph = SLIST_NEXT(periph, periph_links);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
while (next_periph != NULL &&
(next_periph->flags & CAM_PERIPH_FREE) != 0)
next_periph = SLIST_NEXT(next_periph, periph_links);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (next_periph)
next_periph->refcount++;
mtx_unlock(&bus->eb_mtx);
xpt_unlock_buses();
cam_periph_release_locked(periph);
}
return(retval);
}
static int
xptpdrvtraverse(struct periph_driver **start_pdrv,
xpt_pdrvfunc_t *tr_func, void *arg)
{
struct periph_driver **pdrv;
int retval;
retval = 1;
/*
* We don't traverse the peripheral driver list like we do the
* other lists, because it is a linker set, and therefore cannot be
* changed during runtime. If the peripheral driver list is ever
* re-done to be something other than a linker set (i.e. it can
* change while the system is running), the list traversal should
* be modified to work like the other traversal functions.
*/
for (pdrv = (start_pdrv ? start_pdrv : periph_drivers);
*pdrv != NULL; pdrv++) {
retval = tr_func(pdrv, arg);
if (retval == 0)
return(retval);
}
return(retval);
}
static int
xptpdperiphtraverse(struct periph_driver **pdrv,
struct cam_periph *start_periph,
xpt_periphfunc_t *tr_func, void *arg)
{
struct cam_periph *periph, *next_periph;
int retval;
retval = 1;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (start_periph)
periph = start_periph;
else {
xpt_lock_buses();
periph = TAILQ_FIRST(&(*pdrv)->units);
while (periph != NULL && (periph->flags & CAM_PERIPH_FREE) != 0)
periph = TAILQ_NEXT(periph, unit_links);
if (periph == NULL) {
xpt_unlock_buses();
return (retval);
}
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
periph->refcount++;
xpt_unlock_buses();
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
for (; periph != NULL; periph = next_periph) {
cam_periph_lock(periph);
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
retval = tr_func(periph, arg);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
cam_periph_unlock(periph);
if (retval == 0) {
cam_periph_release(periph);
break;
}
xpt_lock_buses();
Fix a race condition in CAM peripheral free handling, locking in the CAM XPT bus traversal code, and a number of other periph level issues. cam_periph.h, cam_periph.c: Modify cam_periph_acquire() to test the CAM_PERIPH_INVALID flag prior to allowing a reference count to be gained on a peripheral. Callers of this function will receive CAM_REQ_CMP_ERR status in the situation of attempting to reference an invalidated periph. This guarantees that a peripheral scheduled for a deferred free will not be accessed during its wait for destruction. Panic during attempts to drop a reference count on a peripheral that already has a zero reference count. In cam_periph_list(), use a local sbuf with SBUF_FIXEDLEN set so that mallocs do not occur while the xpt topology lock is held, regardless of the allocation policy of the passed in sbuf. Add a new routine, cam_periph_release_locked_buses(), that can be called when the caller already holds the CAM topology lock. Add some extra debugging for duplicate peripheral allocations in cam_periph_alloc(). Treat CAM_DEV_NOT_THERE much the same as a selection timeout (AC_LOST_DEVICE is emitted), but forgo retries. cam_xpt.c: Revamp the way the EDT traversal code does locking and reference counting. This was broken, since it assumed that the EDT would not change during traversal, but that assumption is no longer valid. So, to prevent devices from going away while we traverse the EDT, make sure we properly lock everything and hold references on devices that we are using. The two peripheral driver traversal routines should be examined. xptpdperiphtraverse() holds the topology lock for the entire time it runs. xptperiphtraverse() is now locked properly, but only holds the topology lock while it is traversing the list, and not while the traversal function is running. The bus locking code in xptbustraverse() should also be revisited at a later time, since it is complex and should probably be simplified. scsi_da.c: Pay attention to the return value from cam_periph_acquire(). Return 0 always from daclose() even if the disk is now gone. Add some rudimentary error injection support. scsi_sg.c: Fix reference counting in the sg(4) driver. The sg driver was calling cam_periph_release() on close, but never called cam_periph_acquire() (which increments the reference count) on open. The periph code correctly complained that the sg(4) driver was trying to decrement the refcount when it was already 0. Sponsored by: Spectra Logic MFC after: 2 weeks
2012-01-12 00:41:48 +00:00
next_periph = TAILQ_NEXT(periph, unit_links);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
while (next_periph != NULL &&
(next_periph->flags & CAM_PERIPH_FREE) != 0)
next_periph = TAILQ_NEXT(next_periph, unit_links);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (next_periph)
next_periph->refcount++;
xpt_unlock_buses();
cam_periph_release(periph);
}
return(retval);
}
static int
xptdefbusfunc(struct cam_eb *bus, void *arg)
{
struct xpt_traverse_config *tr_config;
tr_config = (struct xpt_traverse_config *)arg;
if (tr_config->depth == XPT_DEPTH_BUS) {
xpt_busfunc_t *tr_func;
tr_func = (xpt_busfunc_t *)tr_config->tr_func;
return(tr_func(bus, tr_config->tr_arg));
} else
return(xpttargettraverse(bus, NULL, xptdeftargetfunc, arg));
}
static int
xptdeftargetfunc(struct cam_et *target, void *arg)
{
struct xpt_traverse_config *tr_config;
tr_config = (struct xpt_traverse_config *)arg;
if (tr_config->depth == XPT_DEPTH_TARGET) {
xpt_targetfunc_t *tr_func;
tr_func = (xpt_targetfunc_t *)tr_config->tr_func;
return(tr_func(target, tr_config->tr_arg));
} else
return(xptdevicetraverse(target, NULL, xptdefdevicefunc, arg));
}
static int
xptdefdevicefunc(struct cam_ed *device, void *arg)
{
struct xpt_traverse_config *tr_config;
tr_config = (struct xpt_traverse_config *)arg;
if (tr_config->depth == XPT_DEPTH_DEVICE) {
xpt_devicefunc_t *tr_func;
tr_func = (xpt_devicefunc_t *)tr_config->tr_func;
return(tr_func(device, tr_config->tr_arg));
} else
return(xptperiphtraverse(device, NULL, xptdefperiphfunc, arg));
}
static int
xptdefperiphfunc(struct cam_periph *periph, void *arg)
{
struct xpt_traverse_config *tr_config;
xpt_periphfunc_t *tr_func;
tr_config = (struct xpt_traverse_config *)arg;
tr_func = (xpt_periphfunc_t *)tr_config->tr_func;
/*
* Unlike the other default functions, we don't check for depth
* here. The peripheral driver level is the last level in the EDT,
* so if we're here, we should execute the function in question.
*/
return(tr_func(periph, tr_config->tr_arg));
}
/*
* Execute the given function for every bus in the EDT.
*/
static int
xpt_for_all_busses(xpt_busfunc_t *tr_func, void *arg)
{
struct xpt_traverse_config tr_config;
tr_config.depth = XPT_DEPTH_BUS;
tr_config.tr_func = tr_func;
tr_config.tr_arg = arg;
return(xptbustraverse(NULL, xptdefbusfunc, &tr_config));
}
/*
* Execute the given function for every device in the EDT.
*/
static int
xpt_for_all_devices(xpt_devicefunc_t *tr_func, void *arg)
{
struct xpt_traverse_config tr_config;
tr_config.depth = XPT_DEPTH_DEVICE;
tr_config.tr_func = tr_func;
tr_config.tr_arg = arg;
return(xptbustraverse(NULL, xptdefbusfunc, &tr_config));
}
static int
xptsetasyncfunc(struct cam_ed *device, void *arg)
{
struct cam_path path;
struct ccb_getdev cgd;
struct ccb_setasync *csa = (struct ccb_setasync *)arg;
/*
* Don't report unconfigured devices (Wildcard devs,
* devices only for target mode, device instances
* that have been invalidated but are waiting for
* their last reference count to be released).
*/
if ((device->flags & CAM_DEV_UNCONFIGURED) != 0)
return (1);
xpt_compile_path(&path,
NULL,
device->target->bus->path_id,
device->target->target_id,
device->lun_id);
xpt_setup_ccb(&cgd.ccb_h, &path, CAM_PRIORITY_NORMAL);
cgd.ccb_h.func_code = XPT_GDEV_TYPE;
xpt_action((union ccb *)&cgd);
csa->callback(csa->callback_arg,
AC_FOUND_DEVICE,
&path, &cgd);
xpt_release_path(&path);
return(1);
}
static int
xptsetasyncbusfunc(struct cam_eb *bus, void *arg)
{
struct cam_path path;
struct ccb_pathinq cpi;
struct ccb_setasync *csa = (struct ccb_setasync *)arg;
xpt_compile_path(&path, /*periph*/NULL,
bus->path_id,
CAM_TARGET_WILDCARD,
CAM_LUN_WILDCARD);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_path_lock(&path);
xpt_path_inq(&cpi, &path);
csa->callback(csa->callback_arg,
AC_PATH_REGISTERED,
&path, &cpi);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_path_unlock(&path);
xpt_release_path(&path);
return(1);
}
void
xpt_action(union ccb *start_ccb)
{
CAM_DEBUG(start_ccb->ccb_h.path, CAM_DEBUG_TRACE,
("xpt_action: func %#x %s\n", start_ccb->ccb_h.func_code,
xpt_action_name(start_ccb->ccb_h.func_code)));
start_ccb->ccb_h.status = CAM_REQ_INPROG;
(*(start_ccb->ccb_h.path->bus->xport->ops->action))(start_ccb);
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
}
void
xpt_action_default(union ccb *start_ccb)
{
struct cam_path *path;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_sim *sim;
struct mtx *mtx;
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
path = start_ccb->ccb_h.path;
CAM_DEBUG(path, CAM_DEBUG_TRACE,
("xpt_action_default: func %#x %s\n", start_ccb->ccb_h.func_code,
xpt_action_name(start_ccb->ccb_h.func_code)));
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
switch (start_ccb->ccb_h.func_code) {
case XPT_SCSI_IO:
{
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
struct cam_ed *device;
/*
* For the sake of compatibility with SCSI-1
* devices that may not understand the identify
* message, we include lun information in the
* second byte of all commands. SCSI-1 specifies
* that luns are a 3 bit value and reserves only 3
* bits for lun information in the CDB. Later
* revisions of the SCSI spec allow for more than 8
* luns, but have deprecated lun information in the
* CDB. So, if the lun won't fit, we must omit.
*
* Also be aware that during initial probing for devices,
* the inquiry information is unknown but initialized to 0.
* This means that this code will be exercised while probing
* devices with an ANSI revision greater than 2.
*/
device = path->device;
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
if (device->protocol_version <= SCSI_REV_2
&& start_ccb->ccb_h.target_lun < 8
&& (start_ccb->ccb_h.flags & CAM_CDB_POINTER) == 0) {
start_ccb->csio.cdb_io.cdb_bytes[1] |=
start_ccb->ccb_h.target_lun << 5;
}
start_ccb->csio.scsi_status = SCSI_STATUS_OK;
}
/* FALLTHROUGH */
case XPT_TARGET_IO:
case XPT_CONT_TARGET_IO:
start_ccb->csio.sense_resid = 0;
start_ccb->csio.resid = 0;
/* FALLTHROUGH */
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
case XPT_ATA_IO:
if (start_ccb->ccb_h.func_code == XPT_ATA_IO)
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
start_ccb->ataio.resid = 0;
/* FALLTHROUGH */
case XPT_NVME_IO:
case XPT_NVME_ADMIN:
case XPT_MMC_IO:
case XPT_RESET_DEV:
case XPT_ENG_EXEC:
Add Serial Management Protocol (SMP) passthrough support to CAM. This includes support in the kernel, camcontrol(8), libcam and the mps(4) driver for SMP passthrough. The CAM SCSI probe code has been modified to fetch Inquiry VPD page 0x00 to determine supported pages, and will now fetch page 0x83 in addition to page 0x80 if supported. Add two new CAM CCBs, XPT_SMP_IO, and XPT_GDEV_ADVINFO. The SMP CCB is intended for SMP requests and responses. The ADVINFO is currently used to fetch cached VPD page 0x83 data from the transport layer, but is intended to be extensible to fetch other types of device-specific data. SMP-only devices are not currently represented in the CAM topology, and so the current semantics are that the SIM will route SMP CCBs to either the addressed device, if it contains an SMP target, or its parent, if it contains an SMP target. (This is noted in cam_ccb.h, since it will change later once we have the ability to have SMP-only devices in CAM's topology.) smp_all.c, smp_all.h: New helper routines for SMP. This includes SMP request building routines, response parsing routines, error decoding routines, and structure definitions for a number of SMP commands. libcam/Makefile: Add smp_all.c to libcam, so that SMP functionality is available to userland applications. camcontrol.8, camcontrol.c: Add smp passthrough support to camcontrol. Several new subcommands are now available: 'smpcmd' functions much like 'cmd', except that it allows the user to send generic SMP commands. 'smprg' sends the SMP report general command, and displays the decoded output. It will automatically fetch extended output if it is available. 'smppc' sends the SMP phy control command, with any number of potential options. Among other things, this allows the user to reset a phy on a SAS expander, or disable a phy on an expander. 'smpmaninfo' sends the SMP report manufacturer information and displays the decoded output. 'smpphylist' displays a list of phys on an expander, and the CAM devices attached to those phys, if any. cam.h, cam.c: Add a status value for SMP errors (CAM_SMP_STATUS_ERROR). Add a missing description for CAM_SCSI_IT_NEXUS_LOST. Add support for SMP commands to cam_error_string(). cam_ccb.h: Rename the CAM_DIR_RESV flag to CAM_DIR_BOTH. SMP commands are by nature bi-directional, and we may need to support bi-directional SCSI commands later. Add the XPT_SMP_IO CCB. Since SMP commands are bi-directional, there are pointers for both the request and response. Add a fill routine for SMP CCBs. Add the XPT_GDEV_ADVINFO CCB. This is currently used to fetch cached page 0x83 data from the transport later, but is extensible to fetch many other types of data. cam_periph.c: Add support in cam_periph_mapmem() for XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. cam_xpt.c: Add support for executing XPT_SMP_IO CCBs. cam_xpt_internal.h: Add fields for VPD pages 0x00 and 0x83 in struct cam_ed. scsi_all.c: Add scsi_get_sas_addr(), a function that parses VPD page 0x83 data and pulls out a SAS address. scsi_all.h: Add VPD page 0x00 and 0x83 structures, and a prototype for scsi_get_sas_addr(). scsi_pass.c: Add support for mapping buffers in XPT_SMP_IO and XPT_GDEV_ADVINFO CCBs. scsi_xpt.c: In the SCSI probe code, first ask the device for VPD page 0x00. If any VPD pages are supported, that page is required to be implemented. Based on the response, we may probe for the serial number (page 0x80) or device id (page 0x83). Add support for the XPT_GDEV_ADVINFO CCB. sys/conf/files: Add smp_all.c. mps.c: Add support for passing in a uio in mps_map_command(), so we can map a S/G list at once. Add support for SMP passthrough commands in mps_data_cb(). SMP is a special case, because the first buffer in the S/G list is outbound and the second buffer is inbound. Add support for warning the user if the busdma code comes back with more buffers than will work for the command. This will, for example, help the user determine why an SMP command failed if busdma comes back with three buffers. mps_pci.c: Add sys/uio.h. mps_sas.c: Add the SAS address and the parent handle to the list of fields we pull from device page 0 and cache in struct mpssas_target. These are needed for SMP passthrough. Add support for the XPT_SMP_IO CCB. For now, this CCB is routed to the addressed device if it supports SMP, or to its parent if it does not and the parent does. This is necessary because CAM does not currently support SMP-only nodes in the topology. Make SMP passthrough support conditional on __FreeBSD_version >= 900026. This will make it easier to MFC this change to the driver without MFCing the CAM changes as well. mps_user.c: Un-staticize mpi_init_sge() so we can use it for the SMP passthrough code. mpsvar.h: Add a uio and iovecs into struct mps_command for SMP passthrough commands. Add a cm_max_segs field to struct mps_command so that we can warn the user if busdma comes back with too many segments. Clear the cm_reply when a command gets freed. If it is not cleared, reply frames will eventually get freed into the pool multiple times and corrupt the pool. (This fix is from scottl.) Add a prototype for mpi_init_sge(). sys/param.h: Bump __FreeBSD_version to 900026 for the for the inclusion of the XPT_GDEV_ADVINFO and XPT_SMP_IO CAM CCBs.
2010-11-30 22:39:46 +00:00
case XPT_SMP_IO:
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
{
struct cam_devq *devq;
devq = path->bus->sim->devq;
mtx_lock(&devq->send_mtx);
cam_ccbq_insert_ccb(&path->device->ccbq, start_ccb);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (xpt_schedule_devq(devq, path->device) != 0)
xpt_run_devq(devq);
mtx_unlock(&devq->send_mtx);
break;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
case XPT_CALC_GEOMETRY:
/* Filter out garbage */
if (start_ccb->ccg.block_size == 0
|| start_ccb->ccg.volume_size == 0) {
start_ccb->ccg.cylinders = 0;
start_ccb->ccg.heads = 0;
start_ccb->ccg.secs_per_track = 0;
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
goto call_sim;
case XPT_ABORT:
{
union ccb* abort_ccb;
abort_ccb = start_ccb->cab.abort_ccb;
if (XPT_FC_IS_DEV_QUEUED(abort_ccb)) {
struct cam_ed *device;
struct cam_devq *devq;
device = abort_ccb->ccb_h.path->device;
devq = device->sim->devq;
mtx_lock(&devq->send_mtx);
if (abort_ccb->ccb_h.pinfo.index > 0) {
cam_ccbq_remove_ccb(&device->ccbq, abort_ccb);
abort_ccb->ccb_h.status =
CAM_REQ_ABORTED|CAM_DEV_QFRZN;
xpt_freeze_devq_device(device, 1);
mtx_unlock(&devq->send_mtx);
xpt_done(abort_ccb);
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
mtx_unlock(&devq->send_mtx);
if (abort_ccb->ccb_h.pinfo.index == CAM_UNQUEUED_INDEX
&& (abort_ccb->ccb_h.status & CAM_SIM_QUEUED) == 0) {
/*
* We've caught this ccb en route to
* the SIM. Flag it for abort and the
* SIM will do so just before starting
* real work on the CCB.
*/
abort_ccb->ccb_h.status =
CAM_REQ_ABORTED|CAM_DEV_QFRZN;
xpt_freeze_devq(abort_ccb->ccb_h.path, 1);
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
2008-01-02 01:45:31 +00:00
}
if (XPT_FC_IS_QUEUED(abort_ccb)
&& (abort_ccb->ccb_h.pinfo.index == CAM_DONEQ_INDEX)) {
/*
* It's already completed but waiting
* for our SWI to get to it.
*/
start_ccb->ccb_h.status = CAM_UA_ABORT;
break;
}
/*
* If we weren't able to take care of the abort request
* in the XPT, pass the request down to the SIM for processing.
*/
}
/* FALLTHROUGH */
case XPT_ACCEPT_TARGET_IO:
case XPT_EN_LUN:
case XPT_IMMED_NOTIFY:
case XPT_NOTIFY_ACK:
case XPT_RESET_BUS:
case XPT_IMMEDIATE_NOTIFY:
case XPT_NOTIFY_ACKNOWLEDGE:
case XPT_GET_SIM_KNOB_OLD:
case XPT_GET_SIM_KNOB:
case XPT_SET_SIM_KNOB:
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
case XPT_GET_TRAN_SETTINGS:
case XPT_SET_TRAN_SETTINGS:
case XPT_PATH_INQ:
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
call_sim:
sim = path->bus->sim;
mtx = sim->mtx;
if (mtx && !mtx_owned(mtx))
mtx_lock(mtx);
else
mtx = NULL;
CAM_DEBUG(path, CAM_DEBUG_TRACE,
("Calling sim->sim_action(): func=%#x\n", start_ccb->ccb_h.func_code));
(*(sim->sim_action))(sim, start_ccb);
CAM_DEBUG(path, CAM_DEBUG_TRACE,
("sim->sim_action returned: status=%#x\n", start_ccb->ccb_h.status));
if (mtx)
mtx_unlock(mtx);
break;
case XPT_PATH_STATS:
start_ccb->cpis.last_reset = path->bus->last_reset;
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
case XPT_GDEV_TYPE:
{
struct cam_ed *dev;
dev = path->device;
if ((dev->flags & CAM_DEV_UNCONFIGURED) != 0) {
start_ccb->ccb_h.status = CAM_DEV_NOT_THERE;
} else {
struct ccb_getdev *cgd;
cgd = &start_ccb->cgd;
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
cgd->protocol = dev->protocol;
cgd->inq_data = dev->inq_data;
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
cgd->ident_data = dev->ident_data;
cgd->inq_flags = dev->inq_flags;
cgd->ccb_h.status = CAM_REQ_CMP;
cgd->serial_num_len = dev->serial_num_len;
if ((dev->serial_num_len > 0)
&& (dev->serial_num != NULL))
bcopy(dev->serial_num, cgd->serial_num,
dev->serial_num_len);
}
2008-01-02 01:45:31 +00:00
break;
}
case XPT_GDEV_STATS:
{
struct ccb_getdevstats *cgds = &start_ccb->cgds;
struct cam_ed *dev = path->device;
struct cam_eb *bus = path->bus;
struct cam_et *tar = path->target;
struct cam_devq *devq = bus->sim->devq;
mtx_lock(&devq->send_mtx);
cgds->dev_openings = dev->ccbq.dev_openings;
cgds->dev_active = dev->ccbq.dev_active;
cgds->allocated = dev->ccbq.allocated;
cgds->queued = cam_ccbq_pending_ccb_count(&dev->ccbq);
cgds->held = cgds->allocated - cgds->dev_active - cgds->queued;
cgds->last_reset = tar->last_reset;
cgds->maxtags = dev->maxtags;
cgds->mintags = dev->mintags;
if (timevalcmp(&tar->last_reset, &bus->last_reset, <))
cgds->last_reset = bus->last_reset;
mtx_unlock(&devq->send_mtx);
cgds->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_GDEVLIST:
{
struct cam_periph *nperiph;
struct periph_list *periph_head;
struct ccb_getdevlist *cgdl;
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
u_int i;
struct cam_ed *device;
int found;
found = 0;
/*
* Don't want anyone mucking with our data.
*/
device = path->device;
periph_head = &device->periphs;
cgdl = &start_ccb->cgdl;
/*
* Check and see if the list has changed since the user
* last requested a list member. If so, tell them that the
2008-01-02 01:45:31 +00:00
* list has changed, and therefore they need to start over
* from the beginning.
*/
2008-01-02 01:45:31 +00:00
if ((cgdl->index != 0) &&
(cgdl->generation != device->generation)) {
cgdl->status = CAM_GDEVLIST_LIST_CHANGED;
break;
}
/*
2008-01-02 01:45:31 +00:00
* Traverse the list of peripherals and attempt to find
* the requested peripheral.
*/
for (nperiph = SLIST_FIRST(periph_head), i = 0;
(nperiph != NULL) && (i <= cgdl->index);
nperiph = SLIST_NEXT(nperiph, periph_links), i++) {
if (i == cgdl->index) {
strlcpy(cgdl->periph_name,
nperiph->periph_name,
sizeof(cgdl->periph_name));
cgdl->unit_number = nperiph->unit_number;
found = 1;
}
}
if (found == 0) {
cgdl->status = CAM_GDEVLIST_ERROR;
break;
}
if (nperiph == NULL)
cgdl->status = CAM_GDEVLIST_LAST_DEVICE;
else
cgdl->status = CAM_GDEVLIST_MORE_DEVS;
cgdl->index++;
cgdl->generation = device->generation;
cgdl->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_DEV_MATCH:
{
dev_pos_type position_type;
struct ccb_dev_match *cdm;
cdm = &start_ccb->cdm;
/*
* There are two ways of getting at information in the EDT.
* The first way is via the primary EDT tree. It starts
* with a list of buses, then a list of targets on a bus,
* then devices/luns on a target, and then peripherals on a
* device/lun. The "other" way is by the peripheral driver
* lists. The peripheral driver lists are organized by
* peripheral driver. (obviously) So it makes sense to
* use the peripheral driver list if the user is looking
* for something like "da1", or all "da" devices. If the
* user is looking for something on a particular bus/target
* or lun, it's generally better to go through the EDT tree.
*/
if (cdm->pos.position_type != CAM_DEV_POS_NONE)
position_type = cdm->pos.position_type;
else {
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
u_int i;
position_type = CAM_DEV_POS_NONE;
for (i = 0; i < cdm->num_patterns; i++) {
if ((cdm->patterns[i].type == DEV_MATCH_BUS)
||(cdm->patterns[i].type == DEV_MATCH_DEVICE)){
position_type = CAM_DEV_POS_EDT;
break;
}
}
if (cdm->num_patterns == 0)
position_type = CAM_DEV_POS_EDT;
else if (position_type == CAM_DEV_POS_NONE)
position_type = CAM_DEV_POS_PDRV;
}
switch(position_type & CAM_DEV_POS_TYPEMASK) {
case CAM_DEV_POS_EDT:
xptedtmatch(cdm);
break;
case CAM_DEV_POS_PDRV:
xptperiphlistmatch(cdm);
break;
default:
cdm->status = CAM_DEV_MATCH_ERROR;
break;
}
if (cdm->status == CAM_DEV_MATCH_ERROR)
start_ccb->ccb_h.status = CAM_REQ_CMP_ERR;
else
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_SASYNC_CB:
{
struct ccb_setasync *csa;
struct async_node *cur_entry;
struct async_list *async_head;
u_int32_t added;
csa = &start_ccb->csa;
added = csa->event_enable;
async_head = &path->device->asyncs;
/*
* If there is already an entry for us, simply
* update it.
*/
cur_entry = SLIST_FIRST(async_head);
while (cur_entry != NULL) {
if ((cur_entry->callback_arg == csa->callback_arg)
&& (cur_entry->callback == csa->callback))
break;
cur_entry = SLIST_NEXT(cur_entry, links);
}
if (cur_entry != NULL) {
/*
* If the request has no flags set,
* remove the entry.
*/
added &= ~cur_entry->event_enable;
if (csa->event_enable == 0) {
SLIST_REMOVE(async_head, cur_entry,
async_node, links);
xpt_release_device(path->device);
free(cur_entry, M_CAMXPT);
} else {
cur_entry->event_enable = csa->event_enable;
}
csa->event_enable = added;
} else {
cur_entry = malloc(sizeof(*cur_entry), M_CAMXPT,
M_NOWAIT);
if (cur_entry == NULL) {
csa->ccb_h.status = CAM_RESRC_UNAVAIL;
break;
}
cur_entry->event_enable = csa->event_enable;
cur_entry->event_lock = (path->bus->sim->mtx &&
mtx_owned(path->bus->sim->mtx)) ? 1 : 0;
cur_entry->callback_arg = csa->callback_arg;
cur_entry->callback = csa->callback;
SLIST_INSERT_HEAD(async_head, cur_entry, links);
xpt_acquire_device(path->device);
}
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_REL_SIMQ:
{
struct ccb_relsim *crs;
struct cam_ed *dev;
crs = &start_ccb->crs;
dev = path->device;
if (dev == NULL) {
crs->ccb_h.status = CAM_DEV_NOT_THERE;
break;
}
if ((crs->release_flags & RELSIM_ADJUST_OPENINGS) != 0) {
/* Don't ever go below one opening */
if (crs->openings > 0) {
xpt_dev_ccbq_resize(path, crs->openings);
if (bootverbose) {
xpt_print(path,
"number of openings is now %d\n",
crs->openings);
}
}
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&dev->sim->devq->send_mtx);
if ((crs->release_flags & RELSIM_RELEASE_AFTER_TIMEOUT) != 0) {
if ((dev->flags & CAM_DEV_REL_TIMEOUT_PENDING) != 0) {
/*
* Just extend the old timeout and decrement
* the freeze count so that a single timeout
* is sufficient for releasing the queue.
*/
start_ccb->ccb_h.flags &= ~CAM_DEV_QFREEZE;
callout_stop(&dev->callout);
} else {
start_ccb->ccb_h.flags |= CAM_DEV_QFREEZE;
}
callout_reset_sbt(&dev->callout,
SBT_1MS * crs->release_timeout, 0,
xpt_release_devq_timeout, dev, 0);
dev->flags |= CAM_DEV_REL_TIMEOUT_PENDING;
}
if ((crs->release_flags & RELSIM_RELEASE_AFTER_CMDCMPLT) != 0) {
if ((dev->flags & CAM_DEV_REL_ON_COMPLETE) != 0) {
/*
* Decrement the freeze count so that a single
* completion is still sufficient to unfreeze
* the queue.
*/
start_ccb->ccb_h.flags &= ~CAM_DEV_QFREEZE;
} else {
2008-01-02 01:45:31 +00:00
dev->flags |= CAM_DEV_REL_ON_COMPLETE;
start_ccb->ccb_h.flags |= CAM_DEV_QFREEZE;
}
}
if ((crs->release_flags & RELSIM_RELEASE_AFTER_QEMPTY) != 0) {
if ((dev->flags & CAM_DEV_REL_ON_QUEUE_EMPTY) != 0
|| (dev->ccbq.dev_active == 0)) {
start_ccb->ccb_h.flags &= ~CAM_DEV_QFREEZE;
} else {
2008-01-02 01:45:31 +00:00
dev->flags |= CAM_DEV_REL_ON_QUEUE_EMPTY;
start_ccb->ccb_h.flags |= CAM_DEV_QFREEZE;
}
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&dev->sim->devq->send_mtx);
if ((start_ccb->ccb_h.flags & CAM_DEV_QFREEZE) == 0)
xpt_release_devq(path, /*count*/1, /*run_queue*/TRUE);
start_ccb->crs.qfrozen_cnt = dev->ccbq.queue.qfrozen_cnt;
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_DEBUG: {
struct cam_path *oldpath;
/* Check that all request bits are supported. */
if (start_ccb->cdbg.flags & ~(CAM_DEBUG_COMPILE)) {
start_ccb->ccb_h.status = CAM_FUNC_NOTAVAIL;
break;
}
cam_dflags = CAM_DEBUG_NONE;
if (cam_dpath != NULL) {
oldpath = cam_dpath;
cam_dpath = NULL;
xpt_free_path(oldpath);
}
if (start_ccb->cdbg.flags != CAM_DEBUG_NONE) {
if (xpt_create_path(&cam_dpath, NULL,
start_ccb->ccb_h.path_id,
start_ccb->ccb_h.target_id,
start_ccb->ccb_h.target_lun) !=
CAM_REQ_CMP) {
start_ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
} else {
cam_dflags = start_ccb->cdbg.flags;
start_ccb->ccb_h.status = CAM_REQ_CMP;
xpt_print(cam_dpath, "debugging flags now %x\n",
cam_dflags);
}
} else
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_NOOP:
if ((start_ccb->ccb_h.flags & CAM_DEV_QFREEZE) != 0)
xpt_freeze_devq(path, 1);
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
case XPT_REPROBE_LUN:
xpt_async(AC_INQ_CHANGED, path, NULL);
start_ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(start_ccb);
break;
case XPT_ASYNC:
start_ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(start_ccb);
break;
default:
case XPT_SDEV_TYPE:
case XPT_TERM_IO:
case XPT_ENG_INQ:
/* XXX Implement */
xpt_print(start_ccb->ccb_h.path,
"%s: CCB type %#x %s not supported\n", __func__,
start_ccb->ccb_h.func_code,
xpt_action_name(start_ccb->ccb_h.func_code));
start_ccb->ccb_h.status = CAM_PROVIDE_FAIL;
if (start_ccb->ccb_h.func_code & XPT_FC_DEV_QUEUED) {
xpt_done(start_ccb);
}
break;
}
CAM_DEBUG(path, CAM_DEBUG_TRACE,
("xpt_action_default: func= %#x %s status %#x\n",
start_ccb->ccb_h.func_code,
xpt_action_name(start_ccb->ccb_h.func_code),
start_ccb->ccb_h.status));
}
/*
* Call the sim poll routine to allow the sim to complete
* any inflight requests, then call camisr_runqueue to
* complete any CCB that the polling completed.
*/
void
xpt_sim_poll(struct cam_sim *sim)
{
struct mtx *mtx;
mtx = sim->mtx;
if (mtx)
mtx_lock(mtx);
(*(sim->sim_poll))(sim);
if (mtx)
mtx_unlock(mtx);
camisr_runqueue();
}
uint32_t
xpt_poll_setup(union ccb *start_ccb)
{
u_int32_t timeout;
2008-01-02 01:45:31 +00:00
struct cam_sim *sim;
struct cam_devq *devq;
struct cam_ed *dev;
timeout = start_ccb->ccb_h.timeout * 10;
sim = start_ccb->ccb_h.path->bus->sim;
devq = sim->devq;
dev = start_ccb->ccb_h.path->device;
/*
* Steal an opening so that no other queued requests
* can get it before us while we simulate interrupts.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&devq->send_mtx);
2008-01-02 01:45:31 +00:00
dev->ccbq.dev_openings--;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
while((devq->send_openings <= 0 || dev->ccbq.dev_openings < 0) &&
(--timeout > 0)) {
mtx_unlock(&devq->send_mtx);
DELAY(100);
xpt_sim_poll(sim);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&devq->send_mtx);
}
dev->ccbq.dev_openings++;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&devq->send_mtx);
2008-01-02 01:45:31 +00:00
return (timeout);
}
void
xpt_pollwait(union ccb *start_ccb, uint32_t timeout)
{
while (--timeout > 0) {
xpt_sim_poll(start_ccb->ccb_h.path->bus->sim);
if ((start_ccb->ccb_h.status & CAM_STATUS_MASK)
!= CAM_REQ_INPROG)
break;
DELAY(100);
}
if (timeout == 0) {
/*
* XXX Is it worth adding a sim_timeout entry
* point so we can attempt recovery? If
* this is only used for dumps, I don't think
* it is.
*/
start_ccb->ccb_h.status = CAM_CMD_TIMEOUT;
}
}
void
xpt_polled_action(union ccb *start_ccb)
{
uint32_t timeout;
struct cam_ed *dev;
timeout = start_ccb->ccb_h.timeout * 10;
dev = start_ccb->ccb_h.path->device;
mtx_unlock(&dev->device_mtx);
timeout = xpt_poll_setup(start_ccb);
if (timeout > 0) {
xpt_action(start_ccb);
xpt_pollwait(start_ccb, timeout);
} else {
start_ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&dev->device_mtx);
}
2008-01-02 01:45:31 +00:00
/*
2014-09-17 18:59:25 +00:00
* Schedule a peripheral driver to receive a ccb when its
* target device has space for more transactions.
*/
void
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_schedule(struct cam_periph *periph, u_int32_t new_priority)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("xpt_schedule\n"));
cam_periph_assert(periph, MA_OWNED);
if (new_priority < periph->scheduled_priority) {
periph->scheduled_priority = new_priority;
xpt_run_allocq(periph, 0);
}
}
/*
* Schedule a device to run on a given queue.
* If the device was inserted as a new entry on the queue,
* return 1 meaning the device queue should be run. If we
* were already queued, implying someone else has already
* started the queue, return 0 so the caller doesn't attempt
* to run the queue.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static int
xpt_schedule_dev(struct camq *queue, cam_pinfo *pinfo,
u_int32_t new_priority)
{
int retval;
u_int32_t old_priority;
CAM_DEBUG_PRINT(CAM_DEBUG_XPT, ("xpt_schedule_dev\n"));
old_priority = pinfo->priority;
/*
* Are we already queued?
*/
if (pinfo->index != CAM_UNQUEUED_INDEX) {
/* Simply reorder based on new priority */
if (new_priority < old_priority) {
camq_change_priority(queue, pinfo->index,
new_priority);
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
("changed priority to %d\n",
new_priority));
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
retval = 1;
} else
retval = 0;
} else {
/* New entry on the queue */
if (new_priority < old_priority)
pinfo->priority = new_priority;
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
("Inserting onto queue\n"));
pinfo->generation = ++queue->generation;
camq_insert(queue, pinfo);
retval = 1;
}
return (retval);
}
static void
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_run_allocq_task(void *context, int pending)
{
struct cam_periph *periph = context;
cam_periph_lock(periph);
periph->flags &= ~CAM_PERIPH_RUN_TASK;
xpt_run_allocq(periph, 1);
cam_periph_unlock(periph);
cam_periph_release(periph);
}
static void
xpt_run_allocq(struct cam_periph *periph, int sleep)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_ed *device;
union ccb *ccb;
uint32_t prio;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
cam_periph_assert(periph, MA_OWNED);
if (periph->periph_allocating)
return;
cam_periph_doacquire(periph);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
periph->periph_allocating = 1;
CAM_DEBUG_PRINT(CAM_DEBUG_XPT, ("xpt_run_allocq(%p)\n", periph));
device = periph->path->device;
ccb = NULL;
restart:
while ((prio = min(periph->scheduled_priority,
periph->immediate_priority)) != CAM_PRIORITY_NONE &&
(periph->periph_allocated - (ccb != NULL ? 1 : 0) <
device->ccbq.total_openings || prio <= CAM_PRIORITY_OOB)) {
if (ccb == NULL &&
(ccb = xpt_get_ccb_nowait(periph)) == NULL) {
if (sleep) {
ccb = xpt_get_ccb(periph);
goto restart;
}
if (periph->flags & CAM_PERIPH_RUN_TASK)
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
break;
cam_periph_doacquire(periph);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
periph->flags |= CAM_PERIPH_RUN_TASK;
taskqueue_enqueue(xsoftc.xpt_taskq,
&periph->periph_run_task);
break;
}
xpt_setup_ccb(&ccb->ccb_h, periph->path, prio);
if (prio == periph->immediate_priority) {
periph->immediate_priority = CAM_PRIORITY_NONE;
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
("waking cam_periph_getccb()\n"));
SLIST_INSERT_HEAD(&periph->ccb_list, &ccb->ccb_h,
periph_links.sle);
wakeup(&periph->ccb_list);
} else {
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
periph->scheduled_priority = CAM_PRIORITY_NONE;
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
("calling periph_start()\n"));
periph->periph_start(periph, ccb);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
ccb = NULL;
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (ccb != NULL)
xpt_release_ccb(ccb);
periph->periph_allocating = 0;
cam_periph_release_locked(periph);
}
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
static void
xpt_run_devq(struct cam_devq *devq)
{
struct mtx *mtx;
CAM_DEBUG_PRINT(CAM_DEBUG_XPT, ("xpt_run_devq\n"));
devq->send_queue.qfrozen_cnt++;
while ((devq->send_queue.entries > 0)
&& (devq->send_openings > 0)
&& (devq->send_queue.qfrozen_cnt <= 1)) {
struct cam_ed *device;
union ccb *work_ccb;
struct cam_sim *sim;
struct xpt_proto *proto;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
device = (struct cam_ed *)camq_remove(&devq->send_queue,
CAMQ_HEAD);
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
("running device %p\n", device));
work_ccb = cam_ccbq_peek_ccb(&device->ccbq, CAMQ_HEAD);
if (work_ccb == NULL) {
printf("device on run queue with no ccbs???\n");
continue;
}
if ((work_ccb->ccb_h.flags & CAM_HIGH_POWER) != 0) {
mtx_lock(&xsoftc.xpt_highpower_lock);
if (xsoftc.num_highpower <= 0) {
/*
* We got a high power command, but we
* don't have any available slots. Freeze
* the device queue until we have a slot
* available.
*/
xpt_freeze_devq_device(device, 1);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
STAILQ_INSERT_TAIL(&xsoftc.highpowerq, device,
highpowerq_entry);
mtx_unlock(&xsoftc.xpt_highpower_lock);
continue;
} else {
/*
* Consume a high power slot while
* this ccb runs.
*/
xsoftc.num_highpower--;
}
mtx_unlock(&xsoftc.xpt_highpower_lock);
}
cam_ccbq_remove_ccb(&device->ccbq, work_ccb);
cam_ccbq_send_ccb(&device->ccbq, work_ccb);
devq->send_openings--;
2008-01-02 01:45:31 +00:00
devq->send_active++;
xpt_schedule_devq(devq, device);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&devq->send_mtx);
if ((work_ccb->ccb_h.flags & CAM_DEV_QFREEZE) != 0) {
/*
* The client wants to freeze the queue
* after this CCB is sent.
*/
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_freeze_devq(work_ccb->ccb_h.path, 1);
}
/* In Target mode, the peripheral driver knows best... */
if (work_ccb->ccb_h.func_code == XPT_SCSI_IO) {
if ((device->inq_flags & SID_CmdQue) != 0
&& work_ccb->csio.tag_action != CAM_TAG_ACTION_NONE)
work_ccb->ccb_h.flags |= CAM_TAG_ACTION_VALID;
else
/*
* Clear this in case of a retried CCB that
* failed due to a rejected tag.
*/
work_ccb->ccb_h.flags &= ~CAM_TAG_ACTION_VALID;
}
KASSERT(device == work_ccb->ccb_h.path->device,
("device (%p) / path->device (%p) mismatch",
device, work_ccb->ccb_h.path->device));
proto = xpt_proto_find(device->protocol);
if (proto && proto->ops->debug_out)
proto->ops->debug_out(work_ccb);
/*
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
* Device queues can be shared among multiple SIM instances
* that reside on different buses. Use the SIM from the
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
* queued device, rather than the one from the calling bus.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
sim = device->sim;
mtx = sim->mtx;
if (mtx && !mtx_owned(mtx))
mtx_lock(mtx);
else
mtx = NULL;
work_ccb->ccb_h.qos.periph_data = cam_iosched_now();
(*(sim->sim_action))(sim, work_ccb);
if (mtx)
mtx_unlock(mtx);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&devq->send_mtx);
}
devq->send_queue.qfrozen_cnt--;
}
/*
* This function merges stuff from the src ccb into the dst ccb, while keeping
* important fields in the dst ccb constant.
*/
void
xpt_merge_ccb(union ccb *dst_ccb, union ccb *src_ccb)
{
/*
* Pull fields that are valid for peripheral drivers to set
* into the dst CCB along with the CCB "payload".
*/
dst_ccb->ccb_h.retry_count = src_ccb->ccb_h.retry_count;
dst_ccb->ccb_h.func_code = src_ccb->ccb_h.func_code;
dst_ccb->ccb_h.timeout = src_ccb->ccb_h.timeout;
dst_ccb->ccb_h.flags = src_ccb->ccb_h.flags;
bcopy(&(&src_ccb->ccb_h)[1], &(&dst_ccb->ccb_h)[1],
sizeof(union ccb) - sizeof(struct ccb_hdr));
}
void
Add asynchronous command support to the pass(4) driver, and the new camdd(8) utility. CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and completed CCBs may be retrieved via the CAMIOGET ioctl. User processes can use poll(2) or kevent(2) to get notification when I/O has completed. While the existing CAMIOCOMMAND blocking ioctl interface only supports user virtual data pointers in a CCB (generally only one per CCB), the new CAMIOQUEUE ioctl supports user virtual and physical address pointers, as well as user virtual and physical scatter/gather lists. This allows user applications to have more flexibility in their data handling operations. Kernel memory for data transferred via the queued interface is allocated from the zone allocator in MAXPHYS sized chunks, and user data is copied in and out. This is likely faster than the vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in configurations with many processors (there are more TLB shootdowns caused by the mapping/unmapping operation) but may not be as fast as running with unmapped I/O. The new memory handling model for user requests also allows applications to send CCBs with request sizes that are larger than MAXPHYS. The pass(4) driver now limits queued requests to the I/O size listed by the SIM driver in the maxio field in the Path Inquiry (XPT_PATH_INQ) CCB. There are some things things would be good to add: 1. Come up with a way to do unmapped I/O on multiple buffers. Currently the unmapped I/O interface operates on a struct bio, which includes only one address and length. It would be nice to be able to send an unmapped scatter/gather list down to busdma. This would allow eliminating the copy we currently do for data. 2. Add an ioctl to list currently outstanding CCBs in the various queues. 3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do that. 4. Test physical address support. Virtual pointers and scatter gather lists have been tested, but I have not yet tested physical addresses or scatter/gather lists. 5. Investigate multiple queue support. At the moment there is one queue of commands per pass(4) device. If multiple processes open the device, they will submit I/O into the same queue and get events for the same completions. This is probably the right model for most applications, but it is something that could be changed later on. Also, add a new utility, camdd(8) that uses the asynchronous pass(4) driver interface. This utility is intended to be a basic data transfer/copy utility, a simple benchmark utility, and an example of how to use the asynchronous pass(4) interface. It can copy data to and from pass(4) devices using any target queue depth, starting offset and blocksize for the input and ouptut devices. It currently only supports SCSI devices, but could be easily extended to support ATA devices. It can also copy data to and from regular files, block devices, tape devices, pipes, stdin, and stdout. It does not support queueing multiple commands to any of those targets, since it uses the standard read(2)/write(2)/writev(2)/readv(2) system calls. The I/O is done by two threads, one for the reader and one for the writer. The reader thread sends completed read requests to the writer thread in strictly sequential order, even if they complete out of order. That could be modified later on for random I/O patterns or slightly out of order I/O. camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from the pass(4) driver and also to send request notifications internally. For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR) per CAM CCB on the reading side, and a scatter/gather list (CAM_DATA_SG) on the writing side. In addition to testing both interfaces, this makes any potential reblocking of I/O easier. No data is copied between the reader and the writer, but rather the reader's buffers are split into multiple I/O requests or combined into a single I/O request depending on the input and output blocksize. For the file I/O path, camdd(8) also uses a single buffer (read(2), write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list (readv(2), writev(2), preadv(2), pwritev(2)) on writes. Things that would be nice to do for camdd(8) eventually: 1. Add support for I/O pattern generation. Patterns like all zeros, all ones, LBA-based patterns, random patterns, etc. Right Now you can always use /dev/zero, /dev/random, etc. 2. Add support for a "sink" mode, so we do only reads with no writes. Right now, you can use /dev/null. 3. Add support for automatic queue depth probing, so that we can figure out the right queue depth on the input and output side for maximum throughput. At the moment it defaults to 6. 4. Add support for SATA device passthrough I/O. 5. Add support for random LBAs and/or lengths on the input and output sides. 6. Track average per-I/O latency and busy time. The busy time and latency could also feed in to the automatic queue depth determination. sys/cam/scsi/scsi_pass.h: Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue and fetch asynchronous CAM CCBs respectively. Although these ioctls do not have a declared argument, they both take a union ccb pointer. If we declare a size here, the ioctl code in sys/kern/sys_generic.c will malloc and free a buffer for either the CCB or the CCB pointer (depending on how it is declared). Since we have to keep a copy of the CCB (which is fairly large) anyway, having the ioctl malloc and free a CCB for each call is wasteful. sys/cam/scsi/scsi_pass.c: Add asynchronous CCB support. Add two new ioctls, CAMIOQUEUE and CAMIOGET. CAMIOQUEUE adds a CCB to the incoming queue. The CCB is executed immediately (and moved to the active queue) if it is an immediate CCB, but otherwise it will be executed in passstart() when a CCB is available from the transport layer. When CCBs are completed (because they are immediate or passdone() if they are queued), they are put on the done queue. If we get the final close on the device before all pending I/O is complete, all active I/O is moved to the abandoned queue and we increment the peripheral reference count so that the peripheral driver instance doesn't go away before all pending I/O is done. The new passcreatezone() function is called on the first call to the CAMIOQUEUE ioctl on a given device to allocate the UMA zones for I/O requests and S/G list buffers. This may be good to move off to a taskqueue at some point. The new passmemsetup() function allocates memory and scatter/gather lists to hold the user's data, and copies in any data that needs to be written. For virtual pointers (CAM_DATA_VADDR), the kernel buffer is malloced from the new pass(4) driver malloc bucket. For virtual scatter/gather lists (CAM_DATA_SG), buffers are allocated from a new per-pass(9) UMA zone in MAXPHYS-sized chunks. Physical pointers are passed in unchanged. We have support for up to 16 scatter/gather segments (for the user and kernel S/G lists) in the default struct pass_io_req, so requests with longer S/G lists require an extra kernel malloc. The new passcopysglist() function copies a user scatter/gather list to a kernel scatter/gather list. The number of elements in each list may be different, but (obviously) the amount of data stored has to be identical. The new passmemdone() function copies data out for the CAM_DATA_VADDR and CAM_DATA_SG cases. The new passiocleanup() function restores data pointers in user CCBs and frees memory. Add new functions to support kqueue(2)/kevent(2): passreadfilt() tells kevent whether or not the done queue is empty. passkqfilter() adds a knote to our list. passreadfiltdetach() removes a knote from our list. Add a new function, passpoll(), for poll(2)/select(2) to use. Add devstat(9) support for the queued CCB path. sys/cam/ata/ata_da.c: Add support for the BIO_VLIST bio type. sys/cam/cam_ccb.h: Add a new enumeration for the xflags field in the CCB header. (This doesn't change the CCB header, just adds an enumeration to use.) sys/cam/cam_xpt.c: Add a new function, xpt_setup_ccb_flags(), that allows specifying CCB flags. sys/cam/cam_xpt.h: Add a prototype for xpt_setup_ccb_flags(). sys/cam/scsi/scsi_da.c: Add support for BIO_VLIST. sys/dev/md/md.c: Add BIO_VLIST support to md(4). sys/geom/geom_disk.c: Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size limiting code in g_disk_start() a bit. sys/kern/subr_bus_dma.c: Change _bus_dmamap_load_vlist() to take a starting offset and length. Add a new function, _bus_dmamap_load_pages(), that will load a list of physical pages starting at an offset. Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios. Allow unmapped I/O to start at an offset. sys/kern/subr_uio.c: Add two new functions, physcopyin_vlist() and physcopyout_vlist(). sys/pc98/include/bus.h: Guard kernel-only parts of the pc98 machine/bus.h header with #ifdef _KERNEL. This allows userland programs to include <machine/bus.h> to get the definition of bus_addr_t and bus_size_t. sys/sys/bio.h: Add a new bio flag, BIO_VLIST. sys/sys/uio.h: Add prototypes for physcopyin_vlist() and physcopyout_vlist(). share/man/man4/pass.4: Document the CAMIOQUEUE and CAMIOGET ioctls. usr.sbin/Makefile: Add camdd. usr.sbin/camdd/Makefile: Add a makefile for camdd(8). usr.sbin/camdd/camdd.8: Man page for camdd(8). usr.sbin/camdd/camdd.c: The new camdd(8) utility. Sponsored by: Spectra Logic MFC after: 1 week
2015-12-03 20:54:55 +00:00
xpt_setup_ccb_flags(struct ccb_hdr *ccb_h, struct cam_path *path,
u_int32_t priority, u_int32_t flags)
{
CAM_DEBUG(path, CAM_DEBUG_TRACE, ("xpt_setup_ccb\n"));
ccb_h->pinfo.priority = priority;
ccb_h->path = path;
ccb_h->path_id = path->bus->path_id;
if (path->target)
ccb_h->target_id = path->target->target_id;
else
ccb_h->target_id = CAM_TARGET_WILDCARD;
if (path->device) {
ccb_h->target_lun = path->device->lun_id;
ccb_h->pinfo.generation = ++path->device->ccbq.queue.generation;
} else {
ccb_h->target_lun = CAM_TARGET_WILDCARD;
}
ccb_h->pinfo.index = CAM_UNQUEUED_INDEX;
Add asynchronous command support to the pass(4) driver, and the new camdd(8) utility. CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and completed CCBs may be retrieved via the CAMIOGET ioctl. User processes can use poll(2) or kevent(2) to get notification when I/O has completed. While the existing CAMIOCOMMAND blocking ioctl interface only supports user virtual data pointers in a CCB (generally only one per CCB), the new CAMIOQUEUE ioctl supports user virtual and physical address pointers, as well as user virtual and physical scatter/gather lists. This allows user applications to have more flexibility in their data handling operations. Kernel memory for data transferred via the queued interface is allocated from the zone allocator in MAXPHYS sized chunks, and user data is copied in and out. This is likely faster than the vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in configurations with many processors (there are more TLB shootdowns caused by the mapping/unmapping operation) but may not be as fast as running with unmapped I/O. The new memory handling model for user requests also allows applications to send CCBs with request sizes that are larger than MAXPHYS. The pass(4) driver now limits queued requests to the I/O size listed by the SIM driver in the maxio field in the Path Inquiry (XPT_PATH_INQ) CCB. There are some things things would be good to add: 1. Come up with a way to do unmapped I/O on multiple buffers. Currently the unmapped I/O interface operates on a struct bio, which includes only one address and length. It would be nice to be able to send an unmapped scatter/gather list down to busdma. This would allow eliminating the copy we currently do for data. 2. Add an ioctl to list currently outstanding CCBs in the various queues. 3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do that. 4. Test physical address support. Virtual pointers and scatter gather lists have been tested, but I have not yet tested physical addresses or scatter/gather lists. 5. Investigate multiple queue support. At the moment there is one queue of commands per pass(4) device. If multiple processes open the device, they will submit I/O into the same queue and get events for the same completions. This is probably the right model for most applications, but it is something that could be changed later on. Also, add a new utility, camdd(8) that uses the asynchronous pass(4) driver interface. This utility is intended to be a basic data transfer/copy utility, a simple benchmark utility, and an example of how to use the asynchronous pass(4) interface. It can copy data to and from pass(4) devices using any target queue depth, starting offset and blocksize for the input and ouptut devices. It currently only supports SCSI devices, but could be easily extended to support ATA devices. It can also copy data to and from regular files, block devices, tape devices, pipes, stdin, and stdout. It does not support queueing multiple commands to any of those targets, since it uses the standard read(2)/write(2)/writev(2)/readv(2) system calls. The I/O is done by two threads, one for the reader and one for the writer. The reader thread sends completed read requests to the writer thread in strictly sequential order, even if they complete out of order. That could be modified later on for random I/O patterns or slightly out of order I/O. camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from the pass(4) driver and also to send request notifications internally. For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR) per CAM CCB on the reading side, and a scatter/gather list (CAM_DATA_SG) on the writing side. In addition to testing both interfaces, this makes any potential reblocking of I/O easier. No data is copied between the reader and the writer, but rather the reader's buffers are split into multiple I/O requests or combined into a single I/O request depending on the input and output blocksize. For the file I/O path, camdd(8) also uses a single buffer (read(2), write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list (readv(2), writev(2), preadv(2), pwritev(2)) on writes. Things that would be nice to do for camdd(8) eventually: 1. Add support for I/O pattern generation. Patterns like all zeros, all ones, LBA-based patterns, random patterns, etc. Right Now you can always use /dev/zero, /dev/random, etc. 2. Add support for a "sink" mode, so we do only reads with no writes. Right now, you can use /dev/null. 3. Add support for automatic queue depth probing, so that we can figure out the right queue depth on the input and output side for maximum throughput. At the moment it defaults to 6. 4. Add support for SATA device passthrough I/O. 5. Add support for random LBAs and/or lengths on the input and output sides. 6. Track average per-I/O latency and busy time. The busy time and latency could also feed in to the automatic queue depth determination. sys/cam/scsi/scsi_pass.h: Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue and fetch asynchronous CAM CCBs respectively. Although these ioctls do not have a declared argument, they both take a union ccb pointer. If we declare a size here, the ioctl code in sys/kern/sys_generic.c will malloc and free a buffer for either the CCB or the CCB pointer (depending on how it is declared). Since we have to keep a copy of the CCB (which is fairly large) anyway, having the ioctl malloc and free a CCB for each call is wasteful. sys/cam/scsi/scsi_pass.c: Add asynchronous CCB support. Add two new ioctls, CAMIOQUEUE and CAMIOGET. CAMIOQUEUE adds a CCB to the incoming queue. The CCB is executed immediately (and moved to the active queue) if it is an immediate CCB, but otherwise it will be executed in passstart() when a CCB is available from the transport layer. When CCBs are completed (because they are immediate or passdone() if they are queued), they are put on the done queue. If we get the final close on the device before all pending I/O is complete, all active I/O is moved to the abandoned queue and we increment the peripheral reference count so that the peripheral driver instance doesn't go away before all pending I/O is done. The new passcreatezone() function is called on the first call to the CAMIOQUEUE ioctl on a given device to allocate the UMA zones for I/O requests and S/G list buffers. This may be good to move off to a taskqueue at some point. The new passmemsetup() function allocates memory and scatter/gather lists to hold the user's data, and copies in any data that needs to be written. For virtual pointers (CAM_DATA_VADDR), the kernel buffer is malloced from the new pass(4) driver malloc bucket. For virtual scatter/gather lists (CAM_DATA_SG), buffers are allocated from a new per-pass(9) UMA zone in MAXPHYS-sized chunks. Physical pointers are passed in unchanged. We have support for up to 16 scatter/gather segments (for the user and kernel S/G lists) in the default struct pass_io_req, so requests with longer S/G lists require an extra kernel malloc. The new passcopysglist() function copies a user scatter/gather list to a kernel scatter/gather list. The number of elements in each list may be different, but (obviously) the amount of data stored has to be identical. The new passmemdone() function copies data out for the CAM_DATA_VADDR and CAM_DATA_SG cases. The new passiocleanup() function restores data pointers in user CCBs and frees memory. Add new functions to support kqueue(2)/kevent(2): passreadfilt() tells kevent whether or not the done queue is empty. passkqfilter() adds a knote to our list. passreadfiltdetach() removes a knote from our list. Add a new function, passpoll(), for poll(2)/select(2) to use. Add devstat(9) support for the queued CCB path. sys/cam/ata/ata_da.c: Add support for the BIO_VLIST bio type. sys/cam/cam_ccb.h: Add a new enumeration for the xflags field in the CCB header. (This doesn't change the CCB header, just adds an enumeration to use.) sys/cam/cam_xpt.c: Add a new function, xpt_setup_ccb_flags(), that allows specifying CCB flags. sys/cam/cam_xpt.h: Add a prototype for xpt_setup_ccb_flags(). sys/cam/scsi/scsi_da.c: Add support for BIO_VLIST. sys/dev/md/md.c: Add BIO_VLIST support to md(4). sys/geom/geom_disk.c: Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size limiting code in g_disk_start() a bit. sys/kern/subr_bus_dma.c: Change _bus_dmamap_load_vlist() to take a starting offset and length. Add a new function, _bus_dmamap_load_pages(), that will load a list of physical pages starting at an offset. Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios. Allow unmapped I/O to start at an offset. sys/kern/subr_uio.c: Add two new functions, physcopyin_vlist() and physcopyout_vlist(). sys/pc98/include/bus.h: Guard kernel-only parts of the pc98 machine/bus.h header with #ifdef _KERNEL. This allows userland programs to include <machine/bus.h> to get the definition of bus_addr_t and bus_size_t. sys/sys/bio.h: Add a new bio flag, BIO_VLIST. sys/sys/uio.h: Add prototypes for physcopyin_vlist() and physcopyout_vlist(). share/man/man4/pass.4: Document the CAMIOQUEUE and CAMIOGET ioctls. usr.sbin/Makefile: Add camdd. usr.sbin/camdd/Makefile: Add a makefile for camdd(8). usr.sbin/camdd/camdd.8: Man page for camdd(8). usr.sbin/camdd/camdd.c: The new camdd(8) utility. Sponsored by: Spectra Logic MFC after: 1 week
2015-12-03 20:54:55 +00:00
ccb_h->flags = flags;
ccb_h->xflags = 0;
}
Add asynchronous command support to the pass(4) driver, and the new camdd(8) utility. CCBs may be queued to the driver via the new CAMIOQUEUE ioctl, and completed CCBs may be retrieved via the CAMIOGET ioctl. User processes can use poll(2) or kevent(2) to get notification when I/O has completed. While the existing CAMIOCOMMAND blocking ioctl interface only supports user virtual data pointers in a CCB (generally only one per CCB), the new CAMIOQUEUE ioctl supports user virtual and physical address pointers, as well as user virtual and physical scatter/gather lists. This allows user applications to have more flexibility in their data handling operations. Kernel memory for data transferred via the queued interface is allocated from the zone allocator in MAXPHYS sized chunks, and user data is copied in and out. This is likely faster than the vmapbuf()/vunmapbuf() method used by the CAMIOCOMMAND ioctl in configurations with many processors (there are more TLB shootdowns caused by the mapping/unmapping operation) but may not be as fast as running with unmapped I/O. The new memory handling model for user requests also allows applications to send CCBs with request sizes that are larger than MAXPHYS. The pass(4) driver now limits queued requests to the I/O size listed by the SIM driver in the maxio field in the Path Inquiry (XPT_PATH_INQ) CCB. There are some things things would be good to add: 1. Come up with a way to do unmapped I/O on multiple buffers. Currently the unmapped I/O interface operates on a struct bio, which includes only one address and length. It would be nice to be able to send an unmapped scatter/gather list down to busdma. This would allow eliminating the copy we currently do for data. 2. Add an ioctl to list currently outstanding CCBs in the various queues. 3. Add an ioctl to cancel a request, or use the XPT_ABORT CCB to do that. 4. Test physical address support. Virtual pointers and scatter gather lists have been tested, but I have not yet tested physical addresses or scatter/gather lists. 5. Investigate multiple queue support. At the moment there is one queue of commands per pass(4) device. If multiple processes open the device, they will submit I/O into the same queue and get events for the same completions. This is probably the right model for most applications, but it is something that could be changed later on. Also, add a new utility, camdd(8) that uses the asynchronous pass(4) driver interface. This utility is intended to be a basic data transfer/copy utility, a simple benchmark utility, and an example of how to use the asynchronous pass(4) interface. It can copy data to and from pass(4) devices using any target queue depth, starting offset and blocksize for the input and ouptut devices. It currently only supports SCSI devices, but could be easily extended to support ATA devices. It can also copy data to and from regular files, block devices, tape devices, pipes, stdin, and stdout. It does not support queueing multiple commands to any of those targets, since it uses the standard read(2)/write(2)/writev(2)/readv(2) system calls. The I/O is done by two threads, one for the reader and one for the writer. The reader thread sends completed read requests to the writer thread in strictly sequential order, even if they complete out of order. That could be modified later on for random I/O patterns or slightly out of order I/O. camdd(8) uses kqueue(2)/kevent(2) to get I/O completion events from the pass(4) driver and also to send request notifications internally. For pass(4) devcies, camdd(8) uses a single buffer (CAM_DATA_VADDR) per CAM CCB on the reading side, and a scatter/gather list (CAM_DATA_SG) on the writing side. In addition to testing both interfaces, this makes any potential reblocking of I/O easier. No data is copied between the reader and the writer, but rather the reader's buffers are split into multiple I/O requests or combined into a single I/O request depending on the input and output blocksize. For the file I/O path, camdd(8) also uses a single buffer (read(2), write(2), pread(2) or pwrite(2)) on reads, and a scatter/gather list (readv(2), writev(2), preadv(2), pwritev(2)) on writes. Things that would be nice to do for camdd(8) eventually: 1. Add support for I/O pattern generation. Patterns like all zeros, all ones, LBA-based patterns, random patterns, etc. Right Now you can always use /dev/zero, /dev/random, etc. 2. Add support for a "sink" mode, so we do only reads with no writes. Right now, you can use /dev/null. 3. Add support for automatic queue depth probing, so that we can figure out the right queue depth on the input and output side for maximum throughput. At the moment it defaults to 6. 4. Add support for SATA device passthrough I/O. 5. Add support for random LBAs and/or lengths on the input and output sides. 6. Track average per-I/O latency and busy time. The busy time and latency could also feed in to the automatic queue depth determination. sys/cam/scsi/scsi_pass.h: Define two new ioctls, CAMIOQUEUE and CAMIOGET, that queue and fetch asynchronous CAM CCBs respectively. Although these ioctls do not have a declared argument, they both take a union ccb pointer. If we declare a size here, the ioctl code in sys/kern/sys_generic.c will malloc and free a buffer for either the CCB or the CCB pointer (depending on how it is declared). Since we have to keep a copy of the CCB (which is fairly large) anyway, having the ioctl malloc and free a CCB for each call is wasteful. sys/cam/scsi/scsi_pass.c: Add asynchronous CCB support. Add two new ioctls, CAMIOQUEUE and CAMIOGET. CAMIOQUEUE adds a CCB to the incoming queue. The CCB is executed immediately (and moved to the active queue) if it is an immediate CCB, but otherwise it will be executed in passstart() when a CCB is available from the transport layer. When CCBs are completed (because they are immediate or passdone() if they are queued), they are put on the done queue. If we get the final close on the device before all pending I/O is complete, all active I/O is moved to the abandoned queue and we increment the peripheral reference count so that the peripheral driver instance doesn't go away before all pending I/O is done. The new passcreatezone() function is called on the first call to the CAMIOQUEUE ioctl on a given device to allocate the UMA zones for I/O requests and S/G list buffers. This may be good to move off to a taskqueue at some point. The new passmemsetup() function allocates memory and scatter/gather lists to hold the user's data, and copies in any data that needs to be written. For virtual pointers (CAM_DATA_VADDR), the kernel buffer is malloced from the new pass(4) driver malloc bucket. For virtual scatter/gather lists (CAM_DATA_SG), buffers are allocated from a new per-pass(9) UMA zone in MAXPHYS-sized chunks. Physical pointers are passed in unchanged. We have support for up to 16 scatter/gather segments (for the user and kernel S/G lists) in the default struct pass_io_req, so requests with longer S/G lists require an extra kernel malloc. The new passcopysglist() function copies a user scatter/gather list to a kernel scatter/gather list. The number of elements in each list may be different, but (obviously) the amount of data stored has to be identical. The new passmemdone() function copies data out for the CAM_DATA_VADDR and CAM_DATA_SG cases. The new passiocleanup() function restores data pointers in user CCBs and frees memory. Add new functions to support kqueue(2)/kevent(2): passreadfilt() tells kevent whether or not the done queue is empty. passkqfilter() adds a knote to our list. passreadfiltdetach() removes a knote from our list. Add a new function, passpoll(), for poll(2)/select(2) to use. Add devstat(9) support for the queued CCB path. sys/cam/ata/ata_da.c: Add support for the BIO_VLIST bio type. sys/cam/cam_ccb.h: Add a new enumeration for the xflags field in the CCB header. (This doesn't change the CCB header, just adds an enumeration to use.) sys/cam/cam_xpt.c: Add a new function, xpt_setup_ccb_flags(), that allows specifying CCB flags. sys/cam/cam_xpt.h: Add a prototype for xpt_setup_ccb_flags(). sys/cam/scsi/scsi_da.c: Add support for BIO_VLIST. sys/dev/md/md.c: Add BIO_VLIST support to md(4). sys/geom/geom_disk.c: Add BIO_VLIST support to the GEOM disk class. Re-factor the I/O size limiting code in g_disk_start() a bit. sys/kern/subr_bus_dma.c: Change _bus_dmamap_load_vlist() to take a starting offset and length. Add a new function, _bus_dmamap_load_pages(), that will load a list of physical pages starting at an offset. Update _bus_dmamap_load_bio() to allow loading BIO_VLIST bios. Allow unmapped I/O to start at an offset. sys/kern/subr_uio.c: Add two new functions, physcopyin_vlist() and physcopyout_vlist(). sys/pc98/include/bus.h: Guard kernel-only parts of the pc98 machine/bus.h header with #ifdef _KERNEL. This allows userland programs to include <machine/bus.h> to get the definition of bus_addr_t and bus_size_t. sys/sys/bio.h: Add a new bio flag, BIO_VLIST. sys/sys/uio.h: Add prototypes for physcopyin_vlist() and physcopyout_vlist(). share/man/man4/pass.4: Document the CAMIOQUEUE and CAMIOGET ioctls. usr.sbin/Makefile: Add camdd. usr.sbin/camdd/Makefile: Add a makefile for camdd(8). usr.sbin/camdd/camdd.8: Man page for camdd(8). usr.sbin/camdd/camdd.c: The new camdd(8) utility. Sponsored by: Spectra Logic MFC after: 1 week
2015-12-03 20:54:55 +00:00
void
xpt_setup_ccb(struct ccb_hdr *ccb_h, struct cam_path *path, u_int32_t priority)
{
xpt_setup_ccb_flags(ccb_h, path, priority, /*flags*/ 0);
}
/* Path manipulation functions */
cam_status
xpt_create_path(struct cam_path **new_path_ptr, struct cam_periph *perph,
path_id_t path_id, target_id_t target_id, lun_id_t lun_id)
{
struct cam_path *path;
cam_status status;
path = (struct cam_path *)malloc(sizeof(*path), M_CAMPATH, M_NOWAIT);
if (path == NULL) {
status = CAM_RESRC_UNAVAIL;
return(status);
}
status = xpt_compile_path(path, perph, path_id, target_id, lun_id);
if (status != CAM_REQ_CMP) {
free(path, M_CAMPATH);
path = NULL;
}
*new_path_ptr = path;
return (status);
}
cam_status
xpt_create_path_unlocked(struct cam_path **new_path_ptr,
struct cam_periph *periph, path_id_t path_id,
target_id_t target_id, lun_id_t lun_id)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
return (xpt_create_path(new_path_ptr, periph, path_id, target_id,
lun_id));
}
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
cam_status
xpt_compile_path(struct cam_path *new_path, struct cam_periph *perph,
path_id_t path_id, target_id_t target_id, lun_id_t lun_id)
{
struct cam_eb *bus;
struct cam_et *target;
struct cam_ed *device;
cam_status status;
status = CAM_REQ_CMP; /* Completed without error */
target = NULL; /* Wildcarded */
device = NULL; /* Wildcarded */
/*
* We will potentially modify the EDT, so block interrupts
* that may attempt to create cam paths.
*/
bus = xpt_find_bus(path_id);
if (bus == NULL) {
status = CAM_PATH_INVALID;
} else {
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_lock_buses();
mtx_lock(&bus->eb_mtx);
target = xpt_find_target(bus, target_id);
if (target == NULL) {
/* Create one */
struct cam_et *new_target;
new_target = xpt_alloc_target(bus, target_id);
if (new_target == NULL) {
status = CAM_RESRC_UNAVAIL;
} else {
target = new_target;
}
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_unlock_buses();
if (target != NULL) {
device = xpt_find_device(target, lun_id);
if (device == NULL) {
/* Create one */
struct cam_ed *new_device;
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
new_device =
(*(bus->xport->ops->alloc_device))(bus,
target,
lun_id);
if (new_device == NULL) {
status = CAM_RESRC_UNAVAIL;
} else {
device = new_device;
}
}
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&bus->eb_mtx);
}
/*
* Only touch the user's data if we are successful.
*/
if (status == CAM_REQ_CMP) {
new_path->periph = perph;
new_path->bus = bus;
new_path->target = target;
new_path->device = device;
CAM_DEBUG(new_path, CAM_DEBUG_TRACE, ("xpt_compile_path\n"));
} else {
if (device != NULL)
xpt_release_device(device);
if (target != NULL)
xpt_release_target(target);
if (bus != NULL)
xpt_release_bus(bus);
}
return (status);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
cam_status
xpt_clone_path(struct cam_path **new_path_ptr, struct cam_path *path)
{
struct cam_path *new_path;
new_path = (struct cam_path *)malloc(sizeof(*path), M_CAMPATH, M_NOWAIT);
if (new_path == NULL)
return(CAM_RESRC_UNAVAIL);
*new_path = *path;
if (path->bus != NULL)
xpt_acquire_bus(path->bus);
if (path->target != NULL)
xpt_acquire_target(path->target);
if (path->device != NULL)
xpt_acquire_device(path->device);
*new_path_ptr = new_path;
return (CAM_REQ_CMP);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
void
xpt_release_path(struct cam_path *path)
{
CAM_DEBUG(path, CAM_DEBUG_TRACE, ("xpt_release_path\n"));
if (path->device != NULL) {
xpt_release_device(path->device);
path->device = NULL;
}
if (path->target != NULL) {
xpt_release_target(path->target);
path->target = NULL;
}
if (path->bus != NULL) {
xpt_release_bus(path->bus);
path->bus = NULL;
}
}
void
xpt_free_path(struct cam_path *path)
{
CAM_DEBUG(path, CAM_DEBUG_TRACE, ("xpt_free_path\n"));
xpt_release_path(path);
free(path, M_CAMPATH);
}
void
xpt_path_counts(struct cam_path *path, uint32_t *bus_ref,
uint32_t *periph_ref, uint32_t *target_ref, uint32_t *device_ref)
{
xpt_lock_buses();
if (bus_ref) {
if (path->bus)
*bus_ref = path->bus->refcount;
else
*bus_ref = 0;
}
if (periph_ref) {
if (path->periph)
*periph_ref = path->periph->refcount;
else
*periph_ref = 0;
}
xpt_unlock_buses();
if (target_ref) {
if (path->target)
*target_ref = path->target->refcount;
else
*target_ref = 0;
}
if (device_ref) {
if (path->device)
*device_ref = path->device->refcount;
else
*device_ref = 0;
}
}
/*
* Return -1 for failure, 0 for exact match, 1 for match with wildcards
* in path1, 2 for match with wildcards in path2.
*/
int
xpt_path_comp(struct cam_path *path1, struct cam_path *path2)
{
int retval = 0;
if (path1->bus != path2->bus) {
if (path1->bus->path_id == CAM_BUS_WILDCARD)
retval = 1;
else if (path2->bus->path_id == CAM_BUS_WILDCARD)
retval = 2;
else
return (-1);
}
if (path1->target != path2->target) {
if (path1->target->target_id == CAM_TARGET_WILDCARD) {
if (retval == 0)
retval = 1;
} else if (path2->target->target_id == CAM_TARGET_WILDCARD)
retval = 2;
else
return (-1);
}
if (path1->device != path2->device) {
if (path1->device->lun_id == CAM_LUN_WILDCARD) {
if (retval == 0)
retval = 1;
} else if (path2->device->lun_id == CAM_LUN_WILDCARD)
retval = 2;
else
return (-1);
}
return (retval);
}
int
xpt_path_comp_dev(struct cam_path *path, struct cam_ed *dev)
{
int retval = 0;
if (path->bus != dev->target->bus) {
if (path->bus->path_id == CAM_BUS_WILDCARD)
retval = 1;
else if (dev->target->bus->path_id == CAM_BUS_WILDCARD)
retval = 2;
else
return (-1);
}
if (path->target != dev->target) {
if (path->target->target_id == CAM_TARGET_WILDCARD) {
if (retval == 0)
retval = 1;
} else if (dev->target->target_id == CAM_TARGET_WILDCARD)
retval = 2;
else
return (-1);
}
if (path->device != dev) {
if (path->device->lun_id == CAM_LUN_WILDCARD) {
if (retval == 0)
retval = 1;
} else if (dev->lun_id == CAM_LUN_WILDCARD)
retval = 2;
else
return (-1);
}
return (retval);
}
void
xpt_print_path(struct cam_path *path)
{
struct sbuf sb;
char buffer[XPT_PRINT_LEN];
sbuf_new(&sb, buffer, XPT_PRINT_LEN, SBUF_FIXEDLEN);
xpt_path_sbuf(path, &sb);
sbuf_finish(&sb);
printf("%s", sbuf_data(&sb));
sbuf_delete(&sb);
}
void
xpt_print_device(struct cam_ed *device)
{
if (device == NULL)
printf("(nopath): ");
else {
printf("(noperiph:%s%d:%d:%d:%jx): ", device->sim->sim_name,
device->sim->unit_number,
device->sim->bus_id,
device->target->target_id,
(uintmax_t)device->lun_id);
}
}
void
xpt_print(struct cam_path *path, const char *fmt, ...)
{
va_list ap;
struct sbuf sb;
char buffer[XPT_PRINT_LEN];
sbuf_new(&sb, buffer, XPT_PRINT_LEN, SBUF_FIXEDLEN);
xpt_path_sbuf(path, &sb);
va_start(ap, fmt);
sbuf_vprintf(&sb, fmt, ap);
va_end(ap);
sbuf_finish(&sb);
printf("%s", sbuf_data(&sb));
sbuf_delete(&sb);
}
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
int
xpt_path_string(struct cam_path *path, char *str, size_t str_len)
{
struct sbuf sb;
int len;
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
sbuf_new(&sb, str, str_len, 0);
len = xpt_path_sbuf(path, &sb);
sbuf_finish(&sb);
return (len);
}
int
xpt_path_sbuf(struct cam_path *path, struct sbuf *sb)
{
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
if (path == NULL)
sbuf_printf(sb, "(nopath): ");
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
else {
if (path->periph != NULL)
sbuf_printf(sb, "(%s%d:", path->periph->periph_name,
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
path->periph->unit_number);
else
sbuf_printf(sb, "(noperiph:");
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
if (path->bus != NULL)
sbuf_printf(sb, "%s%d:%d:", path->bus->sim->sim_name,
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
path->bus->sim->unit_number,
path->bus->sim->bus_id);
else
sbuf_printf(sb, "nobus:");
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
if (path->target != NULL)
sbuf_printf(sb, "%d:", path->target->target_id);
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
else
sbuf_printf(sb, "X:");
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
if (path->device != NULL)
sbuf_printf(sb, "%jx): ",
(uintmax_t)path->device->lun_id);
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
else
sbuf_printf(sb, "X): ");
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
}
return(sbuf_len(sb));
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
}
path_id_t
xpt_path_path_id(struct cam_path *path)
{
return(path->bus->path_id);
}
target_id_t
xpt_path_target_id(struct cam_path *path)
{
if (path->target != NULL)
return (path->target->target_id);
else
return (CAM_TARGET_WILDCARD);
}
lun_id_t
xpt_path_lun_id(struct cam_path *path)
{
if (path->device != NULL)
return (path->device->lun_id);
else
return (CAM_LUN_WILDCARD);
}
struct cam_sim *
xpt_path_sim(struct cam_path *path)
{
return (path->bus->sim);
}
struct cam_periph*
xpt_path_periph(struct cam_path *path)
{
return (path->periph);
}
/*
* Release a CAM control block for the caller. Remit the cost of the structure
* to the device referenced by the path. If the this device had no 'credits'
* and peripheral drivers have registered async callbacks for this notification
* call them now.
*/
void
xpt_release_ccb(union ccb *free_ccb)
{
struct cam_ed *device;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_periph *periph;
CAM_DEBUG_PRINT(CAM_DEBUG_XPT, ("xpt_release_ccb\n"));
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_path_assert(free_ccb->ccb_h.path, MA_OWNED);
device = free_ccb->ccb_h.path->device;
periph = free_ccb->ccb_h.path->periph;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_free_ccb(free_ccb);
periph->periph_allocated--;
cam_ccbq_release_opening(&device->ccbq);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_run_allocq(periph, 0);
}
/* Functions accessed by SIM drivers */
static struct xpt_xport_ops xport_default_ops = {
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
.alloc_device = xpt_alloc_device_default,
.action = xpt_action_default,
.async = xpt_dev_async_default,
};
static struct xpt_xport xport_default = {
.xport = XPORT_UNKNOWN,
.name = "unknown",
.ops = &xport_default_ops,
};
CAM_XPT_XPORT(xport_default);
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
/*
* A sim structure, listing the SIM entry points and instance
* identification info is passed to xpt_bus_register to hook the SIM
* into the CAM framework. xpt_bus_register creates a cam_eb entry
* for this new bus and places it in the array of buses and assigns
* it a path_id. The path_id may be influenced by "hard wiring"
* information specified by the user. Once interrupt services are
* available, the bus will be probed.
*/
int32_t
xpt_bus_register(struct cam_sim *sim, device_t parent, u_int32_t bus)
{
struct cam_eb *new_bus;
struct cam_eb *old_bus;
struct ccb_pathinq cpi;
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
struct cam_path *path;
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
cam_status status;
sim->bus_id = bus;
new_bus = (struct cam_eb *)malloc(sizeof(*new_bus),
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
M_CAMXPT, M_NOWAIT|M_ZERO);
if (new_bus == NULL) {
/* Couldn't satisfy request */
return (CAM_RESRC_UNAVAIL);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_init(&new_bus->eb_mtx, "CAM bus lock", NULL, MTX_DEF);
TAILQ_INIT(&new_bus->et_entries);
cam_sim_hold(sim);
new_bus->sim = sim;
timevalclear(&new_bus->last_reset);
new_bus->flags = 0;
new_bus->refcount = 1; /* Held until a bus_deregister event */
new_bus->generation = 0;
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
xpt_lock_buses();
sim->path_id = new_bus->path_id =
xptpathid(sim->sim_name, sim->unit_number, sim->bus_id);
old_bus = TAILQ_FIRST(&xsoftc.xpt_busses);
while (old_bus != NULL
&& old_bus->path_id < new_bus->path_id)
old_bus = TAILQ_NEXT(old_bus, links);
if (old_bus != NULL)
TAILQ_INSERT_BEFORE(old_bus, new_bus, links);
else
TAILQ_INSERT_TAIL(&xsoftc.xpt_busses, new_bus, links);
xsoftc.bus_generation++;
xpt_unlock_buses();
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
/*
* Set a default transport so that a PATH_INQ can be issued to
* the SIM. This will then allow for probing and attaching of
* a more appropriate transport.
*/
new_bus->xport = &xport_default;
status = xpt_create_path(&path, /*periph*/NULL, sim->path_id,
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD);
if (status != CAM_REQ_CMP) {
xpt_release_bus(new_bus);
return (CAM_RESRC_UNAVAIL);
}
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
xpt_path_inq(&cpi, path);
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
if (cpi.ccb_h.status == CAM_REQ_CMP) {
struct xpt_xport **xpt;
SET_FOREACH(xpt, cam_xpt_xport_set) {
if ((*xpt)->xport == cpi.transport) {
new_bus->xport = *xpt;
break;
}
}
if (new_bus->xport == NULL) {
xpt_print(path,
"No transport found for %d\n", cpi.transport);
xpt_release_bus(new_bus);
free(path, M_CAMXPT);
return (CAM_RESRC_UNAVAIL);
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
}
}
/* Notify interested parties */
if (sim->path_id != CAM_XPT_PATH_ID) {
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_async(AC_PATH_REGISTERED, path, &cpi);
CAM and mps(4) driver scanning changes. Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM not to perform a rescan on a bus when it is registered. We now use this flag in the mps(4) driver. Since it knows what devices it has attached, it is more efficient for it to just issue a target rescan on the targets that are attached. Also, remove the private rescan thread from the mps(4) driver in favor of the rescan thread already built into CAM. Without this change, but with the change above, the MPS scanner could run before or during CAM's initial setup, which would cause duplicate device reprobes and announcements. sys/param.h: Bump __FreeBSD_version to 1000039 for the inclusion of the PIM_RESCAN CAM path inquiry flag. sys/cam/cam_ccb.h: sys/cam/cam_xpt.c: Added a PIM_NOSCAN flag. If a SIM sets this in the path inquiry ccb, then CAM won't rescan the bus in xpt_bus_regsister. sys/dev/mps/mps_sas.c For versions of FreeBSD that have the PIM_NOSCAN path inquiry flag, don't freeze the sim queue during scanning, because CAM won't be scanning this bus. Instead, hold up the boot. Don't call mpssas_rescan_target in mpssas_startup_decrement; it's redundant and I don't know why it was in there. Set PIM_NOSCAN in path inquiry CCBs. Remove methods related to the internal rescan daemon. Always use async events to trigger a probe for EEDP support. In older versions of FreeBSD where AC_ADVINFO_CHANGED is not available, use AC_FOUND_DEVICE and issue the necessary READ CAPACITY manually. Provide a path to xpt_register_async() so that we only receive events for our own SCSI domain. Improve error reporting in cases where setup for EEDP detection fails. sys/dev/mps/mps_sas.h: Remove softc flags and data related to the scanner thread. sys/dev/mps/mps_sas_lsi.c: Unconditionally rescan the target whenever a device is added. Sponsored by: Spectra Logic MFC after: 1 week
2013-07-22 18:37:07 +00:00
if ((cpi.hba_misc & PIM_NOSCAN) == 0) {
union ccb *scan_ccb;
/* Initiate bus rescan. */
scan_ccb = xpt_alloc_ccb_nowait();
if (scan_ccb != NULL) {
scan_ccb->ccb_h.path = path;
scan_ccb->ccb_h.func_code = XPT_SCAN_BUS;
scan_ccb->crcn.flags = 0;
xpt_rescan(scan_ccb);
} else {
CAM and mps(4) driver scanning changes. Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM not to perform a rescan on a bus when it is registered. We now use this flag in the mps(4) driver. Since it knows what devices it has attached, it is more efficient for it to just issue a target rescan on the targets that are attached. Also, remove the private rescan thread from the mps(4) driver in favor of the rescan thread already built into CAM. Without this change, but with the change above, the MPS scanner could run before or during CAM's initial setup, which would cause duplicate device reprobes and announcements. sys/param.h: Bump __FreeBSD_version to 1000039 for the inclusion of the PIM_RESCAN CAM path inquiry flag. sys/cam/cam_ccb.h: sys/cam/cam_xpt.c: Added a PIM_NOSCAN flag. If a SIM sets this in the path inquiry ccb, then CAM won't rescan the bus in xpt_bus_regsister. sys/dev/mps/mps_sas.c For versions of FreeBSD that have the PIM_NOSCAN path inquiry flag, don't freeze the sim queue during scanning, because CAM won't be scanning this bus. Instead, hold up the boot. Don't call mpssas_rescan_target in mpssas_startup_decrement; it's redundant and I don't know why it was in there. Set PIM_NOSCAN in path inquiry CCBs. Remove methods related to the internal rescan daemon. Always use async events to trigger a probe for EEDP support. In older versions of FreeBSD where AC_ADVINFO_CHANGED is not available, use AC_FOUND_DEVICE and issue the necessary READ CAPACITY manually. Provide a path to xpt_register_async() so that we only receive events for our own SCSI domain. Improve error reporting in cases where setup for EEDP detection fails. sys/dev/mps/mps_sas.h: Remove softc flags and data related to the scanner thread. sys/dev/mps/mps_sas_lsi.c: Unconditionally rescan the target whenever a device is added. Sponsored by: Spectra Logic MFC after: 1 week
2013-07-22 18:37:07 +00:00
xpt_print(path,
"Can't allocate CCB to scan bus\n");
xpt_free_path(path);
}
} else
CAM and mps(4) driver scanning changes. Add a PIM_NOSCAN flag to the CAM path inquiry CCB. This tells CAM not to perform a rescan on a bus when it is registered. We now use this flag in the mps(4) driver. Since it knows what devices it has attached, it is more efficient for it to just issue a target rescan on the targets that are attached. Also, remove the private rescan thread from the mps(4) driver in favor of the rescan thread already built into CAM. Without this change, but with the change above, the MPS scanner could run before or during CAM's initial setup, which would cause duplicate device reprobes and announcements. sys/param.h: Bump __FreeBSD_version to 1000039 for the inclusion of the PIM_RESCAN CAM path inquiry flag. sys/cam/cam_ccb.h: sys/cam/cam_xpt.c: Added a PIM_NOSCAN flag. If a SIM sets this in the path inquiry ccb, then CAM won't rescan the bus in xpt_bus_regsister. sys/dev/mps/mps_sas.c For versions of FreeBSD that have the PIM_NOSCAN path inquiry flag, don't freeze the sim queue during scanning, because CAM won't be scanning this bus. Instead, hold up the boot. Don't call mpssas_rescan_target in mpssas_startup_decrement; it's redundant and I don't know why it was in there. Set PIM_NOSCAN in path inquiry CCBs. Remove methods related to the internal rescan daemon. Always use async events to trigger a probe for EEDP support. In older versions of FreeBSD where AC_ADVINFO_CHANGED is not available, use AC_FOUND_DEVICE and issue the necessary READ CAPACITY manually. Provide a path to xpt_register_async() so that we only receive events for our own SCSI domain. Improve error reporting in cases where setup for EEDP detection fails. sys/dev/mps/mps_sas.h: Remove softc flags and data related to the scanner thread. sys/dev/mps/mps_sas_lsi.c: Unconditionally rescan the target whenever a device is added. Sponsored by: Spectra Logic MFC after: 1 week
2013-07-22 18:37:07 +00:00
xpt_free_path(path);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
} else
xpt_free_path(path);
return (CAM_SUCCESS);
}
int32_t
xpt_bus_deregister(path_id_t pathid)
{
struct cam_path bus_path;
cam_status status;
status = xpt_compile_path(&bus_path, NULL, pathid,
CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD);
if (status != CAM_REQ_CMP)
return (status);
xpt_async(AC_LOST_DEVICE, &bus_path, NULL);
xpt_async(AC_PATH_DEREGISTERED, &bus_path, NULL);
/* Release the reference count held while registered. */
xpt_release_bus(bus_path.bus);
xpt_release_path(&bus_path);
return (CAM_REQ_CMP);
}
static path_id_t
xptnextfreepathid(void)
{
struct cam_eb *bus;
path_id_t pathid;
const char *strval;
mtx_assert(&xsoftc.xpt_topo_lock, MA_OWNED);
pathid = 0;
bus = TAILQ_FIRST(&xsoftc.xpt_busses);
retry:
/* Find an unoccupied pathid */
while (bus != NULL && bus->path_id <= pathid) {
if (bus->path_id == pathid)
pathid++;
bus = TAILQ_NEXT(bus, links);
}
/*
* Ensure that this pathid is not reserved for
* a bus that may be registered in the future.
*/
if (resource_string_value("scbus", pathid, "at", &strval) == 0) {
++pathid;
/* Start the search over */
goto retry;
}
return (pathid);
}
static path_id_t
xptpathid(const char *sim_name, int sim_unit, int sim_bus)
{
path_id_t pathid;
int i, dunit, val;
char buf[32];
const char *dname;
pathid = CAM_XPT_PATH_ID;
snprintf(buf, sizeof(buf), "%s%d", sim_name, sim_unit);
if (strcmp(buf, "xpt0") == 0 && sim_bus == 0)
return (pathid);
i = 0;
while ((resource_find_match(&i, &dname, &dunit, "at", buf)) == 0) {
if (strcmp(dname, "scbus")) {
/* Avoid a bit of foot shooting. */
continue;
}
if (dunit < 0) /* unwired?! */
continue;
if (resource_int_value("scbus", dunit, "bus", &val) == 0) {
if (sim_bus == val) {
pathid = dunit;
break;
}
} else if (sim_bus == 0) {
/* Unspecified matches bus 0 */
pathid = dunit;
break;
} else {
printf("Ambiguous scbus configuration for %s%d "
"bus %d, cannot wire down. The kernel "
"config entry for scbus%d should "
"specify a controller bus.\n"
"Scbus will be assigned dynamically.\n",
sim_name, sim_unit, sim_bus, dunit);
break;
}
}
if (pathid == CAM_XPT_PATH_ID)
pathid = xptnextfreepathid();
return (pathid);
}
static const char *
xpt_async_string(u_int32_t async_code)
{
switch (async_code) {
case AC_BUS_RESET: return ("AC_BUS_RESET");
case AC_UNSOL_RESEL: return ("AC_UNSOL_RESEL");
case AC_SCSI_AEN: return ("AC_SCSI_AEN");
case AC_SENT_BDR: return ("AC_SENT_BDR");
case AC_PATH_REGISTERED: return ("AC_PATH_REGISTERED");
case AC_PATH_DEREGISTERED: return ("AC_PATH_DEREGISTERED");
case AC_FOUND_DEVICE: return ("AC_FOUND_DEVICE");
case AC_LOST_DEVICE: return ("AC_LOST_DEVICE");
case AC_TRANSFER_NEG: return ("AC_TRANSFER_NEG");
case AC_INQ_CHANGED: return ("AC_INQ_CHANGED");
case AC_GETDEV_CHANGED: return ("AC_GETDEV_CHANGED");
case AC_CONTRACT: return ("AC_CONTRACT");
case AC_ADVINFO_CHANGED: return ("AC_ADVINFO_CHANGED");
Implement media change notification for DA and CD removable media devices. It includes three parts: 1) Modifications to CAM to detect media media changes and report them to disk(9) layer. For modern SATA (and potentially UAS) devices it utilizes Asynchronous Notification mechanism to receive events from hardware. Active polling with TEST UNIT READY commands with 3 seconds period is used for incapable hardware. After that both CD and DA drivers work the same way, detecting two conditions: "NOT READY: Medium not present" after medium was detected previously, and "UNIT ATTENTION: Not ready to ready change, medium may have changed". First one reported to disk(9) as media removal, second as media insert/change. To reliably receive second event new AC_UNIT_ATTENTION async added to make UAs broadcasted to all periphs by generic error handling code in cam_periph_error(). 2) Modifications to GEOM core to handle media remove and change events. Media removal handled by spoiling all consumers attached to the provider. Media change event also schedules provider retaste after spoiling to probe new media. New flag G_CF_ORPHAN was added to consumers to reflect that consumer is in process of destruction. It allows retaste to create new geom instance of the same class, while previous one is still dying. 3) Modifications to some GEOM classes: DEV -- to report media change events to devd; VFS -- to handle spoiling same as orphan to prevent accessing replaced media. PART class already handles spoiling alike to orphan. Reviewed by: silence on geom@ and scsi@ Tested by: avg Sponsored by: iXsystems, Inc. / PC-BSD MFC after: 2 months
2012-07-29 11:51:48 +00:00
case AC_UNIT_ATTENTION: return ("AC_UNIT_ATTENTION");
}
return ("AC_UNKNOWN");
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static int
xpt_async_size(u_int32_t async_code)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
switch (async_code) {
case AC_BUS_RESET: return (0);
case AC_UNSOL_RESEL: return (0);
case AC_SCSI_AEN: return (0);
case AC_SENT_BDR: return (0);
case AC_PATH_REGISTERED: return (sizeof(struct ccb_pathinq));
case AC_PATH_DEREGISTERED: return (0);
case AC_FOUND_DEVICE: return (sizeof(struct ccb_getdev));
case AC_LOST_DEVICE: return (0);
case AC_TRANSFER_NEG: return (sizeof(struct ccb_trans_settings));
case AC_INQ_CHANGED: return (0);
case AC_GETDEV_CHANGED: return (0);
case AC_CONTRACT: return (sizeof(struct ac_contract));
case AC_ADVINFO_CHANGED: return (-1);
case AC_UNIT_ATTENTION: return (sizeof(struct ccb_scsiio));
}
return (0);
}
static int
xpt_async_process_dev(struct cam_ed *device, void *arg)
{
union ccb *ccb = arg;
struct cam_path *path = ccb->ccb_h.path;
void *async_arg = ccb->casync.async_arg_ptr;
u_int32_t async_code = ccb->casync.async_code;
int relock;
if (path->device != device
&& path->device->lun_id != CAM_LUN_WILDCARD
&& device->lun_id != CAM_LUN_WILDCARD)
return (1);
/*
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
* The async callback could free the device.
* If it is a broadcast async, it doesn't hold
* device reference, so take our own reference.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_acquire_device(device);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
/*
* If async for specific device is to be delivered to
* the wildcard client, take the specific device lock.
* XXX: We may need a way for client to specify it.
*/
if ((device->lun_id == CAM_LUN_WILDCARD &&
path->device->lun_id != CAM_LUN_WILDCARD) ||
(device->target->target_id == CAM_TARGET_WILDCARD &&
path->target->target_id != CAM_TARGET_WILDCARD) ||
(device->target->bus->path_id == CAM_BUS_WILDCARD &&
path->target->bus->path_id != CAM_BUS_WILDCARD)) {
mtx_unlock(&device->device_mtx);
xpt_path_lock(path);
relock = 1;
} else
relock = 0;
(*(device->target->bus->xport->ops->async))(async_code,
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
device->target->bus, device->target, device, async_arg);
xpt_async_bcast(&device->asyncs, async_code, path, async_arg);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (relock) {
xpt_path_unlock(path);
mtx_lock(&device->device_mtx);
}
xpt_release_device(device);
return (1);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static int
xpt_async_process_tgt(struct cam_et *target, void *arg)
{
union ccb *ccb = arg;
struct cam_path *path = ccb->ccb_h.path;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (path->target != target
&& path->target->target_id != CAM_TARGET_WILDCARD
&& target->target_id != CAM_TARGET_WILDCARD)
return (1);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (ccb->casync.async_code == AC_SENT_BDR) {
/* Update our notion of when the last reset occurred */
microtime(&target->last_reset);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
return (xptdevicetraverse(target, NULL, xpt_async_process_dev, ccb));
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static void
xpt_async_process(struct cam_periph *periph, union ccb *ccb)
{
struct cam_eb *bus;
struct cam_path *path;
void *async_arg;
u_int32_t async_code;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
path = ccb->ccb_h.path;
async_code = ccb->casync.async_code;
async_arg = ccb->casync.async_arg_ptr;
CAM_DEBUG(path, CAM_DEBUG_TRACE | CAM_DEBUG_INFO,
("xpt_async(%s)\n", xpt_async_string(async_code)));
bus = path->bus;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (async_code == AC_BUS_RESET) {
/* Update our notion of when the last reset occurred */
microtime(&bus->last_reset);
}
2008-01-02 01:45:31 +00:00
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpttargettraverse(bus, NULL, xpt_async_process_tgt, ccb);
/*
* If this wasn't a fully wildcarded async, tell all
* clients that want all async events.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (bus != xpt_periph->path->bus) {
xpt_path_lock(xpt_periph->path);
xpt_async_process_dev(xpt_periph->path->device, ccb);
xpt_path_unlock(xpt_periph->path);
}
if (path->device != NULL && path->device->lun_id != CAM_LUN_WILDCARD)
xpt_release_devq(path, 1, TRUE);
else
xpt_release_simq(path->bus->sim, TRUE);
if (ccb->casync.async_arg_size > 0)
free(async_arg, M_CAMXPT);
xpt_free_path(path);
xpt_free_ccb(ccb);
}
static void
xpt_async_bcast(struct async_list *async_head,
u_int32_t async_code,
struct cam_path *path, void *async_arg)
{
struct async_node *cur_entry;
struct mtx *mtx;
cur_entry = SLIST_FIRST(async_head);
while (cur_entry != NULL) {
struct async_node *next_entry;
/*
* Grab the next list entry before we call the current
* entry's callback. This is because the callback function
* can delete its async callback entry.
*/
next_entry = SLIST_NEXT(cur_entry, links);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if ((cur_entry->event_enable & async_code) != 0) {
mtx = cur_entry->event_lock ?
path->device->sim->mtx : NULL;
if (mtx)
mtx_lock(mtx);
cur_entry->callback(cur_entry->callback_arg,
async_code, path,
async_arg);
if (mtx)
mtx_unlock(mtx);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
cur_entry = next_entry;
}
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
void
xpt_async(u_int32_t async_code, struct cam_path *path, void *async_arg)
{
union ccb *ccb;
int size;
ccb = xpt_alloc_ccb_nowait();
if (ccb == NULL) {
xpt_print(path, "Can't allocate CCB to send %s\n",
xpt_async_string(async_code));
return;
}
if (xpt_clone_path(&ccb->ccb_h.path, path) != CAM_REQ_CMP) {
xpt_print(path, "Can't allocate path to send %s\n",
xpt_async_string(async_code));
xpt_free_ccb(ccb);
return;
}
ccb->ccb_h.path->periph = NULL;
ccb->ccb_h.func_code = XPT_ASYNC;
ccb->ccb_h.cbfcnp = xpt_async_process;
ccb->ccb_h.flags |= CAM_UNLOCKED;
ccb->casync.async_code = async_code;
ccb->casync.async_arg_size = 0;
size = xpt_async_size(async_code);
CAM_DEBUG(ccb->ccb_h.path, CAM_DEBUG_TRACE,
("xpt_async: func %#x %s aync_code %d %s\n",
ccb->ccb_h.func_code,
xpt_action_name(ccb->ccb_h.func_code),
async_code,
xpt_async_string(async_code)));
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (size > 0 && async_arg != NULL) {
ccb->casync.async_arg_ptr = malloc(size, M_CAMXPT, M_NOWAIT);
if (ccb->casync.async_arg_ptr == NULL) {
xpt_print(path, "Can't allocate argument to send %s\n",
xpt_async_string(async_code));
xpt_free_path(ccb->ccb_h.path);
xpt_free_ccb(ccb);
return;
}
memcpy(ccb->casync.async_arg_ptr, async_arg, size);
ccb->casync.async_arg_size = size;
} else if (size < 0) {
ccb->casync.async_arg_ptr = async_arg;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
ccb->casync.async_arg_size = size;
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (path->device != NULL && path->device->lun_id != CAM_LUN_WILDCARD)
xpt_freeze_devq(path, 1);
else
xpt_freeze_simq(path->bus->sim, 1);
xpt_action(ccb);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
static void
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
xpt_dev_async_default(u_int32_t async_code, struct cam_eb *bus,
struct cam_et *target, struct cam_ed *device,
void *async_arg)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
/*
* We only need to handle events for real devices.
*/
if (target->target_id == CAM_TARGET_WILDCARD
|| device->lun_id == CAM_LUN_WILDCARD)
return;
printf("%s called\n", __func__);
}
static uint32_t
xpt_freeze_devq_device(struct cam_ed *dev, u_int count)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_devq *devq;
uint32_t freeze;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
devq = dev->sim->devq;
mtx_assert(&devq->send_mtx, MA_OWNED);
CAM_DEBUG_DEV(dev, CAM_DEBUG_TRACE,
("xpt_freeze_devq_device(%d) %u->%u\n", count,
dev->ccbq.queue.qfrozen_cnt, dev->ccbq.queue.qfrozen_cnt + count));
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
freeze = (dev->ccbq.queue.qfrozen_cnt += count);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
/* Remove frozen device from sendq. */
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (device_is_queued(dev))
camq_remove(&devq->send_queue, dev->devq_entry.index);
return (freeze);
}
u_int32_t
xpt_freeze_devq(struct cam_path *path, u_int count)
{
struct cam_ed *dev = path->device;
struct cam_devq *devq;
uint32_t freeze;
devq = dev->sim->devq;
mtx_lock(&devq->send_mtx);
CAM_DEBUG(path, CAM_DEBUG_TRACE, ("xpt_freeze_devq(%d)\n", count));
freeze = xpt_freeze_devq_device(dev, count);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&devq->send_mtx);
return (freeze);
}
u_int32_t
xpt_freeze_simq(struct cam_sim *sim, u_int count)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_devq *devq;
uint32_t freeze;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
devq = sim->devq;
mtx_lock(&devq->send_mtx);
freeze = (devq->send_queue.qfrozen_cnt += count);
mtx_unlock(&devq->send_mtx);
return (freeze);
}
static void
xpt_release_devq_timeout(void *arg)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_ed *dev;
struct cam_devq *devq;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
dev = (struct cam_ed *)arg;
CAM_DEBUG_DEV(dev, CAM_DEBUG_TRACE, ("xpt_release_devq_timeout\n"));
devq = dev->sim->devq;
mtx_assert(&devq->send_mtx, MA_OWNED);
if (xpt_release_devq_device(dev, /*count*/1, /*run_queue*/TRUE))
xpt_run_devq(devq);
}
void
xpt_release_devq(struct cam_path *path, u_int count, int run_queue)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_ed *dev;
struct cam_devq *devq;
CAM_DEBUG(path, CAM_DEBUG_TRACE, ("xpt_release_devq(%d, %d)\n",
count, run_queue));
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
dev = path->device;
devq = dev->sim->devq;
mtx_lock(&devq->send_mtx);
if (xpt_release_devq_device(dev, count, run_queue))
xpt_run_devq(dev->sim->devq);
mtx_unlock(&devq->send_mtx);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static int
xpt_release_devq_device(struct cam_ed *dev, u_int count, int run_queue)
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_assert(&dev->sim->devq->send_mtx, MA_OWNED);
CAM_DEBUG_DEV(dev, CAM_DEBUG_TRACE,
("xpt_release_devq_device(%d, %d) %u->%u\n", count, run_queue,
dev->ccbq.queue.qfrozen_cnt, dev->ccbq.queue.qfrozen_cnt - count));
if (count > dev->ccbq.queue.qfrozen_cnt) {
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
#ifdef INVARIANTS
printf("xpt_release_devq(): requested %u > present %u\n",
count, dev->ccbq.queue.qfrozen_cnt);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
#endif
count = dev->ccbq.queue.qfrozen_cnt;
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
}
dev->ccbq.queue.qfrozen_cnt -= count;
if (dev->ccbq.queue.qfrozen_cnt == 0) {
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
/*
* No longer need to wait for a successful
* command completion.
*/
dev->flags &= ~CAM_DEV_REL_ON_COMPLETE;
/*
* Remove any timeouts that might be scheduled
* to release this queue.
*/
if ((dev->flags & CAM_DEV_REL_TIMEOUT_PENDING) != 0) {
callout_stop(&dev->callout);
dev->flags &= ~CAM_DEV_REL_TIMEOUT_PENDING;
}
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
/*
* Now that we are unfrozen schedule the
* device so any pending transactions are
* run.
*/
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_schedule_devq(dev->sim->devq, dev);
} else
run_queue = 0;
return (run_queue);
}
void
xpt_release_simq(struct cam_sim *sim, int run_queue)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_devq *devq;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
devq = sim->devq;
mtx_lock(&devq->send_mtx);
if (devq->send_queue.qfrozen_cnt <= 0) {
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
#ifdef INVARIANTS
printf("xpt_release_simq: requested 1 > present %u\n",
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
devq->send_queue.qfrozen_cnt);
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
#endif
} else
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
devq->send_queue.qfrozen_cnt--;
if (devq->send_queue.qfrozen_cnt == 0) {
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
/*
* If there is a timeout scheduled to release this
* sim queue, remove it. The queue frozen count is
* already at 0.
*/
if ((sim->flags & CAM_SIM_REL_TIMEOUT_PENDING) != 0){
callout_stop(&sim->callout);
sim->flags &= ~CAM_SIM_REL_TIMEOUT_PENDING;
}
if (run_queue) {
/*
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
* Now that we are unfrozen run the send queue.
*/
xpt_run_devq(sim->devq);
}
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&devq->send_mtx);
}
void
xpt_done(union ccb *done_ccb)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_doneq *queue;
int run, hash;
#if defined(BUF_TRACKING) || defined(FULL_BUF_TRACKING)
if (done_ccb->ccb_h.func_code == XPT_SCSI_IO &&
done_ccb->csio.bio != NULL)
biotrack(done_ccb->csio.bio, __func__);
#endif
CAM_DEBUG(done_ccb->ccb_h.path, CAM_DEBUG_TRACE,
("xpt_done: func= %#x %s status %#x\n",
done_ccb->ccb_h.func_code,
xpt_action_name(done_ccb->ccb_h.func_code),
done_ccb->ccb_h.status));
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if ((done_ccb->ccb_h.func_code & XPT_FC_QUEUED) == 0)
return;
/* Store the time the ccb was in the sim */
done_ccb->ccb_h.qos.periph_data = cam_iosched_delta_t(done_ccb->ccb_h.qos.periph_data);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
hash = (done_ccb->ccb_h.path_id + done_ccb->ccb_h.target_id +
done_ccb->ccb_h.target_lun) % cam_num_doneqs;
queue = &cam_doneqs[hash];
mtx_lock(&queue->cam_doneq_mtx);
run = (queue->cam_doneq_sleep && STAILQ_EMPTY(&queue->cam_doneq));
STAILQ_INSERT_TAIL(&queue->cam_doneq, &done_ccb->ccb_h, sim_links.stqe);
done_ccb->ccb_h.pinfo.index = CAM_DONEQ_INDEX;
mtx_unlock(&queue->cam_doneq_mtx);
if (run)
wakeup(&queue->cam_doneq);
}
void
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_done_direct(union ccb *done_ccb)
{
CAM_DEBUG(done_ccb->ccb_h.path, CAM_DEBUG_TRACE,
("xpt_done_direct: status %#x\n", done_ccb->ccb_h.status));
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if ((done_ccb->ccb_h.func_code & XPT_FC_QUEUED) == 0)
return;
/* Store the time the ccb was in the sim */
done_ccb->ccb_h.qos.periph_data = cam_iosched_delta_t(done_ccb->ccb_h.qos.periph_data);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_done_process(&done_ccb->ccb_h);
}
union ccb *
xpt_alloc_ccb(void)
{
union ccb *new_ccb;
new_ccb = malloc(sizeof(*new_ccb), M_CAMCCB, M_ZERO|M_WAITOK);
return (new_ccb);
}
union ccb *
xpt_alloc_ccb_nowait(void)
{
union ccb *new_ccb;
new_ccb = malloc(sizeof(*new_ccb), M_CAMCCB, M_ZERO|M_NOWAIT);
return (new_ccb);
}
void
xpt_free_ccb(union ccb *free_ccb)
{
free(free_ccb, M_CAMCCB);
}
/* Private XPT functions */
/*
* Get a CAM control block for the caller. Charge the structure to the device
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
* referenced by the path. If we don't have sufficient resources to allocate
* more ccbs, we return NULL.
*/
static union ccb *
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_get_ccb_nowait(struct cam_periph *periph)
{
union ccb *new_ccb;
new_ccb = malloc(sizeof(*new_ccb), M_CAMCCB, M_ZERO|M_NOWAIT);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (new_ccb == NULL)
return (NULL);
periph->periph_allocated++;
cam_ccbq_take_opening(&periph->path->device->ccbq);
return (new_ccb);
}
static union ccb *
xpt_get_ccb(struct cam_periph *periph)
{
union ccb *new_ccb;
cam_periph_unlock(periph);
new_ccb = malloc(sizeof(*new_ccb), M_CAMCCB, M_ZERO|M_WAITOK);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
cam_periph_lock(periph);
periph->periph_allocated++;
cam_ccbq_take_opening(&periph->path->device->ccbq);
return (new_ccb);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
union ccb *
cam_periph_getccb(struct cam_periph *periph, u_int32_t priority)
{
struct ccb_hdr *ccb_h;
CAM_DEBUG(periph->path, CAM_DEBUG_TRACE, ("cam_periph_getccb\n"));
cam_periph_assert(periph, MA_OWNED);
while ((ccb_h = SLIST_FIRST(&periph->ccb_list)) == NULL ||
ccb_h->pinfo.priority != priority) {
if (priority < periph->immediate_priority) {
periph->immediate_priority = priority;
xpt_run_allocq(periph, 0);
} else
cam_periph_sleep(periph, &periph->ccb_list, PRIBIO,
"cgticb", 0);
}
SLIST_REMOVE_HEAD(&periph->ccb_list, periph_links.sle);
return ((union ccb *)ccb_h);
}
static void
xpt_acquire_bus(struct cam_eb *bus)
{
xpt_lock_buses();
bus->refcount++;
xpt_unlock_buses();
}
static void
xpt_release_bus(struct cam_eb *bus)
{
xpt_lock_buses();
KASSERT(bus->refcount >= 1, ("bus->refcount >= 1"));
if (--bus->refcount > 0) {
xpt_unlock_buses();
return;
}
TAILQ_REMOVE(&xsoftc.xpt_busses, bus, links);
xsoftc.bus_generation++;
xpt_unlock_buses();
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
KASSERT(TAILQ_EMPTY(&bus->et_entries),
("destroying bus, but target list is not empty"));
cam_sim_release(bus->sim);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_destroy(&bus->eb_mtx);
free(bus, M_CAMXPT);
}
static struct cam_et *
xpt_alloc_target(struct cam_eb *bus, target_id_t target_id)
{
struct cam_et *cur_target, *target;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_assert(&xsoftc.xpt_topo_lock, MA_OWNED);
mtx_assert(&bus->eb_mtx, MA_OWNED);
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
target = (struct cam_et *)malloc(sizeof(*target), M_CAMXPT,
M_NOWAIT|M_ZERO);
if (target == NULL)
return (NULL);
TAILQ_INIT(&target->ed_entries);
target->bus = bus;
target->target_id = target_id;
target->refcount = 1;
target->generation = 0;
target->luns = NULL;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_init(&target->luns_mtx, "CAM LUNs lock", NULL, MTX_DEF);
timevalclear(&target->last_reset);
/*
* Hold a reference to our parent bus so it
* will not go away before we do.
*/
bus->refcount++;
/* Insertion sort into our bus's target list */
cur_target = TAILQ_FIRST(&bus->et_entries);
while (cur_target != NULL && cur_target->target_id < target_id)
cur_target = TAILQ_NEXT(cur_target, links);
if (cur_target != NULL) {
TAILQ_INSERT_BEFORE(cur_target, target, links);
} else {
TAILQ_INSERT_TAIL(&bus->et_entries, target, links);
}
bus->generation++;
return (target);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static void
xpt_acquire_target(struct cam_et *target)
{
struct cam_eb *bus = target->bus;
mtx_lock(&bus->eb_mtx);
target->refcount++;
mtx_unlock(&bus->eb_mtx);
}
static void
xpt_release_target(struct cam_et *target)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_eb *bus = target->bus;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&bus->eb_mtx);
if (--target->refcount > 0) {
mtx_unlock(&bus->eb_mtx);
return;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
TAILQ_REMOVE(&bus->et_entries, target, links);
bus->generation++;
mtx_unlock(&bus->eb_mtx);
KASSERT(TAILQ_EMPTY(&target->ed_entries),
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
("destroying target, but device list is not empty"));
xpt_release_bus(bus);
mtx_destroy(&target->luns_mtx);
if (target->luns)
free(target->luns, M_CAMXPT);
free(target, M_CAMXPT);
}
static struct cam_ed *
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
xpt_alloc_device_default(struct cam_eb *bus, struct cam_et *target,
lun_id_t lun_id)
{
struct cam_ed *device;
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
device = xpt_alloc_device(bus, target, lun_id);
if (device == NULL)
return (NULL);
device->mintags = 1;
device->maxtags = 1;
return (device);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
static void
xpt_destroy_device(void *context, int pending)
{
struct cam_ed *device = context;
mtx_lock(&device->device_mtx);
mtx_destroy(&device->device_mtx);
free(device, M_CAMDEV);
}
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
struct cam_ed *
xpt_alloc_device(struct cam_eb *bus, struct cam_et *target, lun_id_t lun_id)
{
struct cam_ed *cur_device, *device;
struct cam_devq *devq;
cam_status status;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_assert(&bus->eb_mtx, MA_OWNED);
/* Make space for us in the device queue on our bus */
devq = bus->sim->devq;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&devq->send_mtx);
status = cam_devq_resize(devq, devq->send_queue.array_size + 1);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&devq->send_mtx);
if (status != CAM_REQ_CMP)
return (NULL);
device = (struct cam_ed *)malloc(sizeof(*device),
M_CAMDEV, M_NOWAIT|M_ZERO);
if (device == NULL)
return (NULL);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
cam_init_pinfo(&device->devq_entry);
device->target = target;
device->lun_id = lun_id;
device->sim = bus->sim;
if (cam_ccbq_init(&device->ccbq,
bus->sim->max_dev_openings) != 0) {
free(device, M_CAMDEV);
return (NULL);
}
SLIST_INIT(&device->asyncs);
SLIST_INIT(&device->periphs);
device->generation = 0;
device->flags = CAM_DEV_UNCONFIGURED;
device->tag_delay_count = 0;
device->tag_saved_openings = 0;
device->refcount = 1;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_init(&device->device_mtx, "CAM device lock", NULL, MTX_DEF);
callout_init_mtx(&device->callout, &devq->send_mtx, 0);
TASK_INIT(&device->device_destroy_task, 0, xpt_destroy_device, device);
/*
* Hold a reference to our parent bus so it
* will not go away before we do.
*/
target->refcount++;
cur_device = TAILQ_FIRST(&target->ed_entries);
while (cur_device != NULL && cur_device->lun_id < lun_id)
cur_device = TAILQ_NEXT(cur_device, links);
if (cur_device != NULL)
TAILQ_INSERT_BEFORE(cur_device, device, links);
else
TAILQ_INSERT_TAIL(&target->ed_entries, device, links);
target->generation++;
return (device);
}
void
xpt_acquire_device(struct cam_ed *device)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_eb *bus = device->target->bus;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&bus->eb_mtx);
device->refcount++;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&bus->eb_mtx);
}
void
xpt_release_device(struct cam_ed *device)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct cam_eb *bus = device->target->bus;
struct cam_devq *devq;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&bus->eb_mtx);
if (--device->refcount > 0) {
mtx_unlock(&bus->eb_mtx);
return;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
TAILQ_REMOVE(&device->target->ed_entries, device,links);
device->target->generation++;
mtx_unlock(&bus->eb_mtx);
/* Release our slot in the devq */
devq = bus->sim->devq;
mtx_lock(&devq->send_mtx);
cam_devq_resize(devq, devq->send_queue.array_size - 1);
Add CAM infrastructure to allow reporting when a drive's long read capacity data changes. cam_ccb.h: Add a new advanced information type, CDAI_TYPE_RCAPLONG, for long read capacity data. cam_xpt_internal.h: Add a read capacity data pointer and length to struct cam_ed. cam_xpt.c: Free the read capacity buffer when a device goes away. While we're here, make sure we don't leak memory for other malloced fields in struct cam_ed. scsi_all.c: Update the scsi_read_capacity_16() to take a uint8_t * and a length instead of just a pointer to the parameter data structure. This will hopefully make this function somewhat immune to future changes in the parameter data. scsi_all.h: Add some extra bit definitions to struct scsi_read_capacity_data_long, and bump up the structure size to the full size specified by SBC-3. Change the prototype for scsi_read_capacity_16(). scsi_da.c: Register changes in read capacity data with the transport layer. This allows the transport layer to send out an async notification to interested parties. Update the dasetgeom() API. Use scsi_extract_sense_len() instead of scsi_extract_sense(). scsi_xpt.c: Add support for the new CDAI_TYPE_RCAPLONG advanced information type. Make sure we set the physpath pointer to NULL after freeing it. This allows blindly freeing it in the struct cam_ed destructor. sys/param.h: Bump __FreeBSD_version from 1000005 to 1000006 to make it easier for third party drivers to determine that the read capacity data async notification is available. camcontrol.c, mptutil/mpt_cam.c: Update these for the new scsi_read_capacity_16() argument structure. Sponsored by: Spectra Logic
2012-01-26 18:09:28 +00:00
KASSERT(SLIST_EMPTY(&device->periphs),
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
("destroying device, but periphs list is not empty"));
KASSERT(device->devq_entry.index == CAM_UNQUEUED_INDEX,
("destroying device while still queued for ccbs"));
/* The send_mtx must be held when accessing the callout */
if ((device->flags & CAM_DEV_REL_TIMEOUT_PENDING) != 0)
callout_stop(&device->callout);
mtx_unlock(&devq->send_mtx);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_release_target(device->target);
cam_ccbq_fini(&device->ccbq);
/*
* Free allocated memory. free(9) does nothing if the
* supplied pointer is NULL, so it is safe to call without
* checking.
*/
free(device->supported_vpds, M_CAMXPT);
free(device->device_id, M_CAMXPT);
free(device->ext_inq, M_CAMXPT);
free(device->physpath, M_CAMXPT);
free(device->rcap_buf, M_CAMXPT);
free(device->serial_num, M_CAMXPT);
free(device->nvme_data, M_CAMXPT);
free(device->nvme_cdata, M_CAMXPT);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
taskqueue_enqueue(xsoftc.xpt_taskq, &device->device_destroy_task);
}
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
u_int32_t
xpt_dev_ccbq_resize(struct cam_path *path, int newopenings)
{
int result;
struct cam_ed *dev;
dev = path->device;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&dev->sim->devq->send_mtx);
result = cam_ccbq_resize(&dev->ccbq, newopenings);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&dev->sim->devq->send_mtx);
if ((dev->flags & CAM_DEV_TAG_AFTER_COUNT) != 0
|| (dev->inq_flags & SID_CmdQue) != 0)
dev->tag_saved_openings = newopenings;
return (result);
}
static struct cam_eb *
xpt_find_bus(path_id_t path_id)
{
struct cam_eb *bus;
xpt_lock_buses();
for (bus = TAILQ_FIRST(&xsoftc.xpt_busses);
bus != NULL;
bus = TAILQ_NEXT(bus, links)) {
if (bus->path_id == path_id) {
bus->refcount++;
break;
}
}
xpt_unlock_buses();
return (bus);
}
static struct cam_et *
xpt_find_target(struct cam_eb *bus, target_id_t target_id)
{
struct cam_et *target;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_assert(&bus->eb_mtx, MA_OWNED);
for (target = TAILQ_FIRST(&bus->et_entries);
target != NULL;
target = TAILQ_NEXT(target, links)) {
if (target->target_id == target_id) {
target->refcount++;
break;
}
}
return (target);
}
static struct cam_ed *
xpt_find_device(struct cam_et *target, lun_id_t lun_id)
{
struct cam_ed *device;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_assert(&target->bus->eb_mtx, MA_OWNED);
for (device = TAILQ_FIRST(&target->ed_entries);
device != NULL;
device = TAILQ_NEXT(device, links)) {
if (device->lun_id == lun_id) {
device->refcount++;
break;
}
}
return (device);
}
void
xpt_start_tags(struct cam_path *path)
{
struct ccb_relsim crs;
struct cam_ed *device;
struct cam_sim *sim;
int newopenings;
device = path->device;
sim = path->bus->sim;
device->flags &= ~CAM_DEV_TAG_AFTER_COUNT;
xpt_freeze_devq(path, /*count*/1);
device->inq_flags |= SID_CmdQue;
if (device->tag_saved_openings != 0)
newopenings = device->tag_saved_openings;
else
Separate the parallel scsi knowledge out of the core of the XPT, and modularize it so that new transports can be created. Add a transport for SATA Add a periph+protocol layer for ATA Add a driver for AHCI-compliant hardware. Add a maxio field to CAM so that drivers can advertise their max I/O capability. Modify various drivers so that they are insulated from the value of MAXPHYS. The new ATA/SATA code supports AHCI-compliant hardware, and will override the classic ATA driver if it is loaded as a module at boot time or compiled into the kernel. The stack now support NCQ (tagged queueing) for increased performance on modern SATA drives. It also supports port multipliers. ATA drives are accessed via 'ada' device nodes. ATAPI drives are accessed via 'cd' device nodes. They can all be enumerated and manipulated via camcontrol, just like SCSI drives. SCSI commands are not translated to their ATA equivalents; ATA native commands are used throughout the entire stack, including camcontrol. See the camcontrol manpage for further details. Testing this code may require that you update your fstab, and possibly modify your BIOS to enable AHCI functionality, if available. This code is very experimental at the moment. The userland ABI/API has changed, so applications will need to be recompiled. It may change further in the near future. The 'ada' device name may also change as more infrastructure is completed in this project. The goal is to eventually put all CAM busses and devices until newbus, allowing for interesting topology and management options. Few functional changes will be seen with existing SCSI/SAS/FC drivers, though the userland ABI has still changed. In the future, transports specific modules for SAS and FC may appear in order to better support the topologies and capabilities of these technologies. The modularization of CAM and the addition of the ATA/SATA modules is meant to break CAM out of the mold of being specific to SCSI, letting it grow to be a framework for arbitrary transports and protocols. It also allows drivers to be written to support discrete hardware without jeopardizing the stability of non-related hardware. While only an AHCI driver is provided now, a Silicon Image driver is also in the works. Drivers for ICH1-4, ICH5-6, PIIX, classic IDE, and any other hardware is possible and encouraged. Help with new transports is also encouraged. Submitted by: scottl, mav Approved by: re
2009-07-10 08:18:08 +00:00
newopenings = min(device->maxtags,
sim->max_tagged_dev_openings);
xpt_dev_ccbq_resize(path, newopenings);
xpt_async(AC_GETDEV_CHANGED, path, NULL);
xpt_setup_ccb(&crs.ccb_h, path, CAM_PRIORITY_NORMAL);
crs.ccb_h.func_code = XPT_REL_SIMQ;
crs.release_flags = RELSIM_RELEASE_AFTER_QEMPTY;
crs.openings
2008-01-02 01:45:31 +00:00
= crs.release_timeout
= crs.qfrozen_cnt
= 0;
xpt_action((union ccb *)&crs);
}
void
xpt_stop_tags(struct cam_path *path)
{
struct ccb_relsim crs;
struct cam_ed *device;
struct cam_sim *sim;
device = path->device;
sim = path->bus->sim;
device->flags &= ~CAM_DEV_TAG_AFTER_COUNT;
device->tag_delay_count = 0;
xpt_freeze_devq(path, /*count*/1);
device->inq_flags &= ~SID_CmdQue;
xpt_dev_ccbq_resize(path, sim->max_dev_openings);
xpt_async(AC_GETDEV_CHANGED, path, NULL);
xpt_setup_ccb(&crs.ccb_h, path, CAM_PRIORITY_NORMAL);
crs.ccb_h.func_code = XPT_REL_SIMQ;
crs.release_flags = RELSIM_RELEASE_AFTER_QEMPTY;
crs.openings
= crs.release_timeout
= crs.qfrozen_cnt
= 0;
xpt_action((union ccb *)&crs);
}
/*
* Assume all possible buses are detected by this time, so allow boot
* as soon as they all are scanned.
*/
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
static void
xpt_boot_delay(void *arg)
{
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_release_boot();
}
/*
* Now that all config hooks have completed, start boot_delay timer,
* waiting for possibly still undetected buses (USB) to appear.
*/
static void
xpt_ch_done(void *arg)
{
callout_init(&xsoftc.boot_callout, 1);
callout_reset_sbt(&xsoftc.boot_callout, SBT_1MS * xsoftc.boot_delay, 0,
xpt_boot_delay, NULL, 0);
}
SYSINIT(xpt_hw_delay, SI_SUB_INT_CONFIG_HOOKS, SI_ORDER_ANY, xpt_ch_done, NULL);
/*
* Now that interrupts are enabled, go find our devices
*/
static void
xpt_config(void *arg)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (taskqueue_start_threads(&xsoftc.xpt_taskq, 1, PRIBIO, "CAM taskq"))
printf("xpt_config: failed to create taskqueue thread.\n");
/* Setup debugging path */
if (cam_dflags != CAM_DEBUG_NONE) {
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (xpt_create_path(&cam_dpath, NULL,
CAM_DEBUG_BUS, CAM_DEBUG_TARGET,
CAM_DEBUG_LUN) != CAM_REQ_CMP) {
printf("xpt_config: xpt_create_path() failed for debug"
" target %d:%d:%d, debugging disabled\n",
CAM_DEBUG_BUS, CAM_DEBUG_TARGET, CAM_DEBUG_LUN);
cam_dflags = CAM_DEBUG_NONE;
}
} else
cam_dpath = NULL;
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
periphdriver_init(1);
xpt_hold_boot();
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
/* Fire up rescan thread. */
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (kproc_kthread_add(xpt_scanner_thread, NULL, &cam_proc, NULL, 0, 0,
"cam", "scanner")) {
printf("xpt_config: failed to create rescan thread.\n");
}
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
}
void
xpt_hold_boot_locked(void)
{
if (xsoftc.buses_to_config++ == 0)
root_mount_hold_token("CAM", &xsoftc.xpt_rootmount);
}
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
void
xpt_hold_boot(void)
{
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_lock_buses();
xpt_hold_boot_locked();
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_unlock_buses();
}
void
xpt_release_boot(void)
{
xpt_lock_buses();
if (--xsoftc.buses_to_config == 0) {
if (xsoftc.buses_config_done == 0) {
xsoftc.buses_config_done = 1;
xsoftc.buses_to_config++;
TASK_INIT(&xsoftc.boot_task, 0, xpt_finishconfig_task,
NULL);
taskqueue_enqueue(taskqueue_thread, &xsoftc.boot_task);
} else
root_mount_rel(&xsoftc.xpt_rootmount);
}
xpt_unlock_buses();
}
/*
* If the given device only has one peripheral attached to it, and if that
* peripheral is the passthrough driver, announce it. This insures that the
* user sees some sort of announcement for every peripheral in their system.
*/
static int
xptpassannouncefunc(struct cam_ed *device, void *arg)
{
struct cam_periph *periph;
int i;
for (periph = SLIST_FIRST(&device->periphs), i = 0; periph != NULL;
periph = SLIST_NEXT(periph, periph_links), i++);
periph = SLIST_FIRST(&device->periphs);
if ((i == 1)
&& (strncmp(periph->periph_name, "pass", 4) == 0))
xpt_announce_periph(periph, NULL);
return(1);
}
static void
xpt_finishconfig_task(void *context, int pending)
{
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
periphdriver_init(2);
/*
* Check for devices with no "standard" peripheral driver
* attached. For any devices like that, announce the
* passthrough driver so the user will see something.
*/
if (!bootverbose)
xpt_for_all_devices(xptpassannouncefunc, NULL);
xpt_release_boot();
}
cam_status
xpt_register_async(int event, ac_callback_t *cbfunc, void *cbarg,
struct cam_path *path)
{
struct ccb_setasync csa;
cam_status status;
int xptpath = 0;
if (path == NULL) {
status = xpt_create_path(&path, /*periph*/NULL, CAM_XPT_PATH_ID,
CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (status != CAM_REQ_CMP)
return (status);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_path_lock(path);
xptpath = 1;
}
MFp4: Large set of CAM inprovements. - Unify bus reset/probe sequence. Whenever bus attached at boot or later, CAM will automatically reset and scan it. It allows to remove duplicate code from many drivers. - Any bus, attached before CAM completed it's boot-time initialization, will equally join to the process, delaying boot if needed. - New kern.cam.boot_delay loader tunable should help controllers that are still unable to register their buses in time (such as slow USB/ PCCard/ CardBus devices), by adding one more event to wait on boot. - To allow synchronization between different CAM levels, concept of requests priorities was extended. Priorities now split between several "run levels". Device can be freezed at specified level, allowing higher priority requests to pass. For example, no payload requests allowed, until PMP driver enable port. ATA XPT negotiate transfer parameters, periph driver configure caching and so on. - Frozen requests are no more counted by request allocation scheduler. It fixes deadlocks, when frozen low priority payload requests occupying slots, required by higher levels to manage theit execution. - Two last changes were holding proper ATA reinitialization and error recovery implementation. Now it is done: SATA controllers and Port Multipliers now implement automatic hot-plug and should correctly recover from timeouts and bus resets. - Improve SCSI error recovery for devices on buses without automatic sense reporting, such as ATAPI or USB. For example, it allows CAM to wait, while CD drive loads disk, instead of immediately return error status. - Decapitalize diagnostic messages and make them more readable and sensible. - Teach PMP driver to limit maximum speed on fan-out ports. - Make boot wait for PMP scan completes, and make rescan more reliable. - Fix pass driver, to return CCB to user level in case of error. - Increase number of retries in cd driver, as device may return several UAs.
2010-01-28 08:41:30 +00:00
xpt_setup_ccb(&csa.ccb_h, path, CAM_PRIORITY_NORMAL);
csa.ccb_h.func_code = XPT_SASYNC_CB;
csa.event_enable = event;
csa.callback = cbfunc;
csa.callback_arg = cbarg;
xpt_action((union ccb *)&csa);
status = csa.ccb_h.status;
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
CAM_DEBUG(csa.ccb_h.path, CAM_DEBUG_TRACE,
("xpt_register_async: func %p\n", cbfunc));
if (xptpath) {
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_path_unlock(path);
xpt_free_path(path);
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
}
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
if ((status == CAM_REQ_CMP) &&
(csa.event_enable & AC_FOUND_DEVICE)) {
/*
* Get this peripheral up to date with all
* the currently existing devices.
*/
xpt_for_all_devices(xptsetasyncfunc, &csa);
}
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
if ((status == CAM_REQ_CMP) &&
(csa.event_enable & AC_PATH_REGISTERED)) {
/*
* Get this peripheral up to date with all
* the currently existing buses.
Lay groundwork in CAM for recording and reporting physical path and other device attributes stored in the CAM Existing Device Table (EDT). This includes some infrastructure requried by the enclosure services driver to export physical path information. Make the CAM device advanced info interface accept store requests. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: - Replace scsi_get_sas_addr() with a scsi_get_devid() which takes a callback that decides whether to accept a particular descriptor. Provide callbacks for NAA IEEE Registered addresses and for SAS addresses, replacing the old function. This is needed because the old function doesn't work for an enclosure address for a SAS device, which is not flagged as a SAS address, but is NAA IEEE Registered. It may be worthwhile merging this interface with the devid match interface. - Add a few more defines for some device ID fields. sbin/camcontrol/camcontrol.c: - Update for the CCB_DEV_ADVINFO interface change. cam/cam_xpt_internal.h: - Add the new fields for the physical path string to the CAM EDT. cam/cam_ccb.h: - Rename CCB_GDEV_ADVINFO to simply CCB_DEV_ADVINFO, and the ccb structure to ccb_dev_advinfo. - Add a flag that changes this CCB's action to store, rather than the default, retrieve. - Add a new buffer type, CDAI_TYPE_PHYS_PATH, for the new CAM EDT physpath field. - Remove the never-implemented transport & proto flags. cam/cam_xpt.c: cam/cam_xpt.h: - Add xpt_getattr(), which provides a wrapper for fetching a device's attribute using the GEOM strings as key. This method currently supports "GEOM::ident" and "GEOM::physpath". Submitted by: will Reviewed by : gibbs Extend the XPT_DEV_MATCH api to allow a device search by device ID. As far as the API is concerned, device ID is a binary blob to be interpreted by the transport layer. The SCSI implementation assumes it is an array of VPD device ID descriptors. sys/cam/cam_ccb.h: Create a new structure, device_id_match_pattern, and update the XPT_DEV_MATCH datastructures and flags so that this pattern type can be used. sys/cam/cam_xpt.c: - A single pattern matching on both inquiry data and device ID is invalid. Report any violators. - Pass device ID match requests through to the new routine scsi_devid_match(). The direct call of a SCSI routine is a layering violation, but no worse than the one a few lines up that checks inquiry data. Defer cleaning this up until our future, larger, rototilling of CAM. - Zero out cam_ed and cam_et nodes on allocation. Prior to this change, device_id_len and device_id were not inialized, preventing proper detection of the presence of this information. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add the scsi_match_devid() routine. Add a helper function for extracting peripherial driver names sys/cam/cam_periph.c: sys/cam/cam_periph.h: Add the cam_periph_list() method which fills an sbuf with a comma delimited list of the peripheral instances associated with a given CAM path. Add a helper functions for SCSI commands used by the SES driver. sys/cam/scsi/scsi_all.c: sys/cam/scsi/scsi_all.h: Add structure definitions and csio filling functions for the receive diagnostic results and send diagnostic commands. Misc CAM XPT cleanups. sys/cam/cam_xpt.c: Broadcast AC_FOUND_DEVICE and AC_PATH_REGISTERED events at the time async event handlers are attached even when registering just for events on a partitular SIM. Previously, you had to register for these events on all SIMs in the system in order to get the initial broadcast even though subsequent device and path arrivals would be delivered. sys/cam/cam_xpt.c: Remove SIM mutex held asserts from path accessors. CAM paths are reference counted and it is this reference count, not the sim mutex, that garantees they are stable. Sponsored by: Spectra Logic Corporation
2011-06-14 14:53:17 +00:00
*/
xpt_for_all_busses(xptsetasyncbusfunc, &csa);
}
return (status);
}
static void
xptaction(struct cam_sim *sim, union ccb *work_ccb)
{
CAM_DEBUG(work_ccb->ccb_h.path, CAM_DEBUG_TRACE, ("xptaction\n"));
switch (work_ccb->ccb_h.func_code) {
/* Common cases first */
case XPT_PATH_INQ: /* Path routing inquiry */
{
struct ccb_pathinq *cpi;
cpi = &work_ccb->cpi;
cpi->version_num = 1; /* XXX??? */
cpi->hba_inquiry = 0;
cpi->target_sprt = 0;
cpi->hba_misc = 0;
cpi->hba_eng_cnt = 0;
cpi->max_target = 0;
cpi->max_lun = 0;
cpi->initiator_id = 0;
strlcpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
strlcpy(cpi->hba_vid, "", HBA_IDLEN);
strlcpy(cpi->dev_name, sim->sim_name, DEV_IDLEN);
cpi->unit_number = sim->unit_number;
cpi->bus_id = sim->bus_id;
Add a number of interrelated CAM feature enhancements and bug fixes. NOTE: These changes will require recompilation of any userland applications, like cdrecord, xmcd, etc., that use the CAM passthrough interface. A make world is recommended. camcontrol.[c8]: - We now support two new commands, "tags" and "negotiate". - The tags commands allows users to view the number of tagged openings for a device as well as a number of other related parameters, and it allows users to set tagged openings for a device. - The negotiate command allows users to enable and disable disconnection and tagged queueing, set sync rates, offsets and bus width. Note that not all of those features are available for all controllers. Only the adv, ahc, and ncr drivers fully support all of the features at this point. Some cards do not allow the setting of sync rates, offsets and the like, and some of the drivers don't have any facilities to do so. Some drivers, like the adw driver, only support enabling or disabling sync negotiation, but do not support setting sync rates. - new description in the camcontrol man page of how to format a disk - cleanup of the camcontrol inquiry command - add support in the 'devlist' command for skipping unconfigured devices if -v was not specified on the command line. - make use of the new base_transfer_speed in the path inquiry CCB. - fix CCB bzero cases cam_xpt.c, cam_sim.[ch], cam_ccb.h: - new flags on many CCB function codes to designate whether they're non-immediate, use a user-supplied CCB, and can only be passed from userland programs via the xpt device. Use these flags in the transport layer and pass driver to categorize CCBs. - new flag in the transport layer device matching code for device nodes that indicates whether a device is unconfigured - bump the CAM version from 0x10 to 0x11 - Change the CAM ioctls to use the version as their group code, so we can force users to recompile code even when the CCB size doesn't change. - add + fill in a new value in the path inquiry CCB, base_transfer_speed. Remove a corresponding field from the cam_sim structure, and add code to every SIM to set this field to the proper value. - Fix the set transfer settings code in the transport layer. scsi_cd.c: - make some variables volatile instead of just casting them in various places - fix a race condition in the changer code - attach unless we get a "logical unit not supported" error. This should fix all of the cases where people have devices that return weird errors when they don't have media in the drive. scsi_da.c: - attach unless we get a "logical unit not supported" error scsi_pass.c: - for immediate CCBs, just malloc a CCB to send the user request in. This gets rid of the 'held' count problem in camcontrol tags. scsi_pass.h: - change the CAM ioctls to use the CAM version as their group code. adv driver: - Allow changing the sync rate and offset separately. adw driver - Allow changing the sync rate and offset separately. aha driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. ahc driver: - Allow setting offset and sync rate separately bt driver: - Don't return CAM_REQ_CMP for SET_TRAN_SETTINGS CCBs. NCR driver: - Fix the ultra/ultra 2 negotiation bug - allow setting both the sync rate and offset separately Other HBA drivers: - Put code in to set the base_transfer_speed field for XPT_GET_TRAN_SETTINGS CCBs. Reviewed by: gibbs, mjacob (isp), imp (aha)
1999-05-06 20:16:39 +00:00
cpi->base_transfer_speed = 0;
Rewrite of the CAM error recovery code. Some of the major changes include: - The SCSI error handling portion of cam_periph_error() has been broken out into a number of subfunctions to better modularize the code that handles the hierarchy of SCSI errors. As a result, the code is now much easier to read. - String handling and error printing has been significantly revamped. We now use sbufs to do string formatting instead of using printfs (for the kernel) and snprintf/strncat (for userland) as before. There is a new catchall error printing routine, cam_error_print() and its string-based counterpart, cam_error_string() that allow the kernel and userland applications to pass in a CCB and have errors printed out properly, whether or not they're SCSI errors. Among other things, this helped eliminate a fair amount of duplicate code in camcontrol. We now print out more information than before, including the CAM status and SCSI status and the error recovery action taken to remedy the problem. - sbufs are now available in userland, via libsbuf. This change was necessary since most of the error printing code is shared between libcam and the kernel. - A new transfer settings interface is included in this checkin. This code is #ifdef'ed out, and is primarily intended to aid discussion with HBA driver authors on the final form the interface should take. There is example code in the ahc(4) driver that implements the HBA driver side of the new interface. The new transfer settings code won't be enabled until we're ready to switch all HBA drivers over to the new interface. src/Makefile.inc1, lib/Makefile: Add libsbuf. It must be built before libcam, since libcam uses sbuf routines. libcam/Makefile: libcam now depends on libsbuf. libsbuf/Makefile: Add a makefile for libsbuf. This pulls in the sbuf sources from sys/kern. bsd.libnames.mk: Add LIBSBUF. camcontrol/Makefile: Add -lsbuf. Since camcontrol is statically linked, we can't depend on the dynamic linker to pull in libsbuf. camcontrol.c: Use cam_error_print() instead of checking for CAM_SCSI_STATUS_ERROR on every failed CCB. sbuf.9: Change the prototypes for sbuf_cat() and sbuf_cpy() so that the source string is now a const char *. This is more in line wth the standard system string functions, and helps eliminate warnings when dealing with a const source buffer. Fix a typo. cam.c: Add description strings for the various CAM error status values, as well as routines to look up those strings. Add new cam_error_string() and cam_error_print() routines for userland and the kernel. cam.h: Add a new CAM flag, CAM_RETRY_SELTO. Add enumerated types for the various options available with cam_error_print() and cam_error_string(). cam_ccb.h: Add new transfer negotiation structures/types. Change inq_len in the ccb_getdev structure to be "reserved". This field has never been filled in, and will be removed when we next bump the CAM version. cam_debug.h: Fix typo. cam_periph.c: Modularize cam_periph_error(). The SCSI error handling part of cam_periph_error() is now in camperiphscsistatuserror() and camperiphscsisenseerror(). In cam_periph_lock(), increase the reference count on the periph while we wait for our lock attempt to succeed so that the periph won't go away while we're sleeping. cam_xpt.c: Add new transfer negotiation code. (ifdefed out) Add a new function, xpt_path_string(). This is a string/sbuf analog to xpt_print_path(). scsi_all.c: Revamp string handing and error printing code. We now use sbufs for much of the string formatting code. More of that code is shared between userland the kernel. scsi_all.h: Get rid of SS_TURSTART, it wasn't terribly useful in the first place. Add a new error action, SS_REQSENSE. (Send a request sense and then retry the command.) This is useful when the controller hasn't performed autosense for some reason. Change the default actions around a bit. scsi_cd.c, scsi_da.c, scsi_pt.c, scsi_ses.c: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Selection timeouts shouldn't be covered by a sense flag. scsi_pass.[ch]: SF_RETRY_SELTO -> CAM_RETRY_SELTO. Get rid of the last vestiges of a read/write interface. libkern/bsearch.c, sys/libkern.h, conf/files: Add bsearch.c, which is needed for some of the new table lookup routines. aic7xxx_freebsd.c: Define AHC_NEW_TRAN_SETTINGS if CAM_NEW_TRAN_CODE is defined. sbuf.h, subr_sbuf.c: Add the appropriate #ifdefs so sbufs can compile and run in userland. Change sbuf_printf() to use vsnprintf() instead of kvprintf(), which is only available in the kernel. Change the source string for sbuf_cpy() and sbuf_cat() to be a const char *. Add __BEGIN_DECLS and __END_DECLS around function prototypes since they're now exported to userland. kdump/mkioctls: Include stdio.h before cam.h since cam.h now includes a function with a FILE * argument. Submitted by: gibbs (mostly) Reviewed by: jdp, marcel (libsbuf makefile changes) Reviewed by: des (sbuf changes) Reviewed by: ken
2001-03-27 05:45:52 +00:00
cpi->protocol = PROTO_UNSPECIFIED;
cpi->protocol_version = PROTO_VERSION_UNSPECIFIED;
cpi->transport = XPORT_UNSPECIFIED;
cpi->transport_version = XPORT_VERSION_UNSPECIFIED;
cpi->ccb_h.status = CAM_REQ_CMP;
break;
}
default:
work_ccb->ccb_h.status = CAM_REQ_INVALID;
break;
}
xpt_done(work_ccb);
}
/*
* The xpt as a "controller" has no interrupt sources, so polling
* is a no-op.
*/
static void
xptpoll(struct cam_sim *sim)
{
}
void
xpt_lock_buses(void)
{
mtx_lock(&xsoftc.xpt_topo_lock);
}
void
xpt_unlock_buses(void)
{
mtx_unlock(&xsoftc.xpt_topo_lock);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct mtx *
xpt_path_mtx(struct cam_path *path)
{
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
return (&path->device->device_mtx);
}
static void
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_done_process(struct ccb_hdr *ccb_h)
{
struct cam_sim *sim = NULL;
struct cam_devq *devq = NULL;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
struct mtx *mtx = NULL;
#if defined(BUF_TRACKING) || defined(FULL_BUF_TRACKING)
struct ccb_scsiio *csio;
if (ccb_h->func_code == XPT_SCSI_IO) {
csio = &((union ccb *)ccb_h)->csio;
if (csio->bio != NULL)
biotrack(csio->bio, __func__);
}
#endif
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (ccb_h->flags & CAM_HIGH_POWER) {
struct highpowerlist *hphead;
struct cam_ed *device;
mtx_lock(&xsoftc.xpt_highpower_lock);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
hphead = &xsoftc.highpowerq;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
device = STAILQ_FIRST(hphead);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
/*
* Increment the count since this command is done.
*/
xsoftc.num_highpower++;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
/*
* Any high powered commands queued up?
*/
if (device != NULL) {
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
STAILQ_REMOVE_HEAD(hphead, highpowerq_entry);
mtx_unlock(&xsoftc.xpt_highpower_lock);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&device->sim->devq->send_mtx);
xpt_release_devq_device(device,
/*count*/1, /*runqueue*/TRUE);
mtx_unlock(&device->sim->devq->send_mtx);
} else
mtx_unlock(&xsoftc.xpt_highpower_lock);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
/*
* Insulate against a race where the periph is destroyed but CCBs are
* still not all processed. This shouldn't happen, but allows us better
* bug diagnostic when it does.
*/
if (ccb_h->path->bus)
sim = ccb_h->path->bus->sim;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (ccb_h->status & CAM_RELEASE_SIMQ) {
KASSERT(sim, ("sim missing for CAM_RELEASE_SIMQ request"));
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
xpt_release_simq(sim, /*run_queue*/FALSE);
ccb_h->status &= ~CAM_RELEASE_SIMQ;
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if ((ccb_h->flags & CAM_DEV_QFRZDIS)
&& (ccb_h->status & CAM_DEV_QFRZN)) {
xpt_release_devq(ccb_h->path, /*count*/1, /*run_queue*/TRUE);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
ccb_h->status &= ~CAM_DEV_QFRZN;
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if ((ccb_h->func_code & XPT_FC_USER_CCB) == 0) {
struct cam_ed *dev = ccb_h->path->device;
if (sim)
devq = sim->devq;
KASSERT(devq, ("Periph disappeared with CCB %p %s request pending.",
ccb_h, xpt_action_name(ccb_h->func_code)));
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_lock(&devq->send_mtx);
devq->send_active--;
devq->send_openings++;
cam_ccbq_ccb_done(&dev->ccbq, (union ccb *)ccb_h);
if (((dev->flags & CAM_DEV_REL_ON_QUEUE_EMPTY) != 0
&& (dev->ccbq.dev_active == 0))) {
dev->flags &= ~CAM_DEV_REL_ON_QUEUE_EMPTY;
xpt_release_devq_device(dev, /*count*/1,
/*run_queue*/FALSE);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (((dev->flags & CAM_DEV_REL_ON_COMPLETE) != 0
&& (ccb_h->status&CAM_STATUS_MASK) != CAM_REQUEUE_REQ)) {
dev->flags &= ~CAM_DEV_REL_ON_COMPLETE;
xpt_release_devq_device(dev, /*count*/1,
/*run_queue*/FALSE);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if (!device_is_queued(dev))
(void)xpt_schedule_devq(devq, dev);
xpt_run_devq(devq);
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
mtx_unlock(&devq->send_mtx);
2008-01-02 01:45:31 +00:00
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if ((dev->flags & CAM_DEV_TAG_AFTER_COUNT) != 0) {
mtx = xpt_path_mtx(ccb_h->path);
mtx_lock(mtx);
if ((dev->flags & CAM_DEV_TAG_AFTER_COUNT) != 0
&& (--dev->tag_delay_count == 0))
xpt_start_tags(ccb_h->path);
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
if ((ccb_h->flags & CAM_UNLOCKED) == 0) {
if (mtx == NULL) {
mtx = xpt_path_mtx(ccb_h->path);
mtx_lock(mtx);
}
} else {
if (mtx != NULL) {
mtx_unlock(mtx);
mtx = NULL;
2008-01-02 01:45:31 +00:00
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
/* Call the peripheral driver's callback */
ccb_h->pinfo.index = CAM_UNQUEUED_INDEX;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
(*ccb_h->cbfcnp)(ccb_h->path->periph, (union ccb *)ccb_h);
if (mtx != NULL)
mtx_unlock(mtx);
}
void
xpt_done_td(void *arg)
{
struct cam_doneq *queue = arg;
struct ccb_hdr *ccb_h;
STAILQ_HEAD(, ccb_hdr) doneq;
STAILQ_INIT(&doneq);
mtx_lock(&queue->cam_doneq_mtx);
while (1) {
while (STAILQ_EMPTY(&queue->cam_doneq)) {
queue->cam_doneq_sleep = 1;
msleep(&queue->cam_doneq, &queue->cam_doneq_mtx,
PRIBIO, "-", 0);
queue->cam_doneq_sleep = 0;
}
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
STAILQ_CONCAT(&doneq, &queue->cam_doneq);
mtx_unlock(&queue->cam_doneq_mtx);
THREAD_NO_SLEEPING();
while ((ccb_h = STAILQ_FIRST(&doneq)) != NULL) {
STAILQ_REMOVE_HEAD(&doneq, sim_links.stqe);
xpt_done_process(ccb_h);
}
THREAD_SLEEPING_OK();
mtx_lock(&queue->cam_doneq_mtx);
}
}
static void
camisr_runqueue(void)
{
struct ccb_hdr *ccb_h;
struct cam_doneq *queue;
int i;
Merge CAM locking changes from the projects/camlock branch to radically reduce lock congestion and improve SMP scalability of the SCSI/ATA stack, preparing the ground for the coming next GEOM direct dispatch support. Replace big per-SIM locks with bunch of smaller ones: - per-LUN locks to protect device and peripheral drivers state; - per-target locks to protect list of LUNs on target; - per-bus locks to protect reference counting; - per-send queue locks to protect queue of CCBs to be sent; - per-done queue locks to protect queue of completed CCBs; - remaining per-SIM locks now protect only HBA driver internals. While holding LUN lock it is allowed (while not recommended for performance reasons) to take SIM lock. The opposite acquisition order is forbidden. All the other locks are leaf locks, that can be taken anywhere, but should not be cascaded. Many functions, such as: xpt_action(), xpt_done(), xpt_async(), xpt_create_path(), etc. are no longer require (but allow) SIM lock to be held. To keep compatibility and solve cases where SIM lock can't be dropped, all xpt_async() calls in addition to xpt_done() calls are queued to completion threads for async processing in clean environment without SIM lock held. Instead of single CAM SWI thread, used for commands completion processing before, use multiple (depending on number of CPUs) threads. Load balanced between them using "hash" of the device B:T:L address. HBA drivers that can drop SIM lock during completion processing and have sufficient number of completion threads to efficiently scale to multiple CPUs can use new function xpt_done_direct() to avoid extra context switch. Make ahci(4) driver to use this mechanism depending on hardware setup. Sponsored by: iXsystems, Inc. MFC after: 2 months
2013-10-21 12:00:26 +00:00
/* Process global queues. */
for (i = 0; i < cam_num_doneqs; i++) {
queue = &cam_doneqs[i];
mtx_lock(&queue->cam_doneq_mtx);
while ((ccb_h = STAILQ_FIRST(&queue->cam_doneq)) != NULL) {
STAILQ_REMOVE_HEAD(&queue->cam_doneq, sim_links.stqe);
mtx_unlock(&queue->cam_doneq_mtx);
xpt_done_process(ccb_h);
mtx_lock(&queue->cam_doneq_mtx);
}
mtx_unlock(&queue->cam_doneq_mtx);
}
}
struct kv
{
uint32_t v;
const char *name;
};
static struct kv map[] = {
{ XPT_NOOP, "XPT_NOOP" },
{ XPT_SCSI_IO, "XPT_SCSI_IO" },
{ XPT_GDEV_TYPE, "XPT_GDEV_TYPE" },
{ XPT_GDEVLIST, "XPT_GDEVLIST" },
{ XPT_PATH_INQ, "XPT_PATH_INQ" },
{ XPT_REL_SIMQ, "XPT_REL_SIMQ" },
{ XPT_SASYNC_CB, "XPT_SASYNC_CB" },
{ XPT_SDEV_TYPE, "XPT_SDEV_TYPE" },
{ XPT_SCAN_BUS, "XPT_SCAN_BUS" },
{ XPT_DEV_MATCH, "XPT_DEV_MATCH" },
{ XPT_DEBUG, "XPT_DEBUG" },
{ XPT_PATH_STATS, "XPT_PATH_STATS" },
{ XPT_GDEV_STATS, "XPT_GDEV_STATS" },
{ XPT_DEV_ADVINFO, "XPT_DEV_ADVINFO" },
{ XPT_ASYNC, "XPT_ASYNC" },
{ XPT_ABORT, "XPT_ABORT" },
{ XPT_RESET_BUS, "XPT_RESET_BUS" },
{ XPT_RESET_DEV, "XPT_RESET_DEV" },
{ XPT_TERM_IO, "XPT_TERM_IO" },
{ XPT_SCAN_LUN, "XPT_SCAN_LUN" },
{ XPT_GET_TRAN_SETTINGS, "XPT_GET_TRAN_SETTINGS" },
{ XPT_SET_TRAN_SETTINGS, "XPT_SET_TRAN_SETTINGS" },
{ XPT_CALC_GEOMETRY, "XPT_CALC_GEOMETRY" },
{ XPT_ATA_IO, "XPT_ATA_IO" },
{ XPT_GET_SIM_KNOB, "XPT_GET_SIM_KNOB" },
{ XPT_SET_SIM_KNOB, "XPT_SET_SIM_KNOB" },
2016-06-09 16:05:49 +00:00
{ XPT_NVME_IO, "XPT_NVME_IO" },
{ XPT_MMC_IO, "XPT_MMC_IO" },
{ XPT_SMP_IO, "XPT_SMP_IO" },
{ XPT_SCAN_TGT, "XPT_SCAN_TGT" },
{ XPT_NVME_ADMIN, "XPT_NVME_ADMIN" },
{ XPT_ENG_INQ, "XPT_ENG_INQ" },
{ XPT_ENG_EXEC, "XPT_ENG_EXEC" },
{ XPT_EN_LUN, "XPT_EN_LUN" },
{ XPT_TARGET_IO, "XPT_TARGET_IO" },
{ XPT_ACCEPT_TARGET_IO, "XPT_ACCEPT_TARGET_IO" },
{ XPT_CONT_TARGET_IO, "XPT_CONT_TARGET_IO" },
{ XPT_IMMED_NOTIFY, "XPT_IMMED_NOTIFY" },
{ XPT_NOTIFY_ACK, "XPT_NOTIFY_ACK" },
{ XPT_IMMEDIATE_NOTIFY, "XPT_IMMEDIATE_NOTIFY" },
{ XPT_NOTIFY_ACKNOWLEDGE, "XPT_NOTIFY_ACKNOWLEDGE" },
{ 0, 0 }
};
const char *
xpt_action_name(uint32_t action)
{
static char buffer[32]; /* Only for unknown messages -- racy */
struct kv *walker = map;
while (walker->name != NULL) {
if (walker->v == action)
return (walker->name);
walker++;
}
snprintf(buffer, sizeof(buffer), "%#x", action);
return (buffer);
}