d/freebsd-nq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
f64d4bd0f7
freebsd-nq/sys/cam/cam_xpt.c

7287 lines
191 KiB
C
Raw Normal View History

Start each of the license/copyright comments with /*-
2005-01-05 22:34:37 +00:00
/*-
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* Implementation of the Common Access Method Transport (XPT) layer.
*
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
* Copyright (c) 1997, 1998, 1999 Justin T. Gibbs.
Generalize the quirk entry for the Conner CFP* drives. It did just cover the CFP2107, but it appears (not surprisingly) that the 1 gig and 4 gig versions of that drive have the same problem with tagged queueing. Also, fix the problem reported in PR kern/9482. The XPT_DEV_MATCH case in xptioctl() wasn't putting a proper path in the CCB before it called xpt_action(). When CAMDEBUG is defined, and CAM_DEBUG_TRACE debugging is turned on, the CAM_DEBUG statement at the beginning of xpt_action would end up deferencing a NULL path pointer. That of course caused a panic. My solution is to just stick the xpt peripheral's path in the CCB. PR: kern/9482 Reviewed by: gibbs
1999-01-19 01:02:47 +00:00
* Copyright (c) 1997, 1998, 1999 Kenneth D. Merry.
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* 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.
*/
Use __FBSDID().
2003-06-10 17:50:20 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
#include <sys/param.h>
- Remove the inthand2_t type and use the equivalent driver_intr_t type from newbus for referencing device interrupt handlers. - Move the 'struct intrec' type which describes interrupt sources into sys/interrupt.h instead of making it just be a x86 structure. - Don't create 'ithd' and 'intrec' typedefs, instead, just use 'struct ithd' and 'struct intrec' - Move the code to translate new-bus interrupt flags into an interrupt thread priority out of the x86 nexus code and into a MI ithread_priority() function in sys/kern/kern_intr.c. - Remove now-uneeded x86-specific headers from sys/dev/ata/ata-all.c and sys/pci/pci_compat.c.
2000-09-13 18:33:25 +00:00
#include <sys/bus.h>
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
#include <sys/systm.h>
#include <sys/types.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
#include <sys/time.h>
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
#include <sys/conf.h>
#include <sys/fcntl.h>
#include <sys/md5.h>
a more correct change that should meet the sniff test
1998-12-24 06:01:15 +00:00
#include <sys/interrupt.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>
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
#include <sys/taskqueue.h>
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Provide locking for the ccb_bioq. This allows xpt_done() to be called without Giant held. In camisr(), move the ccb_bioq elements to a temporary local list and then process the elements off of that list. This enables the list to be processed by only taking the ccb_bioq_lock once and only for a very short time. ccb_bioq_lock is a leaf mutex, so it's fine to call xpt_done() with other locks held. This is just a very minor step in the work to lock CAM, but it allows us to avoid some messy locking/unlock dances in certain drivers.
2005-02-09 11:44:15 +00:00
#include <sys/lock.h>
#include <sys/mutex.h>
Make the exploring of all luns supported by an HBA more of a tunable (until we get REPORT LUNS in place). If we're probing luns, and each probe succeeds, we keep going past lun 7 if we're a SCSI3 or better device (until we fail to probe). If we're probing luns, and a probe fails, we only keep going if we're quirked *for* it (CAM_QUIRK_HILUNS), and if we're not quirked *against* it (CAM_QUIRK_NOHILUNS), or we're a SCSI3 or better device and the tunable (kern.cam.cam_srch_hi) is set non-zero. Reviewed by: nate@rootlabs.org, gibbs@scsiguy.com, ken@kdm.com, scottl@samsco.org MFC after: 1 week
2005-09-16 01:26:17 +00:00
#include <sys/sysctl.h>
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
#include <sys/kthread.h>
Provide locking for the ccb_bioq. This allows xpt_done() to be called without Giant held. In camisr(), move the ccb_bioq elements to a temporary local list and then process the elements off of that list. This enables the list to be processed by only taking the ccb_bioq_lock once and only for a very short time. ccb_bioq_lock is a leaf mutex, so it's fine to call xpt_done() with other locks held. This is just a very minor step in the work to lock CAM, but it allows us to avoid some messy locking/unlock dances in certain drivers.
2005-02-09 11:44:15 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
#ifdef PC98
#include <pc98/pc98/pc98_machdep.h> /* geometry translation */
#endif
#include <cam/cam.h>
#include <cam/cam_ccb.h>
#include <cam/cam_periph.h>
#include <cam/cam_sim.h>
#include <cam/cam_xpt.h>
#include <cam/cam_xpt_sim.h>
#include <cam/cam_xpt_periph.h>
#include <cam/cam_debug.h>
#include <cam/scsi/scsi_all.h>
#include <cam/scsi/scsi_message.h>
#include <cam/scsi/scsi_pass.h>
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
#include <machine/stdarg.h> /* for xpt_print below */
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
#include "opt_cam.h"
/* Datastructures internal to the xpt layer */
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
MALLOC_DEFINE(M_CAMXPT, "CAM XPT", "CAM XPT buffers");
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
/* Object for defering XPT actions to a taskqueue */
struct xpt_task {
struct task task;
For the XPT_SASYNC_CB operation, only decouple the broadcast to the bus and device lists instead of decoupling the whole operation. This avoids problems with SIMs going away.
2007-04-16 19:55:36 +00:00
void *data1;
uintptr_t data2;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
};
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Definition of an async handler callback block. These are used to add
* SIMs and peripherals to the async callback lists.
*/
struct async_node {
Back out the previous change to the queue(3) interface. It was not discussed and should probably not happen. Requested by: msmith and others
2000-05-26 02:09:24 +00:00
SLIST_ENTRY(async_node) links;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
u_int32_t event_enable; /* Async Event enables */
void (*callback)(void *arg, u_int32_t code,
struct cam_path *path, void *args);
void *callback_arg;
};
Back out the previous change to the queue(3) interface. It was not discussed and should probably not happen. Requested by: msmith and others
2000-05-26 02:09:24 +00:00
SLIST_HEAD(async_list, async_node);
SLIST_HEAD(periph_list, cam_periph);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* 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
/*
* Structure for queueing a device in a run queue.
* There is one run queue for allocating new ccbs,
* and another for sending ccbs to the controller.
*/
struct cam_ed_qinfo {
cam_pinfo pinfo;
struct cam_ed *device;
};
/*
* The CAM EDT (Existing Device Table) contains the device information for
* all devices for all busses in the system. The table contains a
* cam_ed structure for each device on the bus.
*/
struct cam_ed {
Back out the previous change to the queue(3) interface. It was not discussed and should probably not happen. Requested by: msmith and others
2000-05-26 02:09:24 +00:00
TAILQ_ENTRY(cam_ed) links;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct cam_ed_qinfo alloc_ccb_entry;
struct cam_ed_qinfo send_ccb_entry;
struct cam_et *target;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
struct cam_sim *sim;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
lun_id_t lun_id;
struct camq drvq; /*
* Queue of type drivers wanting to do
* work on this device.
*/
struct cam_ccbq ccbq; /* Queue of pending ccbs */
struct async_list asyncs; /* Async callback info for this B/T/L */
struct periph_list periphs; /* All attached devices */
u_int generation; /* Generation number */
struct cam_periph *owner; /* Peripheral driver's ownership tag */
struct xpt_quirk_entry *quirk; /* Oddities about this device */
/* Storage for the inquiry data */
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
cam_proto protocol;
u_int protocol_version;
cam_xport transport;
u_int transport_version;
struct scsi_inquiry_data inq_data;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
u_int8_t inq_flags; /*
* Current settings for inquiry flags.
* This allows us to override settings
* like disconnection and tagged
* queuing for a device.
*/
u_int8_t queue_flags; /* Queue flags from the control page */
u_int8_t serial_num_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
u_int8_t *serial_num;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
u_int32_t qfrozen_cnt;
u_int32_t flags;
#define CAM_DEV_UNCONFIGURED 0x01
#define CAM_DEV_REL_TIMEOUT_PENDING 0x02
#define CAM_DEV_REL_ON_COMPLETE 0x04
#define CAM_DEV_REL_ON_QUEUE_EMPTY 0x08
#define CAM_DEV_RESIZE_QUEUE_NEEDED 0x10
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
#define CAM_DEV_TAG_AFTER_COUNT 0x20
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
#define CAM_DEV_INQUIRY_DATA_VALID 0x40
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
#define CAM_DEV_IN_DV 0x80
#define CAM_DEV_DV_HIT_BOTTOM 0x100
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
u_int32_t tag_delay_count;
#define CAM_TAG_DELAY_COUNT 5
Restore queue depth settings across tag disable events. The system often disables tag queuing temporarily in order to allow controllers a window to safely perform transfer negotiation with non-compliant devices. Before this change, CAM would restore the queue depth to the controller specified maximum or device quirk level rather than any depth determined by reactions to QUEUE FULL/BUSY events or an explicit user setting. During device probe, initialize the flags field for XPT_SCAN_BUS. The uninitialized value often confused CAM into not bothering to issue an AC_FOUND_DEVICE async event for new devices. The reason this bug wasn't reported earlier is that CAM manually announces devices after the initial system bus scans. MFC: 3 days
2005-03-23 16:43:29 +00:00
u_int32_t tag_saved_openings;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
u_int32_t refcount;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
struct callout callout;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
};
/*
* Each target is represented by an ET (Existing Target). These
* entries are created when a target is successfully probed with an
* identify, and removed when a device fails to respond after a number
* of retries, or a bus rescan finds the device missing.
*/
style(9)
2008-01-02 01:45:31 +00:00
struct cam_et {
Back out the previous change to the queue(3) interface. It was not discussed and should probably not happen. Requested by: msmith and others
2000-05-26 02:09:24 +00:00
TAILQ_HEAD(, cam_ed) ed_entries;
TAILQ_ENTRY(cam_et) links;
style(9)
2008-01-02 01:45:31 +00:00
struct cam_eb *bus;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
target_id_t target_id;
style(9)
2008-01-02 01:45:31 +00:00
u_int32_t refcount;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
u_int generation;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
struct timeval last_reset;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
};
/*
* Each bus is represented by an EB (Existing Bus). These entries
* are created by calls to xpt_bus_register and deleted by calls to
* xpt_bus_deregister.
*/
style(9)
2008-01-02 01:45:31 +00:00
struct cam_eb {
Back out the previous change to the queue(3) interface. It was not discussed and should probably not happen. Requested by: msmith and others
2000-05-26 02:09:24 +00:00
TAILQ_HEAD(, cam_et) et_entries;
TAILQ_ENTRY(cam_eb) links;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
path_id_t path_id;
struct cam_sim *sim;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
struct timeval last_reset;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
u_int32_t flags;
#define CAM_EB_RUNQ_SCHEDULED 0x01
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
u_int32_t refcount;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
u_int generation;
Prepare for future integration between CAM and newbus. xpt_bus_register now takes a device_t to be the parent of the bus that is being created. Most SIMs have been updated with a reasonable argument, but a few exceptions just pass NULL for now. This argument isn't used yet and the newbus integration likely won't be ready until after 7.0-RELEASE.
2007-06-17 05:55:54 +00:00
device_t parent_dev;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
};
struct cam_path {
struct cam_periph *periph;
struct cam_eb *bus;
struct cam_et *target;
struct cam_ed *device;
};
struct xpt_quirk_entry {
struct scsi_inquiry_pattern inq_pat;
u_int8_t quirks;
#define CAM_QUIRK_NOLUNS 0x01
#define CAM_QUIRK_NOSERIAL 0x02
Put in a limit (8) for lun width- do not exceed probing this unless a device is quirked for allowing it. This will allow SCSI cards to specify their true correct lun width (32 for Qlogic SCSI, 16 or 65536 for Qlogic Fibre Channel) without crippling the system because of f/w breakage of a lot of drives. Reviewed by: gibbs@freebsd.org
1999-08-16 17:47:39 +00:00
#define CAM_QUIRK_HILUNS 0x04
This is a somewhat imperfect means to try and bring FreeBSD forward in its ability to automatically scan and attach luns for modern storage which has luns in the 0..1000 range, not 0..7. The correct thing would be to do REPORT LUNS for devices whose LUN0 version shows a version >= SCSI3, but lacking that we should be able to search higher than LUN 7 if we're >= SCSI3 with no ill effects. This change keeps all of the QUIRK_HILUNS quirks, obeys the QUIRK_NOLUNS, and introduces a QUIRK_NOHILUNS which will keep searches above LUN 7 happening for devices that report >= SCSI3 compliance. I doubt the latter will be needed, but you never know. This allowed me to randomly scan and attach > 500 disks at a time in a situation where quirking for QUIRK_HILUNS wasn't practical (the vendor id and product id changes of the virtualization changes constantly). Reviewed by: ken@freebsd.org, scottl@freebsd.org, gibbs@freebsd.org MFC after: 2 weeks
2005-01-22 22:46:45 +00:00
#define CAM_QUIRK_NOHILUNS 0x08
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
u_int mintags;
u_int maxtags;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
};
Make the exploring of all luns supported by an HBA more of a tunable (until we get REPORT LUNS in place). If we're probing luns, and each probe succeeds, we keep going past lun 7 if we're a SCSI3 or better device (until we fail to probe). If we're probing luns, and a probe fails, we only keep going if we're quirked *for* it (CAM_QUIRK_HILUNS), and if we're not quirked *against* it (CAM_QUIRK_NOHILUNS), or we're a SCSI3 or better device and the tunable (kern.cam.cam_srch_hi) is set non-zero. Reviewed by: nate@rootlabs.org, gibbs@scsiguy.com, ken@kdm.com, scottl@samsco.org MFC after: 1 week
2005-09-16 01:26:17 +00:00
static int cam_srch_hi = 0;
TUNABLE_INT("kern.cam.cam_srch_hi", &cam_srch_hi);
static int sysctl_cam_search_luns(SYSCTL_HANDLER_ARGS);
SYSCTL_PROC(_kern_cam, OID_AUTO, cam_srch_hi, CTLTYPE_INT|CTLFLAG_RW, 0, 0,
sysctl_cam_search_luns, "I",
"allow search above LUN 7 for SCSI3 and greater devices");
Put in a limit (8) for lun width- do not exceed probing this unless a device is quirked for allowing it. This will allow SCSI cards to specify their true correct lun width (32 for Qlogic SCSI, 16 or 65536 for Qlogic Fibre Channel) without crippling the system because of f/w breakage of a lot of drives. Reviewed by: gibbs@freebsd.org
1999-08-16 17:47:39 +00:00
#define CAM_SCSI2_MAXLUN 8
This is a somewhat imperfect means to try and bring FreeBSD forward in its ability to automatically scan and attach luns for modern storage which has luns in the 0..1000 range, not 0..7. The correct thing would be to do REPORT LUNS for devices whose LUN0 version shows a version >= SCSI3, but lacking that we should be able to search higher than LUN 7 if we're >= SCSI3 with no ill effects. This change keeps all of the QUIRK_HILUNS quirks, obeys the QUIRK_NOLUNS, and introduces a QUIRK_NOHILUNS which will keep searches above LUN 7 happening for devices that report >= SCSI3 compliance. I doubt the latter will be needed, but you never know. This allowed me to randomly scan and attach > 500 disks at a time in a situation where quirking for QUIRK_HILUNS wasn't practical (the vendor id and product id changes of the virtualization changes constantly). Reviewed by: ken@freebsd.org, scottl@freebsd.org, gibbs@freebsd.org MFC after: 2 weeks
2005-01-22 22:46:45 +00:00
/*
* If we're not quirked to search <= the first 8 luns
* and we are either quirked to search above lun 8,
Make the exploring of all luns supported by an HBA more of a tunable (until we get REPORT LUNS in place). If we're probing luns, and each probe succeeds, we keep going past lun 7 if we're a SCSI3 or better device (until we fail to probe). If we're probing luns, and a probe fails, we only keep going if we're quirked *for* it (CAM_QUIRK_HILUNS), and if we're not quirked *against* it (CAM_QUIRK_NOHILUNS), or we're a SCSI3 or better device and the tunable (kern.cam.cam_srch_hi) is set non-zero. Reviewed by: nate@rootlabs.org, gibbs@scsiguy.com, ken@kdm.com, scottl@samsco.org MFC after: 1 week
2005-09-16 01:26:17 +00:00
* or we're > SCSI-2 and we've enabled hilun searching,
* or we're > SCSI-2 and the last lun was a success,
* we can look for luns above lun 8.
This is a somewhat imperfect means to try and bring FreeBSD forward in its ability to automatically scan and attach luns for modern storage which has luns in the 0..1000 range, not 0..7. The correct thing would be to do REPORT LUNS for devices whose LUN0 version shows a version >= SCSI3, but lacking that we should be able to search higher than LUN 7 if we're >= SCSI3 with no ill effects. This change keeps all of the QUIRK_HILUNS quirks, obeys the QUIRK_NOLUNS, and introduces a QUIRK_NOHILUNS which will keep searches above LUN 7 happening for devices that report >= SCSI3 compliance. I doubt the latter will be needed, but you never know. This allowed me to randomly scan and attach > 500 disks at a time in a situation where quirking for QUIRK_HILUNS wasn't practical (the vendor id and product id changes of the virtualization changes constantly). Reviewed by: ken@freebsd.org, scottl@freebsd.org, gibbs@freebsd.org MFC after: 2 weeks
2005-01-22 22:46:45 +00:00
*/
Make the exploring of all luns supported by an HBA more of a tunable (until we get REPORT LUNS in place). If we're probing luns, and each probe succeeds, we keep going past lun 7 if we're a SCSI3 or better device (until we fail to probe). If we're probing luns, and a probe fails, we only keep going if we're quirked *for* it (CAM_QUIRK_HILUNS), and if we're not quirked *against* it (CAM_QUIRK_NOHILUNS), or we're a SCSI3 or better device and the tunable (kern.cam.cam_srch_hi) is set non-zero. Reviewed by: nate@rootlabs.org, gibbs@scsiguy.com, ken@kdm.com, scottl@samsco.org MFC after: 1 week
2005-09-16 01:26:17 +00:00
#define CAN_SRCH_HI_SPARSE(dv) \
(((dv->quirk->quirks & CAM_QUIRK_NOHILUNS) == 0) \
&& ((dv->quirk->quirks & CAM_QUIRK_HILUNS) \
|| (SID_ANSI_REV(&dv->inq_data) > SCSI_REV_2 && cam_srch_hi)))
#define CAN_SRCH_HI_DENSE(dv) \
This is a somewhat imperfect means to try and bring FreeBSD forward in its ability to automatically scan and attach luns for modern storage which has luns in the 0..1000 range, not 0..7. The correct thing would be to do REPORT LUNS for devices whose LUN0 version shows a version >= SCSI3, but lacking that we should be able to search higher than LUN 7 if we're >= SCSI3 with no ill effects. This change keeps all of the QUIRK_HILUNS quirks, obeys the QUIRK_NOLUNS, and introduces a QUIRK_NOHILUNS which will keep searches above LUN 7 happening for devices that report >= SCSI3 compliance. I doubt the latter will be needed, but you never know. This allowed me to randomly scan and attach > 500 disks at a time in a situation where quirking for QUIRK_HILUNS wasn't practical (the vendor id and product id changes of the virtualization changes constantly). Reviewed by: ken@freebsd.org, scottl@freebsd.org, gibbs@freebsd.org MFC after: 2 weeks
2005-01-22 22:46:45 +00:00
(((dv->quirk->quirks & CAM_QUIRK_NOHILUNS) == 0) \
&& ((dv->quirk->quirks & CAM_QUIRK_HILUNS) \
Make the exploring of all luns supported by an HBA more of a tunable (until we get REPORT LUNS in place). If we're probing luns, and each probe succeeds, we keep going past lun 7 if we're a SCSI3 or better device (until we fail to probe). If we're probing luns, and a probe fails, we only keep going if we're quirked *for* it (CAM_QUIRK_HILUNS), and if we're not quirked *against* it (CAM_QUIRK_NOHILUNS), or we're a SCSI3 or better device and the tunable (kern.cam.cam_srch_hi) is set non-zero. Reviewed by: nate@rootlabs.org, gibbs@scsiguy.com, ken@kdm.com, scottl@samsco.org MFC after: 1 week
2005-09-16 01:26:17 +00:00
|| (SID_ANSI_REV(&dv->inq_data) > SCSI_REV_2)))
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
typedef enum {
XPT_FLAG_OPEN = 0x01
} xpt_flags;
struct xpt_softc {
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xpt_flags flags;
u_int32_t xpt_generation;
/* number of high powered commands that can go through right now */
STAILQ_HEAD(highpowerlist, ccb_hdr) highpowerq;
int num_highpower;
/* queue for handling async rescan requests. */
TAILQ_HEAD(, ccb_hdr) ccb_scanq;
/* Registered busses */
TAILQ_HEAD(,cam_eb) xpt_busses;
u_int bus_generation;
struct intr_config_hook *xpt_config_hook;
struct mtx xpt_topo_lock;
struct mtx xpt_lock;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
};
static const char quantum[] = "QUANTUM";
Add several quirks: Western Digital Enterprise drives have sorry performance (1.5MB/sec versus 8MB/sec) when doing tagged queueing. Disable tagged queueing for them. Submitted by: Andrew Gallatin <gallatin@cs.duke.edu> Some Sony CDROM drives don't like it when we probe more than one LUN. Verified by: Jean-Marc Zucconi <jmz@FreeBSD.ORG> Some Sony CD-R's don't like multi-LUN probing either. Submitted by: Parag Patel <parag@cgt.com>
1998-09-22 20:41:12 +00:00
static const char sony[] = "SONY";
static const char west_digital[] = "WDIGTL";
The Samsung WN34324U is broken for tagged queueing. PR: kern/9535 Submitted by: NAKAJI Hiroyuki <nakaji@zeisei.dpri.kyoto-u.ac.jp>
1999-01-19 16:59:36 +00:00
static const char samsung[] = "SAMSUNG";
Add a quirk entry for the Seagate Medalist Pro 6 gig drive. It seems this drive has very poor write performance (1.4MB/sec vs. 12MB/sec) with anything more than two oustanding transactions. So, this limits the number of tagged commands to two for that drive. Thanks to Paul van der Zwan for doing a whole lot of testing to confirm this. Reported by: Paul van der Zwan <paulz@trantor.xs4all.nl>
1999-02-26 18:38:06 +00:00
static const char seagate[] = "SEAGATE";
disable tag queing for the MICROP 4421-07 drive. It's an old 2gig drive. convert MICROP to a static string
1999-07-11 06:10:47 +00:00
static const char microp[] = "MICROP";
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
style(9)
2008-01-02 01:45:31 +00:00
static struct xpt_quirk_entry xpt_quirk_table[] =
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
{
/* Reports QUEUE FULL for temporary resource shortages */
Change the Atlas II quirk entries so they work with differential Atlas II's. Also, add a quirk entry for the 2 gig Atlas II. Partially Submitted by: Ted Buswell <tbuswell@mediaone.net>
1998-09-18 19:55:34 +00:00
{ T_DIRECT, SIP_MEDIA_FIXED, quantum, "XP39100*", "*" },
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*quirks*/0, /*mintags*/24, /*maxtags*/32
},
{
/* Reports QUEUE FULL for temporary resource shortages */
Change the Atlas II quirk entries so they work with differential Atlas II's. Also, add a quirk entry for the 2 gig Atlas II. Partially Submitted by: Ted Buswell <tbuswell@mediaone.net>
1998-09-18 19:55:34 +00:00
{ T_DIRECT, SIP_MEDIA_FIXED, quantum, "XP34550*", "*" },
/*quirks*/0, /*mintags*/24, /*maxtags*/32
},
{
/* Reports QUEUE FULL for temporary resource shortages */
{ T_DIRECT, SIP_MEDIA_FIXED, quantum, "XP32275*", "*" },
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*quirks*/0, /*mintags*/24, /*maxtags*/32
},
disable tag queing for the MICROP 4421-07 drive. It's an old 2gig drive. convert MICROP to a static string
1999-07-11 06:10:47 +00:00
{
/* Broken tagged queuing drive */
{ T_DIRECT, SIP_MEDIA_FIXED, microp, "4421-07*", "*" },
/*quirks*/0, /*mintags*/0, /*maxtags*/0
},
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
/* Broken tagged queuing drive */
{ T_DIRECT, SIP_MEDIA_FIXED, "HP", "C372*", "*" },
/*quirks*/0, /*mintags*/0, /*maxtags*/0
},
{
/* Broken tagged queuing drive */
disable tag queing for the MICROP 4421-07 drive. It's an old 2gig drive. convert MICROP to a static string
1999-07-11 06:10:47 +00:00
{ T_DIRECT, SIP_MEDIA_FIXED, microp, "3391*", "x43h" },
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*quirks*/0, /*mintags*/0, /*maxtags*/0
},
The Quantum Atlas III evidently has an identical problem to the Atlas II. It keeps returning queue full until we have reduced the number of tagged openings to the minimum. So, put in a quirk entry with the same work-around. This quirk entry is only for the 9G Atlas III, once someone comes up with inquiry information for the 18G version of that drive, we can quirk it as well. Submitted by: "Johan Granlund" <johan@granlund.nu>
1999-01-07 01:11:24 +00:00
{
/*
* Unfortunately, the Quantum Atlas III has the same
* problem as the Atlas II drives above.
* Reported by: "Johan Granlund" <johan@granlund.nu>
*
* For future reference, the drive with the problem was:
* QUANTUM QM39100TD-SW N1B0
style(9)
2008-01-02 01:45:31 +00:00
*
The Quantum Atlas III evidently has an identical problem to the Atlas II. It keeps returning queue full until we have reduced the number of tagged openings to the minimum. So, put in a quirk entry with the same work-around. This quirk entry is only for the 9G Atlas III, once someone comes up with inquiry information for the 18G version of that drive, we can quirk it as well. Submitted by: "Johan Granlund" <johan@granlund.nu>
1999-01-07 01:11:24 +00:00
* It's possible that Quantum will fix the problem in later
* firmware revisions. If that happens, the quirk entry
* will need to be made specific to the firmware revisions
* with the problem.
style(9)
2008-01-02 01:45:31 +00:00
*
The Quantum Atlas III evidently has an identical problem to the Atlas II. It keeps returning queue full until we have reduced the number of tagged openings to the minimum. So, put in a quirk entry with the same work-around. This quirk entry is only for the 9G Atlas III, once someone comes up with inquiry information for the 18G version of that drive, we can quirk it as well. Submitted by: "Johan Granlund" <johan@granlund.nu>
1999-01-07 01:11:24 +00:00
*/
/* Reports QUEUE FULL for temporary resource shortages */
{ T_DIRECT, SIP_MEDIA_FIXED, quantum, "QM39100*", "*" },
/*quirks*/0, /*mintags*/24, /*maxtags*/32
},
Fix sync rate and transfer rate printouts in xpt_announce_periph(). Sync rates like 4.032MHz were getting printed as 4.32MHz. Also, add a quirk entry for the 18G Quantum Atlas III. Like most other recent Quantum drives, it bogusly reports queue full. Thanks to Andre Albsmeier <andre.albsmeier@mchp.siemens.de> for the Atlas III inquiry information. Reviewed by: gibbs
1999-02-18 18:08:39 +00:00
{
/*
* 18 Gig Atlas III, same problem as the 9G version.
* Reported by: Andre Albsmeier
* <andre.albsmeier@mchp.siemens.de>
*
* For future reference, the drive with the problem was:
* QUANTUM QM318000TD-S N491
*/
/* Reports QUEUE FULL for temporary resource shortages */
{ T_DIRECT, SIP_MEDIA_FIXED, quantum, "QM318000*", "*" },
/*quirks*/0, /*mintags*/24, /*maxtags*/32
},
Disable tagged queueing for the Seagate Elite 9GB drives. They tend to get hung up when you send tags to them too quickly. (CAM is able to recover from the problem, but this just avoids it altogether.) Reviewed by: gibbs Reported by: Bret Ford <bford@uop.cs.uop.edu> and: Martin Renters <martin@tdc.on.ca>
1998-10-14 21:17:39 +00:00
{
/*
* Broken tagged queuing drive
* Reported by: Bret Ford <bford@uop.cs.uop.edu>
* and: Martin Renters <martin@tdc.on.ca>
*/
Add a quirk entry for the Seagate Medalist Pro 6 gig drive. It seems this drive has very poor write performance (1.4MB/sec vs. 12MB/sec) with anything more than two oustanding transactions. So, this limits the number of tagged commands to two for that drive. Thanks to Paul van der Zwan for doing a whole lot of testing to confirm this. Reported by: Paul van der Zwan <paulz@trantor.xs4all.nl>
1999-02-26 18:38:06 +00:00
{ T_DIRECT, SIP_MEDIA_FIXED, seagate, "ST410800*", "71*" },
Disable tagged queueing for the Seagate Elite 9GB drives. They tend to get hung up when you send tags to them too quickly. (CAM is able to recover from the problem, but this just avoids it altogether.) Reviewed by: gibbs Reported by: Bret Ford <bford@uop.cs.uop.edu> and: Martin Renters <martin@tdc.on.ca>
1998-10-14 21:17:39 +00:00
/*quirks*/0, /*mintags*/0, /*maxtags*/0
},
Add a quirk entry for the Seagate Medalist Pro 6 gig drive. It seems this drive has very poor write performance (1.4MB/sec vs. 12MB/sec) with anything more than two oustanding transactions. So, this limits the number of tagged commands to two for that drive. Thanks to Paul van der Zwan for doing a whole lot of testing to confirm this. Reported by: Paul van der Zwan <paulz@trantor.xs4all.nl>
1999-02-26 18:38:06 +00:00
/*
The 4 gig version of the Seagate Medalist Pro seems to have the same trouble with tagged queueing as the 6.5 gig version. So, I've added a quirk entry for it to limit it to two outstanding transactions at a time, just like the 6.5G version. While I'm at it, add a quirk for the 9G version of the drive, since it most likely has the same problem. Submitted by: Jeremy Lea <reg@shale.csir.co.za>
1999-03-07 22:48:50 +00:00
* The Seagate Medalist Pro drives have very poor write
* performance with anything more than 2 tags.
style(9)
2008-01-02 01:45:31 +00:00
*
Add a quirk entry for the Seagate Medalist Pro 6 gig drive. It seems this drive has very poor write performance (1.4MB/sec vs. 12MB/sec) with anything more than two oustanding transactions. So, this limits the number of tagged commands to two for that drive. Thanks to Paul van der Zwan for doing a whole lot of testing to confirm this. Reported by: Paul van der Zwan <paulz@trantor.xs4all.nl>
1999-02-26 18:38:06 +00:00
* Reported by: Paul van der Zwan <paulz@trantor.xs4all.nl>
* Drive: <SEAGATE ST36530N 1444>
The 4 gig version of the Seagate Medalist Pro seems to have the same trouble with tagged queueing as the 6.5 gig version. So, I've added a quirk entry for it to limit it to two outstanding transactions at a time, just like the 6.5G version. While I'm at it, add a quirk for the 9G version of the drive, since it most likely has the same problem. Submitted by: Jeremy Lea <reg@shale.csir.co.za>
1999-03-07 22:48:50 +00:00
*
* Reported by: Jeremy Lea <reg@shale.csir.co.za>
* Drive: <SEAGATE ST34520W 1281>
*
* No one has actually reported that the 9G version
* (ST39140*) of the Medalist Pro has the same problem, but
* we're assuming that it does because the 4G and 6.5G
* versions of the drive are broken.
Add a quirk entry for the Seagate Medalist Pro 6 gig drive. It seems this drive has very poor write performance (1.4MB/sec vs. 12MB/sec) with anything more than two oustanding transactions. So, this limits the number of tagged commands to two for that drive. Thanks to Paul van der Zwan for doing a whole lot of testing to confirm this. Reported by: Paul van der Zwan <paulz@trantor.xs4all.nl>
1999-02-26 18:38:06 +00:00
*/
The 4 gig version of the Seagate Medalist Pro seems to have the same trouble with tagged queueing as the 6.5 gig version. So, I've added a quirk entry for it to limit it to two outstanding transactions at a time, just like the 6.5G version. While I'm at it, add a quirk for the 9G version of the drive, since it most likely has the same problem. Submitted by: Jeremy Lea <reg@shale.csir.co.za>
1999-03-07 22:48:50 +00:00
{
{ T_DIRECT, SIP_MEDIA_FIXED, seagate, "ST34520*", "*"},
/*quirks*/0, /*mintags*/2, /*maxtags*/2
},
{
Add a quirk entry for the Seagate Medalist Pro 6 gig drive. It seems this drive has very poor write performance (1.4MB/sec vs. 12MB/sec) with anything more than two oustanding transactions. So, this limits the number of tagged commands to two for that drive. Thanks to Paul van der Zwan for doing a whole lot of testing to confirm this. Reported by: Paul van der Zwan <paulz@trantor.xs4all.nl>
1999-02-26 18:38:06 +00:00
{ T_DIRECT, SIP_MEDIA_FIXED, seagate, "ST36530*", "*"},
/*quirks*/0, /*mintags*/2, /*maxtags*/2
},
The 4 gig version of the Seagate Medalist Pro seems to have the same trouble with tagged queueing as the 6.5 gig version. So, I've added a quirk entry for it to limit it to two outstanding transactions at a time, just like the 6.5G version. While I'm at it, add a quirk for the 9G version of the drive, since it most likely has the same problem. Submitted by: Jeremy Lea <reg@shale.csir.co.za>
1999-03-07 22:48:50 +00:00
{
{ T_DIRECT, SIP_MEDIA_FIXED, seagate, "ST39140*", "*"},
/*quirks*/0, /*mintags*/2, /*maxtags*/2
},
Disable tagged queueing for the IBM DCAS drives. These drives have poor write performance when tagged queueing is enabled. Although the PR was submitted for the 4 gig version of this drive, the assumption is that the 2 gig version has the same problem. Therefore tagged queueing is disabled for both. Also, update the comment for the Western Digital Enterprise drives to note that the best performance for those drives is achieved when tagged queueing is disabled and write caching is enabled. PR: kern/10398 Submitted by: Hideaki Okada <hokada@isl.melco.co.jp>
1999-03-14 05:15:38 +00:00
{
/*
* Slow when tagged queueing is enabled. Write performance
* steadily drops off with more and more concurrent
* transactions. Best sequential write performance with
* tagged queueing turned off and write caching turned on.
*
* PR: kern/10398
* Submitted by: Hideaki Okada <hokada@isl.melco.co.jp>
* Drive: DCAS-34330 w/ "S65A" firmware.
*
* The drive with the problem had the "S65A" firmware
* revision, and has also been reported (by Stephen J.
* Roznowski <sjr@home.net>) for a drive with the "S61A"
* firmware revision.
*
* Although no one has reported problems with the 2 gig
* version of the DCAS drive, the assumption is that it
* has the same problems as the 4 gig version. Therefore
* this quirk entries disables tagged queueing for all
* DCAS drives.
*/
{ T_DIRECT, SIP_MEDIA_FIXED, "IBM", "DCAS*", "*" },
/*quirks*/0, /*mintags*/0, /*maxtags*/0
},
Add a quirk entry for the Seagate Medalist Pro 6 gig drive. It seems this drive has very poor write performance (1.4MB/sec vs. 12MB/sec) with anything more than two oustanding transactions. So, this limits the number of tagged commands to two for that drive. Thanks to Paul van der Zwan for doing a whole lot of testing to confirm this. Reported by: Paul van der Zwan <paulz@trantor.xs4all.nl>
1999-02-26 18:38:06 +00:00
{
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Broken tagged queuing drive */
Add a quirk entry for the Seagate Medalist Pro 6 gig drive. It seems this drive has very poor write performance (1.4MB/sec vs. 12MB/sec) with anything more than two oustanding transactions. So, this limits the number of tagged commands to two for that drive. Thanks to Paul van der Zwan for doing a whole lot of testing to confirm this. Reported by: Paul van der Zwan <paulz@trantor.xs4all.nl>
1999-02-26 18:38:06 +00:00
{ T_DIRECT, SIP_MEDIA_REMOVABLE, "iomega", "jaz*", "*" },
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*quirks*/0, /*mintags*/0, /*maxtags*/0
Add several quirks: Western Digital Enterprise drives have sorry performance (1.5MB/sec versus 8MB/sec) when doing tagged queueing. Disable tagged queueing for them. Submitted by: Andrew Gallatin <gallatin@cs.duke.edu> Some Sony CDROM drives don't like it when we probe more than one LUN. Verified by: Jean-Marc Zucconi <jmz@FreeBSD.ORG> Some Sony CD-R's don't like multi-LUN probing either. Submitted by: Parag Patel <parag@cgt.com>
1998-09-22 20:41:12 +00:00
},
Add a quirk entry for the CFP2107, another drive with broken tagged queuing support. Ensure that we report that a device supports tagged queuing even if the system is waiting a "command count delay" before starting to use them. If a user disables disconnects on a device ensure that tagged queuing is also turned off. We did the right thing during initial configuration, but could be confused by manual changes.
1998-10-07 03:25:21 +00:00
{
style(9)
2008-01-02 01:45:31 +00:00
/* Broken tagged queuing drive */
By popular demand, back out the CONNER CFP* quirk entry, and return it to its original form. (Originally, it only applied to the CFP 2107.) Hopefully we can come to some conclusion about which Conner drives are broken for tagged queueing.
1999-01-20 01:13:20 +00:00
{ T_DIRECT, SIP_MEDIA_FIXED, "CONNER", "CFP2107*", "*" },
Add a quirk entry for the CFP2107, another drive with broken tagged queuing support. Ensure that we report that a device supports tagged queuing even if the system is waiting a "command count delay" before starting to use them. If a user disables disconnects on a device ensure that tagged queuing is also turned off. We did the right thing during initial configuration, but could be confused by manual changes.
1998-10-07 03:25:21 +00:00
/*quirks*/0, /*mintags*/0, /*maxtags*/0
},
VMware disk volumes are only on LUN 0, and some VMware HBAs have known bugs with commands sent to other than LUN 0. MFC after: 1 day
2006-06-24 16:40:59 +00:00
{
null commit to provide commit message to previous at the request of Sam Leffler: The previous commit established min and maxtags for VMware pseudo disks to fix a submitted PR.
2006-09-11 17:57:23 +00:00
/* This does not support other than LUN 0 */
Grumble. VMWare ESX and VMWare WorkStation have *slightly* different inquiry data (pointlessly different I might add). Pick the common parts. MFC after: 1 day
2006-06-26 05:41:11 +00:00
{ T_DIRECT, SIP_MEDIA_FIXED, "VMware*", "*", "*" },
PR: 103130 Submitted by: Shusuke Shinomiya MFC after: 1 day
2006-09-11 17:34:28 +00:00
CAM_QUIRK_NOLUNS, /*mintags*/2, /*maxtags*/255
VMware disk volumes are only on LUN 0, and some VMware HBAs have known bugs with commands sent to other than LUN 0. MFC after: 1 day
2006-06-24 16:40:59 +00:00
},
The Samsung WN34324U is broken for tagged queueing. PR: kern/9535 Submitted by: NAKAJI Hiroyuki <nakaji@zeisei.dpri.kyoto-u.ac.jp>
1999-01-19 16:59:36 +00:00
{
/*
* Broken tagged queuing drive.
* Submitted by:
* NAKAJI Hiroyuki <nakaji@zeisei.dpri.kyoto-u.ac.jp>
* in PR kern/9535
*/
{ T_DIRECT, SIP_MEDIA_FIXED, samsung, "WN34324U*", "*" },
/*quirks*/0, /*mintags*/0, /*maxtags*/0
},
Add several quirks: Western Digital Enterprise drives have sorry performance (1.5MB/sec versus 8MB/sec) when doing tagged queueing. Disable tagged queueing for them. Submitted by: Andrew Gallatin <gallatin@cs.duke.edu> Some Sony CDROM drives don't like it when we probe more than one LUN. Verified by: Jean-Marc Zucconi <jmz@FreeBSD.ORG> Some Sony CD-R's don't like multi-LUN probing either. Submitted by: Parag Patel <parag@cgt.com>
1998-09-22 20:41:12 +00:00
{
/*
* Slow when tagged queueing is enabled. (1.5MB/sec versus
* 8MB/sec.)
* Submitted by: Andrew Gallatin <gallatin@cs.duke.edu>
Disable tagged queueing for the IBM DCAS drives. These drives have poor write performance when tagged queueing is enabled. Although the PR was submitted for the 4 gig version of this drive, the assumption is that the 2 gig version has the same problem. Therefore tagged queueing is disabled for both. Also, update the comment for the Western Digital Enterprise drives to note that the best performance for those drives is achieved when tagged queueing is disabled and write caching is enabled. PR: kern/10398 Submitted by: Hideaki Okada <hokada@isl.melco.co.jp>
1999-03-14 05:15:38 +00:00
* Best performance with these drives is achieved with
* tagged queueing turned off, and write caching turned on.
Add several quirks: Western Digital Enterprise drives have sorry performance (1.5MB/sec versus 8MB/sec) when doing tagged queueing. Disable tagged queueing for them. Submitted by: Andrew Gallatin <gallatin@cs.duke.edu> Some Sony CDROM drives don't like it when we probe more than one LUN. Verified by: Jean-Marc Zucconi <jmz@FreeBSD.ORG> Some Sony CD-R's don't like multi-LUN probing either. Submitted by: Parag Patel <parag@cgt.com>
1998-09-22 20:41:12 +00:00
*/
{ T_DIRECT, SIP_MEDIA_FIXED, west_digital, "WDE*", "*" },
/*quirks*/0, /*mintags*/0, /*maxtags*/0
},
{
/*
* Slow when tagged queueing is enabled. (1.5MB/sec versus
* 8MB/sec.)
* Submitted by: Andrew Gallatin <gallatin@cs.duke.edu>
Disable tagged queueing for the IBM DCAS drives. These drives have poor write performance when tagged queueing is enabled. Although the PR was submitted for the 4 gig version of this drive, the assumption is that the 2 gig version has the same problem. Therefore tagged queueing is disabled for both. Also, update the comment for the Western Digital Enterprise drives to note that the best performance for those drives is achieved when tagged queueing is disabled and write caching is enabled. PR: kern/10398 Submitted by: Hideaki Okada <hokada@isl.melco.co.jp>
1999-03-14 05:15:38 +00:00
* Best performance with these drives is achieved with
* tagged queueing turned off, and write caching turned on.
Add several quirks: Western Digital Enterprise drives have sorry performance (1.5MB/sec versus 8MB/sec) when doing tagged queueing. Disable tagged queueing for them. Submitted by: Andrew Gallatin <gallatin@cs.duke.edu> Some Sony CDROM drives don't like it when we probe more than one LUN. Verified by: Jean-Marc Zucconi <jmz@FreeBSD.ORG> Some Sony CD-R's don't like multi-LUN probing either. Submitted by: Parag Patel <parag@cgt.com>
1998-09-22 20:41:12 +00:00
*/
{ T_DIRECT, SIP_MEDIA_FIXED, west_digital, "ENTERPRISE", "*" },
/*quirks*/0, /*mintags*/0, /*maxtags*/0
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
},
cam_xpt.c: Add quirk entry for a Samsung drive that doesn't like experiencing the queue full condition. Bump the timeouts for all probe activities to 60s. We don't know what the seletion timeout (or equivelent on other mediums) is for controllers, which can make the transactions at the tail end of a parallel probe take a while to complete. The DPT seems to be a card that takes a long time to see a selection timeout. cam_periph.c: Don't call a device "gone" after a single selection timeout. We need to come up with a better policy. Until that time, you'll have to manually re-scan a bus via camcontrol for the system to decide that a device is really gone. This should give devices experiencing temporary insanity to escape death.
1998-09-20 07:14:36 +00:00
{
/*
* Doesn't handle queue full condition correctly,
* so we need to limit maxtags to what the device
* can handle instead of determining this automatically.
*/
The Samsung WN34324U is broken for tagged queueing. PR: kern/9535 Submitted by: NAKAJI Hiroyuki <nakaji@zeisei.dpri.kyoto-u.ac.jp>
1999-01-19 16:59:36 +00:00
{ T_DIRECT, SIP_MEDIA_FIXED, samsung, "WN321010S*", "*" },
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
/*quirks*/0, /*mintags*/2, /*maxtags*/32
cam_xpt.c: Add quirk entry for a Samsung drive that doesn't like experiencing the queue full condition. Bump the timeouts for all probe activities to 60s. We don't know what the seletion timeout (or equivelent on other mediums) is for controllers, which can make the transactions at the tail end of a parallel probe take a while to complete. The DPT seems to be a card that takes a long time to see a selection timeout. cam_periph.c: Don't call a device "gone" after a single selection timeout. We need to come up with a better policy. Until that time, you'll have to manually re-scan a bus via camcontrol for the system to decide that a device is really gone. This should give devices experiencing temporary insanity to escape death.
1998-09-20 07:14:36 +00:00
},
cam.c: Clear up trailing NULs in cam_strvis. cam_xpt.c: Nuke an experimental quirk entry for the Toshiba 3401. The real problem with this device turned out to be a bug in the aic7xxx driver that was fixed months ago. Add a quirk entry to inhibit multiple lun scanning and serial number probing of DPT RAID volumes. My DPT controller hangs up solid when I do either of these things to a RAID 1 volume.
1998-09-22 04:53:23 +00:00
{
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
/* Really only one LUN */
Do NOLUNS dance for oddball Exabyte. We *really* need to do this as hints. Correct match for A5000 SES instance. PR: 19887 MFC after: 2 weeks
2001-06-04 18:20:52 +00:00
{ T_ENCLOSURE, SIP_MEDIA_FIXED, "SUN", "SENA", "*" },
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
cam.c: Clear up trailing NULs in cam_strvis. cam_xpt.c: Nuke an experimental quirk entry for the Toshiba 3401. The real problem with this device turned out to be a bug in the aic7xxx driver that was fixed months ago. Add a quirk entry to inhibit multiple lun scanning and serial number probing of DPT RAID volumes. My DPT controller hangs up solid when I do either of these things to a RAID 1 volume.
1998-09-22 04:53:23 +00:00
},
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
/* I can't believe we need a quirk for DPT volumes. */
Quirk NOLUNS cabletron device. PR: 11677 Submitted by: "Matthew N. Dodd" <winter@jurai.net>
1999-06-20 18:19:44 +00:00
{ T_ANY, SIP_MEDIA_FIXED|SIP_MEDIA_REMOVABLE, "DPT", "*", "*" },
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
CAM_QUIRK_NOLUNS,
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
/*mintags*/0, /*maxtags*/255
Fix a formatting error. Fix a problem reported by bde: setting SCSI_DELAY to 0 doesn't work. Now, when the user sets SCSI_DELAY to 0, we re-set it to the minimum allowable bus settle delay (100ms). Fix a potential panic in xptfinishconfigfunc() if the CCB passed in is NULL. Reported by, I think, Nicolas Souchu. Fix a memory leak in the same function (we created a path, but didn't free it) by allocating the getdev CCB and path on the stack. Reviewed by: gibbs
1998-09-17 23:58:53 +00:00
},
Add several quirks: Western Digital Enterprise drives have sorry performance (1.5MB/sec versus 8MB/sec) when doing tagged queueing. Disable tagged queueing for them. Submitted by: Andrew Gallatin <gallatin@cs.duke.edu> Some Sony CDROM drives don't like it when we probe more than one LUN. Verified by: Jean-Marc Zucconi <jmz@FreeBSD.ORG> Some Sony CD-R's don't like multi-LUN probing either. Submitted by: Parag Patel <parag@cgt.com>
1998-09-22 20:41:12 +00:00
{
/*
Generalize the quirk entry that disables multi-lun probing for Sony CDROM drives. It seems that quite a few (possibly all?) of their drives respond to inquiries on multiple luns. Hopefully we can detect problems like this in the probe phase at some point. For now, this is a pretty functional solution.
1998-11-04 19:56:24 +00:00
* Many Sony CDROM drives don't like multi-LUN probing.
Add several quirks: Western Digital Enterprise drives have sorry performance (1.5MB/sec versus 8MB/sec) when doing tagged queueing. Disable tagged queueing for them. Submitted by: Andrew Gallatin <gallatin@cs.duke.edu> Some Sony CDROM drives don't like it when we probe more than one LUN. Verified by: Jean-Marc Zucconi <jmz@FreeBSD.ORG> Some Sony CD-R's don't like multi-LUN probing either. Submitted by: Parag Patel <parag@cgt.com>
1998-09-22 20:41:12 +00:00
*/
Quirk NOLUNS cabletron device. PR: 11677 Submitted by: "Matthew N. Dodd" <winter@jurai.net>
1999-06-20 18:19:44 +00:00
{ T_CDROM, SIP_MEDIA_REMOVABLE, sony, "CD-ROM CDU*", "*" },
Add several quirks: Western Digital Enterprise drives have sorry performance (1.5MB/sec versus 8MB/sec) when doing tagged queueing. Disable tagged queueing for them. Submitted by: Andrew Gallatin <gallatin@cs.duke.edu> Some Sony CDROM drives don't like it when we probe more than one LUN. Verified by: Jean-Marc Zucconi <jmz@FreeBSD.ORG> Some Sony CD-R's don't like multi-LUN probing either. Submitted by: Parag Patel <parag@cgt.com>
1998-09-22 20:41:12 +00:00
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
{
/*
* This drive doesn't like multiple LUN probing.
* Submitted by: Parag Patel <parag@cgt.com>
*/
Quirk NOLUNS cabletron device. PR: 11677 Submitted by: "Matthew N. Dodd" <winter@jurai.net>
1999-06-20 18:19:44 +00:00
{ T_WORM, SIP_MEDIA_REMOVABLE, sony, "CD-R CDU9*", "*" },
Add several quirks: Western Digital Enterprise drives have sorry performance (1.5MB/sec versus 8MB/sec) when doing tagged queueing. Disable tagged queueing for them. Submitted by: Andrew Gallatin <gallatin@cs.duke.edu> Some Sony CDROM drives don't like it when we probe more than one LUN. Verified by: Jean-Marc Zucconi <jmz@FreeBSD.ORG> Some Sony CD-R's don't like multi-LUN probing either. Submitted by: Parag Patel <parag@cgt.com>
1998-09-22 20:41:12 +00:00
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
The Yamaha CDR100 doesn't seem to like multi-LUN probing. PR: kern/20347 Submitted by: Andre Albsmeier <andre.albsmeier@mchp.siemens.de>
2000-08-08 06:49:33 +00:00
{
{ T_WORM, SIP_MEDIA_REMOVABLE, "YAMAHA", "CDR100*", "*" },
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
Disable multi-lun probing and serial number inquiries for the Exabyte 8200.
1998-10-06 19:27:19 +00:00
/*
* The 8200 doesn't like multi-lun probing, and probably
* don't like serial number requests either.
*/
{
T_SEQUENTIAL, SIP_MEDIA_REMOVABLE, "EXABYTE",
"EXB-8200*", "*"
},
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
Disable multi-lun probing and serial number inquiries for the Exabyte 8200.
1998-10-06 19:27:19 +00:00
},
Do NOLUNS dance for oddball Exabyte. We *really* need to do this as hints. Correct match for A5000 SES instance. PR: 19887 MFC after: 2 weeks
2001-06-04 18:20:52 +00:00
{
/*
* Let's try the same as above, but for a drive that says
* it's an IPL-6860 but is actually an EXB 8200.
*/
{
T_SEQUENTIAL, SIP_MEDIA_REMOVABLE, "EXABYTE",
"IPL-6860*", "*"
},
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
Do NOLUNS dance for oddball Exabyte. We *really* need to do this as hints. Correct match for A5000 SES instance. PR: 19887 MFC after: 2 weeks
2001-06-04 18:20:52 +00:00
},
Disable multi-lun probing on Hitachi DK31* drives. PR: misc/18793 Submitted by: Paul Haddad <paul@pth.com>
2000-06-04 03:17:37 +00:00
{
/*
* These Hitachi drives don't like multi-lun probing.
* The PR submitter has a DK319H, but says that the Linux
* kernel has a similar work-around for the DK312 and DK314,
* so all DK31* drives are quirked here.
* PR: misc/18793
* Submitted by: Paul Haddad <paul@pth.com>
*/
{ T_DIRECT, SIP_MEDIA_FIXED, "HITACHI", "DK31*", "*" },
CAM_QUIRK_NOLUNS, /*mintags*/2, /*maxtags*/255
},
Disabled tagged commands for Hitachi Jura-C series with J8A8 firmware. PR: 23536 Submitted by: amagai@nue.org
2001-11-19 23:23:28 +00:00
{
/*
* The Hitachi CJ series with J8A8 firmware apparantly has
* problems with tagged commands.
* PR: 23536
* Reported by: amagai@nue.org
*/
{ T_DIRECT, SIP_MEDIA_FIXED, "HITACHI", "DK32CJ*", "J8A8" },
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
Note that the large Hitachi's are also CAM_QUIRK_HILUNS
2002-09-23 04:55:32 +00:00
{
/*
* These are the large storage arrays.
* Submitted by: William Carrel <william.carrel@infospace.com>
*/
{ T_DIRECT, SIP_MEDIA_FIXED, "HITACHI", "OPEN*", "*" },
CAM_QUIRK_HILUNS, 2, 1024
},
This old firmware of the TDC3620 hangs the SCSI bus upon serial number requests. Don't ask it so.
1998-11-25 13:50:10 +00:00
{
/*
* This old revision of the TDC3600 is also SCSI-1, and
* hangs upon serial number probing.
*/
{
T_SEQUENTIAL, SIP_MEDIA_REMOVABLE, "TANDBERG",
" TDC 3600", "U07:"
},
CAM_QUIRK_NOSERIAL, /*mintags*/0, /*maxtags*/0
},
Quirk NOLUNS CALIPER CP150.
1999-05-11 02:41:00 +00:00
{
/*
* Would repond to all LUNs if asked for.
*/
{
T_SEQUENTIAL, SIP_MEDIA_REMOVABLE, "CALIPER",
"CP150", "*"
},
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
Quirk this drive- all luns show up the same. Submitted by: Stephen McKay <syssgm@detir.qld.gov.au>
1999-05-11 15:44:39 +00:00
{
/*
* Would repond to all LUNs if asked for.
*/
{
T_SEQUENTIAL, SIP_MEDIA_REMOVABLE, "KENNEDY",
"96X2*", "*"
},
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
Quirk NOLUNS cabletron device. PR: 11677 Submitted by: "Matthew N. Dodd" <winter@jurai.net>
1999-06-20 18:19:44 +00:00
{
/* Submitted by: Matthew Dodd <winter@jurai.net> */
{ T_PROCESSOR, SIP_MEDIA_FIXED, "Cabletrn", "EA41*", "*" },
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
{
/* Submitted by: Matthew Dodd <winter@jurai.net> */
{ T_PROCESSOR, SIP_MEDIA_FIXED, "CABLETRN", "EA41*", "*" },
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
Quirk entry for TeraSolutions TRC-22 RAID controller to keep mintags from dropping below 55 due to temporary resource shortages.
2000-07-13 07:58:04 +00:00
{
/* TeraSolutions special settings for TRC-22 RAID */
{ T_DIRECT, SIP_MEDIA_FIXED, "TERASOLU", "TRC-22", "*" },
/*quirks*/0, /*mintags*/55, /*maxtags*/255
},
Make Veritas Storage Appliance a HILUNS device. MFC after: 1 day
2002-04-24 00:00:55 +00:00
{
/* Veritas Storage Appliance */
{ T_DIRECT, SIP_MEDIA_FIXED, "VERITAS", "*", "*" },
CAM_QUIRK_HILUNS, /*mintags*/2, /*maxtags*/1024
},
Mark the (ancient) MaxOptix Tahiti 1 drive as a known rogue that would respond to all LUNs. Reviewed by: ken
2001-01-23 10:49:30 +00:00
{
/*
* Would respond to all LUNs. Device type and removable
* flag are jumper-selectable.
*/
{ T_ANY, SIP_MEDIA_REMOVABLE|SIP_MEDIA_FIXED, "MaxOptix",
"Tahiti 1", "*"
},
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
Add quirk for easyRAID ER5 alias areca ARC-6010. It reports itself as SCSI-3 but doesnt like getting probed on high luns because it hangs hard after finding itself again on lun 32... Suggested by: Kenneth Merry
2005-01-25 08:59:06 +00:00
{
/* EasyRAID E5A aka. areca ARC-6010 */
{ T_DIRECT, SIP_MEDIA_FIXED, "easyRAID", "*", "*" },
CAM_QUIRK_NOHILUNS, /*mintags*/2, /*maxtags*/255
},
New Dell 1950/2950 SES backplane drops off the bus if you poke at greater then lun 0. MFC after: 1 week
2006-09-16 17:35:47 +00:00
{
{ T_ENCLOSURE, SIP_MEDIA_FIXED, "DP", "BACKPLANE", "*" },
CAM_QUIRK_NOLUNS, /*mintags*/0, /*maxtags*/0
},
Disable multi-lun probing and serial number inquiries for the Exabyte 8200.
1998-10-06 19:27:19 +00:00
{
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Default tagged queuing parameters for all devices */
{
T_ANY, SIP_MEDIA_REMOVABLE|SIP_MEDIA_FIXED,
/*vendor*/"*", /*product*/"*", /*revision*/"*"
},
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
/*quirks*/0, /*mintags*/2, /*maxtags*/255
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
},
};
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
static const int xpt_quirk_table_size =
sizeof(xpt_quirk_table) / sizeof(*xpt_quirk_table);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
/* Queues for our software interrupt handler */
Back out the previous change to the queue(3) interface. It was not discussed and should probably not happen. Requested by: msmith and others
2000-05-26 02:09:24 +00:00
typedef TAILQ_HEAD(cam_isrq, ccb_hdr) cam_isrq_t;
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
typedef TAILQ_HEAD(cam_simq, cam_sim) cam_simq_t;
static cam_simq_t cam_simq;
static struct mtx cam_simq_lock;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
/* Pointers to software interrupt handlers */
static void *cambio_ih;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
*sigh* I must have been on something that night. Make xpt_periph an extern with the original in cam_xpt.c instead of replicating xpt_periph in all the sources using it (and hence not initialising it)
2000-05-07 18:04:50 +00:00
struct cam_periph *xpt_periph;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static periph_init_t xpt_periph_init;
static periph_init_t probe_periph_init;
static struct periph_driver xpt_driver =
{
xpt_periph_init, "xpt",
TAILQ_HEAD_INITIALIZER(xpt_driver.units)
};
static struct periph_driver probe_driver =
{
probe_periph_init, "probe",
TAILQ_HEAD_INITIALIZER(probe_driver.units)
};
Change the peripheral driver list from a linker set to module driven driver registration. This should allow things like da, sa, cd etc to be in seperate KLD's to the cam core and make them preloadable.
2001-02-07 07:05:59 +00:00
PERIPHDRIVER_DECLARE(xpt, xpt_driver);
PERIPHDRIVER_DECLARE(probe, probe_driver);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static d_open_t xptopen;
static d_close_t xptclose;
static d_ioctl_t xptioctl;
This commit should be a extensive NO-OP: Reformat and initialize correctly all "struct cdevsw". Initialize the d_maj and d_bmaj fields. The d_reset field was not removed, although it is never used. I used a program to do most of this, so all the files now use the same consistent format. Please keep it that way. Vinum and i4b not modified, patches emailed to respective authors.
1999-05-30 16:53:49 +00:00
static struct cdevsw xpt_cdevsw = {
Device megapatch 4/6: Introduce d_version field in struct cdevsw, this must always be initialized to D_VERSION. Flip sense of D_NOGIANT flag to D_NEEDGIANT, this involves removing four D_NOGIANT flags and adding 145 D_NEEDGIANT flags.
2004-02-21 21:10:55 +00:00
.d_version = D_VERSION,
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
.d_flags = 0,
Gigacommit to improve device-driver source compatibility between branches: Initialize struct cdevsw using C99 sparse initializtion and remove all initializations to default values. This patch is automatically generated and has been tested by compiling LINT with all the fields in struct cdevsw in reverse order on alpha, sparc64 and i386. Approved by: re(scottl)
2003-03-03 12:15:54 +00:00
.d_open = xptopen,
.d_close = xptclose,
.d_ioctl = xptioctl,
.d_name = "xpt",
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
};
/* Storage for debugging datastructures */
#ifdef CAMDEBUG
struct cam_path *cam_dpath;
u_int32_t cam_dflags;
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
u_int32_t cam_debug_delay;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
#endif
#if defined(CAM_DEBUG_FLAGS) && !defined(CAMDEBUG)
#error "You must have options CAMDEBUG to use options CAM_DEBUG_FLAGS"
#endif
/*
* In order to enable the CAM_DEBUG_* options, the user must have CAMDEBUG
* enabled. Also, the user must have either none, or all of CAM_DEBUG_BUS,
* CAM_DEBUG_TARGET, and CAM_DEBUG_LUN specified.
*/
#if defined(CAM_DEBUG_BUS) || defined(CAM_DEBUG_TARGET) \
|| defined(CAM_DEBUG_LUN)
#ifdef CAMDEBUG
#if !defined(CAM_DEBUG_BUS) || !defined(CAM_DEBUG_TARGET) \
|| !defined(CAM_DEBUG_LUN)
#error "You must define all or none of CAM_DEBUG_BUS, CAM_DEBUG_TARGET \
and CAM_DEBUG_LUN"
#endif /* !CAM_DEBUG_BUS || !CAM_DEBUG_TARGET || !CAM_DEBUG_LUN */
#else /* !CAMDEBUG */
#error "You must use options CAMDEBUG if you use the CAM_DEBUG_* options"
#endif /* CAMDEBUG */
#endif /* CAM_DEBUG_BUS || CAM_DEBUG_TARGET || CAM_DEBUG_LUN */
Call the xpt_init() hook during at the start of the configure() process via SYSINIT(). This gets a little closer to making cam useable as a module.
1999-01-19 00:13:05 +00:00
/* Our boot-time initialization hook */
CAM, the module: scbus, da, cd, and st wrapped in one module. Make the umass driver depend on this module. Makes it possible to compile the kernel without SCSI support and load it when for example a USB floppy is conencted.
2000-08-13 18:49:40 +00:00
static int cam_module_event_handler(module_t, int /*modeventtype_t*/, void *);
static moduledata_t cam_moduledata = {
"cam",
cam_module_event_handler,
NULL
};
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
static int xpt_init(void *);
CAM, the module: scbus, da, cd, and st wrapped in one module. Make the umass driver depend on this module. Makes it possible to compile the kernel without SCSI support and load it when for example a USB floppy is conencted.
2000-08-13 18:49:40 +00:00
DECLARE_MODULE(cam, cam_moduledata, SI_SUB_CONFIGURE, SI_ORDER_SECOND);
MODULE_VERSION(cam, 1);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static 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);
static void xpt_release_path(struct cam_path *path);
static void xpt_async_bcast(struct async_list *async_head,
u_int32_t async_code,
struct cam_path *path,
void *async_arg);
Fix async notifications for listners registered to wildcard nodes. For example, a client registered to receive specific events for bus 0, target *, lun *, was not receiving notifications. Reviewed by: ken@FreeBSD.org
2000-11-06 20:12:07 +00:00
static void xpt_dev_async(u_int32_t async_code,
struct cam_eb *bus,
struct cam_et *target,
struct cam_ed *device,
void *async_arg);
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
static path_id_t xptnextfreepathid(void);
static path_id_t xptpathid(const char *sim_name, int sim_unit, int sim_bus);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static union ccb *xpt_get_ccb(struct cam_ed *device);
static int xpt_schedule_dev(struct camq *queue, cam_pinfo *dev_pinfo,
u_int32_t new_priority);
static void xpt_run_dev_allocq(struct cam_eb *bus);
static void xpt_run_dev_sendq(struct cam_eb *bus);
static timeout_t xpt_release_devq_timeout;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
static void xpt_release_simq_timeout(void *arg) __unused;
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
static void xpt_release_bus(struct cam_eb *bus);
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
static void xpt_release_devq_device(struct cam_ed *dev, u_int count,
int run_queue);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static struct cam_et*
xpt_alloc_target(struct cam_eb *bus, target_id_t target_id);
static void xpt_release_target(struct cam_eb *bus, struct cam_et *target);
static struct cam_ed*
xpt_alloc_device(struct cam_eb *bus, struct cam_et *target,
lun_id_t lun_id);
static void xpt_release_device(struct cam_eb *bus, struct cam_et *target,
struct cam_ed *device);
static u_int32_t xpt_dev_ccbq_resize(struct cam_path *path, int newopenings);
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_scan_bus(struct cam_periph *periph, union ccb *ccb);
static void xpt_scan_lun(struct cam_periph *periph,
struct cam_path *path, cam_flags flags,
union ccb *ccb);
static void xptscandone(struct cam_periph *periph, union ccb *done_ccb);
static xpt_busfunc_t xptconfigbuscountfunc;
static xpt_busfunc_t xptconfigfunc;
static void xpt_config(void *arg);
static xpt_devicefunc_t xptpassannouncefunc;
static void xpt_finishconfig(struct cam_periph *periph, union ccb *ccb);
static void xptaction(struct cam_sim *sim, union ccb *work_ccb);
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
static void xptpoll(struct cam_sim *sim);
- Overhaul the software interrupt code to use interrupt threads for each type of software interrupt. Roughly, what used to be a bit in spending now maps to a swi thread. Each thread can have multiple handlers, just like a hardware interrupt thread. - Instead of using a bitmask of pending interrupts, we schedule the specific software interrupt thread to run, so spending, NSWI, and the shandlers array are no longer needed. We can now have an arbitrary number of software interrupt threads. When you register a software interrupt thread via sinthand_add(), you get back a struct intrhand that you pass to sched_swi() when you wish to schedule your swi thread to run. - Convert the name of 'struct intrec' to 'struct intrhand' as it is a bit more intuitive. Also, prefix all the members of struct intrhand with 'ih_'. - Make swi_net() a MI function since there is now no point in it being MD. Submitted by: cp
2000-10-25 05:19:40 +00:00
static void camisr(void *);
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
static void camisr_runqueue(void *);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
static int xpt_for_all_busses(xpt_busfunc_t *tr_func, void *arg);
static int xpt_for_all_devices(xpt_devicefunc_t *tr_func,
void *arg);
static xpt_devicefunc_t xptsetasyncfunc;
static xpt_busfunc_t xptsetasyncbusfunc;
static cam_status xptregister(struct cam_periph *periph,
void *arg);
static cam_status proberegister(struct cam_periph *periph,
void *arg);
static void probeschedule(struct cam_periph *probe_periph);
static void probestart(struct cam_periph *periph, union ccb *start_ccb);
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
static void proberequestdefaultnegotiation(struct cam_periph *periph);
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
static int proberequestbackoff(struct cam_periph *periph,
struct cam_ed *device);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static void probedone(struct cam_periph *periph, union ccb *done_ccb);
static void probecleanup(struct cam_periph *periph);
static void xpt_find_quirk(struct cam_ed *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
static void xpt_devise_transport(struct cam_path *path);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static void xpt_set_transfer_settings(struct ccb_trans_settings *cts,
Correct problems with xpt_set_transfer_settings and async transfer negotiation changes with wildcarded paths.
1998-09-25 22:35:56 +00:00
struct cam_ed *device,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
int async_update);
Fix a few problems with the tag delay code: - Tagged devices were limited to one transaction (oops) - We revert to untagged with a tag delay if the user changed the transfer negotiation values (via perhaps camcontrol some day). - xpt_async did not use the expanded path in some cases which could cause a panic.
1998-09-24 22:43:54 +00:00
static void xpt_toggle_tags(struct cam_path *path);
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
static void xpt_start_tags(struct cam_path *path);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static __inline int xpt_schedule_dev_allocq(struct cam_eb *bus,
struct cam_ed *dev);
static __inline int xpt_schedule_dev_sendq(struct cam_eb *bus,
struct cam_ed *dev);
static __inline int periph_is_queued(struct cam_periph *periph);
static __inline int device_is_alloc_queued(struct cam_ed *device);
static __inline int device_is_send_queued(struct cam_ed *device);
static __inline int dev_allocq_is_runnable(struct cam_devq *devq);
static __inline int
xpt_schedule_dev_allocq(struct cam_eb *bus, struct cam_ed *dev)
{
int retval;
if (dev->ccbq.devq_openings > 0) {
if ((dev->flags & CAM_DEV_RESIZE_QUEUE_NEEDED) != 0) {
cam_ccbq_resize(&dev->ccbq,
dev->ccbq.dev_openings
+ dev->ccbq.dev_active);
dev->flags &= ~CAM_DEV_RESIZE_QUEUE_NEEDED;
}
cam_periph.c: Move handling of CAM_AUTOSENSE_FAIL into block dealing with all other scsi status errors. cam_queue.c: cam_queue.h: Fix 'off by one' heap bug in a more efficient manner. Since heap algorithms like to deal with indexes started from 1, offset our heap array pointer at allocation time to make this so for a C environment. This makes the implementation of the algorithm a bit more efficient. cam_xpt.c: Use macros for accessing the head of the heap so that code is isolated from implementation details of the heap.
1999-04-19 21:26:08 +00:00
/*
* The priority of a device waiting for CCB resources
* is that of the the highest priority peripheral driver
* enqueued.
*/
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
retval = xpt_schedule_dev(&bus->sim->devq->alloc_queue,
&dev->alloc_ccb_entry.pinfo,
style(9)
2008-01-02 01:45:31 +00:00
CAMQ_GET_HEAD(&dev->drvq)->priority);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} else {
retval = 0;
}
return (retval);
}
static __inline int
xpt_schedule_dev_sendq(struct cam_eb *bus, struct cam_ed *dev)
{
int retval;
if (dev->ccbq.dev_openings > 0) {
cam_periph.c: Move handling of CAM_AUTOSENSE_FAIL into block dealing with all other scsi status errors. cam_queue.c: cam_queue.h: Fix 'off by one' heap bug in a more efficient manner. Since heap algorithms like to deal with indexes started from 1, offset our heap array pointer at allocation time to make this so for a C environment. This makes the implementation of the algorithm a bit more efficient. cam_xpt.c: Use macros for accessing the head of the heap so that code is isolated from implementation details of the heap.
1999-04-19 21:26:08 +00:00
/*
* The priority of a device waiting for controller
* resources is that of the the highest priority CCB
* enqueued.
*/
retval =
xpt_schedule_dev(&bus->sim->devq->send_queue,
&dev->send_ccb_entry.pinfo,
CAMQ_GET_HEAD(&dev->ccbq.queue)->priority);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} else {
retval = 0;
}
return (retval);
}
static __inline int
periph_is_queued(struct cam_periph *periph)
{
return (periph->pinfo.index != CAM_UNQUEUED_INDEX);
}
static __inline int
device_is_alloc_queued(struct cam_ed *device)
{
return (device->alloc_ccb_entry.pinfo.index != CAM_UNQUEUED_INDEX);
}
static __inline int
device_is_send_queued(struct cam_ed *device)
{
return (device->send_ccb_entry.pinfo.index != CAM_UNQUEUED_INDEX);
}
static __inline int
dev_allocq_is_runnable(struct cam_devq *devq)
{
/*
* Have work to do.
* Have space to do more work.
* Allowed to do work.
*/
return ((devq->alloc_queue.qfrozen_cnt == 0)
&& (devq->alloc_queue.entries > 0)
&& (devq->alloc_openings > 0));
}
static void
xpt_periph_init()
{
Convert the xpt, pass, pt and target drivers to use the new make_dev()/destroy_dev() interface.
1999-11-17 04:59:09 +00:00
make_dev(&xpt_cdevsw, 0, UID_ROOT, GID_OPERATOR, 0600, "xpt0");
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
static void
probe_periph_init()
{
}
static void
xptdone(struct cam_periph *periph, union ccb *done_ccb)
{
/* Caller will release the CCB */
wakeup(&done_ccb->ccb_h.cbfcnp);
}
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
xptopen(struct cdev *dev, int flags, int fmt, struct thread *td)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
Check to make sure that this device is opened read-write, not just read only. Previously, if the device was chmoded 644, someone could open it with the O_RDONLY flag and issue any ioctl to the device. Reviewed by: imp, gibbs
1998-09-16 00:11:53 +00:00
/*
* Only allow read-write access.
*/
if (((flags & FWRITE) == 0) || ((flags & FREAD) == 0))
return(EPERM);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* We don't allow nonblocking access.
*/
if ((flags & O_NONBLOCK) != 0) {
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
printf("%s: can't do nonblocking access\n", devtoname(dev));
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(ENODEV);
}
/* Mark ourselves open */
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xsoftc.flags |= XPT_FLAG_OPEN;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_unlock(&xsoftc.xpt_lock);
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(0);
}
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
xptclose(struct cdev *dev, int flag, int fmt, struct thread *td)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
/* Mark ourselves closed */
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xsoftc.flags &= ~XPT_FLAG_OPEN;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_unlock(&xsoftc.xpt_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(0);
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
/*
* 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.
*/
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
xptioctl(struct cdev *dev, u_long cmd, caddr_t addr, int flag, struct thread *td)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
int error;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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
Allow XPT_PATH_INQ to be performed on the XPT device, as per the CAM spec. Return ENOTSUP for XPT_ENG_INQ and XPT_ENG_EXEC when performed on the XPT device. Reviewed by: gibbs
2000-10-30 23:30:28 +00:00
* passthrough driver. XPT_PATH_INQ is an exception to this, as stated
* in the CAM spec.
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
*/
case CAMIOCOMMAND: {
union ccb *ccb;
union ccb *inccb;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
struct cam_eb *bus;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
inccb = (union ccb *)addr;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
bus = xpt_find_bus(inccb->ccb_h.path_id);
if (bus == NULL) {
error = EINVAL;
break;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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)) {
error = EINVAL;
break;
}
/* FALLTHROUGH */
Allow XPT_PATH_INQ to be performed on the XPT device, as per the CAM spec. Return ENOTSUP for XPT_ENG_INQ and XPT_ENG_EXEC when performed on the XPT device. Reviewed by: gibbs
2000-10-30 23:30:28 +00:00
case XPT_PATH_INQ:
Treat engine inquiry the same as path inquiry in xptioctl(). Return ENOTSUP for any opcode that is not supported by the XPT device. Add back a missing local declaration that seems to have been deleted by my last commit.
2000-10-31 22:07:02 +00:00
case XPT_ENG_INQ:
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
case XPT_SCAN_LUN:
Revert a driver API change to xpt_alloc_ccb that isn't necessary. Fix a couple of associated error checks.
2007-04-18 04:58:53 +00:00
ccb = xpt_alloc_ccb();
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
CAM_SIM_LOCK(bus->sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Create a path using the bus, target, and lun the
* user passed in.
*/
if (xpt_create_path(&ccb->ccb_h.path, xpt_periph,
inccb->ccb_h.path_id,
inccb->ccb_h.target_id,
inccb->ccb_h.target_lun) !=
CAM_REQ_CMP){
error = EINVAL;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
CAM_SIM_UNLOCK(bus->sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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);
ccb->ccb_h.cbfcnp = xptdone;
cam_periph_runccb(ccb, NULL, 0, 0, NULL);
bcopy(ccb, inccb, sizeof(union ccb));
xpt_free_path(ccb->ccb_h.path);
xpt_free_ccb(ccb);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
CAM_SIM_UNLOCK(bus->sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
case XPT_DEBUG: {
union ccb ccb;
/*
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
* This is an immediate CCB, so it's okay to
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* allocate it on the stack.
*/
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
CAM_SIM_LOCK(bus->sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Create a path using the bus, target, and lun the
* user passed in.
*/
if (xpt_create_path(&ccb.ccb_h.path, xpt_periph,
inccb->ccb_h.path_id,
inccb->ccb_h.target_id,
inccb->ccb_h.target_lun) !=
CAM_REQ_CMP){
error = EINVAL;
Add a missing unlock to an error path fro XPT_DEBUG handling.
2008-02-26 08:09:29 +00:00
CAM_SIM_UNLOCK(bus->sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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);
ccb.ccb_h.cbfcnp = xptdone;
xpt_action(&ccb);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
CAM_SIM_UNLOCK(bus->sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
bcopy(&ccb, inccb, sizeof(union ccb));
xpt_free_path(ccb.ccb_h.path);
break;
}
case XPT_DEV_MATCH: {
struct cam_periph_map_info mapinfo;
Generalize the quirk entry for the Conner CFP* drives. It did just cover the CFP2107, but it appears (not surprisingly) that the 1 gig and 4 gig versions of that drive have the same problem with tagged queueing. Also, fix the problem reported in PR kern/9482. The XPT_DEV_MATCH case in xptioctl() wasn't putting a proper path in the CCB before it called xpt_action(). When CAMDEBUG is defined, and CAM_DEBUG_TRACE debugging is turned on, the CAM_DEBUG statement at the beginning of xpt_action would end up deferencing a NULL path pointer. That of course caused a panic. My solution is to just stick the xpt peripheral's path in the CCB. PR: kern/9482 Reviewed by: gibbs
1999-01-19 01:02:47 +00:00
struct cam_path *old_path;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* We can't deal with physical addresses for this
* type of transaction.
*/
if (inccb->ccb_h.flags & CAM_DATA_PHYS) {
error = EINVAL;
break;
}
Generalize the quirk entry for the Conner CFP* drives. It did just cover the CFP2107, but it appears (not surprisingly) that the 1 gig and 4 gig versions of that drive have the same problem with tagged queueing. Also, fix the problem reported in PR kern/9482. The XPT_DEV_MATCH case in xptioctl() wasn't putting a proper path in the CCB before it called xpt_action(). When CAMDEBUG is defined, and CAM_DEBUG_TRACE debugging is turned on, the CAM_DEBUG statement at the beginning of xpt_action would end up deferencing a NULL path pointer. That of course caused a panic. My solution is to just stick the xpt peripheral's path in the CCB. PR: kern/9482 Reviewed by: gibbs
1999-01-19 01:02:47 +00:00
/*
* 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;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
bzero(&mapinfo, sizeof(mapinfo));
/*
* Map the pattern and match buffers into kernel
* virtual address space.
*/
error = cam_periph_mapmem(inccb, &mapinfo);
Generalize the quirk entry for the Conner CFP* drives. It did just cover the CFP2107, but it appears (not surprisingly) that the 1 gig and 4 gig versions of that drive have the same problem with tagged queueing. Also, fix the problem reported in PR kern/9482. The XPT_DEV_MATCH case in xptioctl() wasn't putting a proper path in the CCB before it called xpt_action(). When CAMDEBUG is defined, and CAM_DEBUG_TRACE debugging is turned on, the CAM_DEBUG statement at the beginning of xpt_action would end up deferencing a NULL path pointer. That of course caused a panic. My solution is to just stick the xpt peripheral's path in the CCB. PR: kern/9482 Reviewed by: gibbs
1999-01-19 01:02:47 +00:00
if (error) {
inccb->ccb_h.path = old_path;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
Generalize the quirk entry for the Conner CFP* drives. It did just cover the CFP2107, but it appears (not surprisingly) that the 1 gig and 4 gig versions of that drive have the same problem with tagged queueing. Also, fix the problem reported in PR kern/9482. The XPT_DEV_MATCH case in xptioctl() wasn't putting a proper path in the CCB before it called xpt_action(). When CAMDEBUG is defined, and CAM_DEBUG_TRACE debugging is turned on, the CAM_DEBUG statement at the beginning of xpt_action would end up deferencing a NULL path pointer. That of course caused a panic. My solution is to just stick the xpt peripheral's path in the CCB. PR: kern/9482 Reviewed by: gibbs
1999-01-19 01:02:47 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* 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);
Generalize the quirk entry for the Conner CFP* drives. It did just cover the CFP2107, but it appears (not surprisingly) that the 1 gig and 4 gig versions of that drive have the same problem with tagged queueing. Also, fix the problem reported in PR kern/9482. The XPT_DEV_MATCH case in xptioctl() wasn't putting a proper path in the CCB before it called xpt_action(). When CAMDEBUG is defined, and CAM_DEBUG_TRACE debugging is turned on, the CAM_DEBUG statement at the beginning of xpt_action would end up deferencing a NULL path pointer. That of course caused a panic. My solution is to just stick the xpt peripheral's path in the CCB. PR: kern/9482 Reviewed by: gibbs
1999-01-19 01:02:47 +00:00
inccb->ccb_h.path = old_path;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
error = 0;
break;
}
default:
Treat engine inquiry the same as path inquiry in xptioctl(). Return ENOTSUP for any opcode that is not supported by the XPT device. Add back a missing local declaration that seems to have been deleted by my last commit.
2000-10-31 22:07:02 +00:00
error = ENOTSUP;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
}
Release the bus reference that is acquired when doing a CAMIOCOMMAND ioctl. This should make SIM removal a bit more robust.
2007-05-16 16:57:21 +00:00
xpt_release_bus(bus);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
* we do it with lock protection.
style(9)
2008-01-02 01:45:31 +00:00
*
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +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;
u_int cur_generation;
Add a "dummy light" (actually two dummy lights) to catch people who don't have the passthrough device configured in their kernel. This will hopefully reduce the number of people complaining that they can't get {camcontrol, xmcd, tosha, cdrecord, etc.} to work. Reviewed by: gibbs
1998-10-12 21:54:13 +00:00
int base_periph_found;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
int splbreaknum;
ccb = (union ccb *)addr;
unit = ccb->cgdl.unit_number;
name = ccb->cgdl.periph_name;
/*
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
* Every 100 devices, we want to drop our lock protection to
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* give the software interrupt handler a chance to run.
* Most systems won't run into this check, but this should
* avoid starvation in the software interrupt handler in
* large systems.
*/
splbreaknum = 100;
ccb = (union ccb *)addr;
Add a "dummy light" (actually two dummy lights) to catch people who don't have the passthrough device configured in their kernel. This will hopefully reduce the number of people complaining that they can't get {camcontrol, xmcd, tosha, cdrecord, etc.} to work. Reviewed by: gibbs
1998-10-12 21:54:13 +00:00
base_periph_found = 0;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* 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 */
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
ptstartover:
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
cur_generation = xsoftc.xpt_generation;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* first find our driver in the list of drivers */
Change the peripheral driver list from a linker set to module driven driver registration. This should allow things like da, sa, cd etc to be in seperate KLD's to the cam core and make them preloadable.
2001-02-07 07:05:59 +00:00
for (p_drv = periph_drivers; *p_drv != NULL; p_drv++)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (strcmp((*p_drv)->driver_name, name) == 0)
break;
if (*p_drv == NULL) {
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
mtx_unlock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
style(9)
2008-01-02 01:45:31 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +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;
} else if (--splbreaknum == 0) {
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
mtx_unlock(&xsoftc.xpt_topo_lock);
mtx_lock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
splbreaknum = 100;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if (cur_generation != xsoftc.xpt_generation)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
goto ptstartover;
}
}
/*
* 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;
Add a "dummy light" (actually two dummy lights) to catch people who don't have the passthrough device configured in their kernel. This will hopefully reduce the number of people complaining that they can't get {camcontrol, xmcd, tosha, cdrecord, etc.} to work. Reviewed by: gibbs
1998-10-12 21:54:13 +00:00
base_periph_found = 1;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
device = periph->path->device;
Mechanical change to use <sys/queue.h> macro API instead of fondling implementation details. Created with: sed(1) Reviewed by: md5(1)
2001-02-04 13:13:25 +00:00
for (i = 0, periph = SLIST_FIRST(&device->periphs);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
periph != NULL;
Mechanical change to use <sys/queue.h> macro API instead of fondling implementation details. Created with: sed(1) Reviewed by: md5(1)
2001-02-04 13:13:25 +00:00
periph = SLIST_NEXT(periph, periph_links), i++) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Check to see whether we have a
style(9)
2008-01-02 01:45:31 +00:00
* passthrough device or not.
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
*/
if (strcmp(periph->periph_name, "pass") == 0) {
/*
* Fill in the getdevlist fields.
*/
strcpy(ccb->cgdl.periph_name,
periph->periph_name);
ccb->cgdl.unit_number =
periph->unit_number;
Mechanical change to use <sys/queue.h> macro API instead of fondling implementation details. Created with: sed(1) Reviewed by: md5(1)
2001-02-04 13:13:25 +00:00
if (SLIST_NEXT(periph, periph_links))
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
Add a "dummy light" (actually two dummy lights) to catch people who don't have the passthrough device configured in their kernel. This will hopefully reduce the number of people complaining that they can't get {camcontrol, xmcd, tosha, cdrecord, etc.} to work. Reviewed by: gibbs
1998-10-12 21:54:13 +00:00
/*
* 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");
Correct the kernel config hint in a printf statement. Submitted by: R.Mahmatkhanov <cvs-src@yandex.ru> Reviewed by: scottl Approved by: rwatson (mentor) MFC after: 3 days
2008-01-28 18:13:03 +00:00
printf("xptioctl: put \"device pass\" in "
Add a "dummy light" (actually two dummy lights) to catch people who don't have the passthrough device configured in their kernel. This will hopefully reduce the number of people complaining that they can't get {camcontrol, xmcd, tosha, cdrecord, etc.} to work. Reviewed by: gibbs
1998-10-12 21:54:13 +00:00
"your kernel config file\n");
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
mtx_unlock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
}
default:
error = ENOTTY;
break;
}
return(error);
}
CAM, the module: scbus, da, cd, and st wrapped in one module. Make the umass driver depend on this module. Makes it possible to compile the kernel without SCSI support and load it when for example a USB floppy is conencted.
2000-08-13 18:49:40 +00:00
static int
cam_module_event_handler(module_t mod, int what, void *arg)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
int error;
switch (what) {
case MOD_LOAD:
if ((error = xpt_init(NULL)) != 0)
return (error);
break;
case MOD_UNLOAD:
CAM, the module: scbus, da, cd, and st wrapped in one module. Make the umass driver depend on this module. Makes it possible to compile the kernel without SCSI support and load it when for example a USB floppy is conencted.
2000-08-13 18:49:40 +00:00
return EBUSY;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
default:
Do a pass over all modules in the kernel and make them return EOPNOTSUPP for unknown events. A number of modules return EINVAL in this instance, and I have left those alone for now and instead taught MOD_QUIESCE to accept this as "didn't do anything".
2004-07-15 08:26:07 +00:00
return EOPNOTSUPP;
CAM, the module: scbus, da, cd, and st wrapped in one module. Make the umass driver depend on this module. Makes it possible to compile the kernel without SCSI support and load it when for example a USB floppy is conencted.
2000-08-13 18:49:40 +00:00
}
return 0;
}
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
/* thread to handle bus rescans */
static void
xpt_scanner_thread(void *dummy)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
cam_isrq_t queue;
union ccb *ccb;
struct cam_sim *sim;
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
for (;;) {
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
/*
* Wait for a rescan request to come in. When it does, splice
* it onto a queue from local storage so that the xpt lock
* doesn't need to be held while the requests are being
* processed.
*/
xpt_lock_buses();
msleep(&xsoftc.ccb_scanq, &xsoftc.xpt_topo_lock, PRIBIO,
"ccb_scanq", 0);
TAILQ_INIT(&queue);
TAILQ_CONCAT(&queue, &xsoftc.ccb_scanq, sim_links.tqe);
xpt_unlock_buses();
while ((ccb = (union ccb *)TAILQ_FIRST(&queue)) != NULL) {
TAILQ_REMOVE(&queue, &ccb->ccb_h, sim_links.tqe);
sim = ccb->ccb_h.path->bus->sim;
Inline cam_periph_lock|unlock to make debugging easier. Use more CAM_SIM_LOCK() more uniformly.
2007-04-19 22:46:26 +00:00
CAM_SIM_LOCK(sim);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
ccb->ccb_h.func_code = XPT_SCAN_BUS;
ccb->ccb_h.cbfcnp = xptdone;
xpt_setup_ccb(&ccb->ccb_h, ccb->ccb_h.path, 5);
cam_periph_runccb(ccb, NULL, 0, 0, NULL);
xpt_free_path(ccb->ccb_h.path);
xpt_free_ccb(ccb);
Inline cam_periph_lock|unlock to make debugging easier. Use more CAM_SIM_LOCK() more uniformly.
2007-04-19 22:46:26 +00:00
CAM_SIM_UNLOCK(sim);
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
}
}
}
void
xpt_rescan(union ccb *ccb)
{
struct ccb_hdr *hdr;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
/*
* Don't make duplicate entries for the same paths.
*/
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xpt_lock_buses();
TAILQ_FOREACH(hdr, &xsoftc.ccb_scanq, sim_links.tqe) {
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
if (xpt_path_comp(hdr->path, ccb->ccb_h.path) == 0) {
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xpt_unlock_buses();
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
xpt_print(ccb->ccb_h.path, "rescan already queued\n");
xpt_free_path(ccb->ccb_h.path);
xpt_free_ccb(ccb);
return;
}
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
TAILQ_INSERT_TAIL(&xsoftc.ccb_scanq, &ccb->ccb_h, sim_links.tqe);
wakeup(&xsoftc.ccb_scanq);
xpt_unlock_buses();
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Functions accessed by the peripheral drivers */
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
static int
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
xpt_init(void *dummy)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
struct cam_sim *xpt_sim;
struct cam_path *path;
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
struct cam_devq *devq;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
cam_status status;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
TAILQ_INIT(&xsoftc.xpt_busses);
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
TAILQ_INIT(&cam_simq);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
TAILQ_INIT(&xsoftc.ccb_scanq);
STAILQ_INIT(&xsoftc.highpowerq);
xsoftc.num_highpower = CAM_MAX_HIGHPOWER;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
mtx_init(&cam_simq_lock, "CAM SIMQ lock", NULL, MTX_DEF);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_init(&xsoftc.xpt_lock, "XPT lock", NULL, MTX_DEF);
mtx_init(&xsoftc.xpt_topo_lock, "XPT topology lock", NULL, MTX_DEF);
Provide locking for the ccb_bioq. This allows xpt_done() to be called without Giant held. In camisr(), move the ccb_bioq elements to a temporary local list and then process the elements off of that list. This enables the list to be processed by only taking the ccb_bioq_lock once and only for a very short time. ccb_bioq_lock is a leaf mutex, so it's fine to call xpt_done() with other locks held. This is just a very minor step in the work to lock CAM, but it allows us to avoid some messy locking/unlock dances in certain drivers.
2005-02-09 11:44:15 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* The xpt layer is, itself, the equivelent of a SIM.
* Allow 16 ccbs in the ccb pool for it. This should
* give decent parallelism when we probe busses and
* perform other XPT functions.
*/
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
devq = cam_simq_alloc(16);
xpt_sim = cam_sim_alloc(xptaction,
xptpoll,
"xpt",
/*softc*/NULL,
/*unit*/0,
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
/*mtx*/&xsoftc.xpt_lock,
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
/*max_dev_transactions*/0,
/*max_tagged_dev_transactions*/0,
devq);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if (xpt_sim == NULL)
return (ENOMEM);
xpt_sim->max_ccbs = 16;
mtx_lock(&xsoftc.xpt_lock);
Prepare for future integration between CAM and newbus. xpt_bus_register now takes a device_t to be the parent of the bus that is being created. Most SIMs have been updated with a reasonable argument, but a few exceptions just pass NULL for now. This argument isn't used yet and the newbus integration likely won't be ready until after 7.0-RELEASE.
2007-06-17 05:55:54 +00:00
if ((status = xpt_bus_register(xpt_sim, NULL, 0)) != CAM_SUCCESS) {
fix botch
2006-12-21 23:20:51 +00:00
printf("xpt_init: xpt_bus_register failed with status %#x,"
Pay attention to return value from xpt_bus_register in xpt_init. Obtained from: Xin Li (Coverity) MFC after: 3 days
2006-12-21 20:06:30 +00:00
" failing attach\n", status);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
return (EINVAL);
Pay attention to return value from xpt_bus_register in xpt_init. Obtained from: Xin Li (Coverity) MFC after: 3 days
2006-12-21 20:06:30 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Looking at the XPT from the SIM layer, the XPT is
* the equivelent 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);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
return (EINVAL);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
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,
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
path, NULL, 0, xpt_sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_free_path(path);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_unlock(&xsoftc.xpt_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Register a callback for when interrupts are enabled.
*/
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xsoftc.xpt_config_hook =
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
(struct intr_config_hook *)malloc(sizeof(struct intr_config_hook),
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
M_CAMXPT, M_NOWAIT | M_ZERO);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if (xsoftc.xpt_config_hook == NULL) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
printf("xpt_init: Cannot malloc config hook "
"- failing attach\n");
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
return (ENOMEM);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xsoftc.xpt_config_hook->ich_func = xpt_config;
if (config_intrhook_establish(xsoftc.xpt_config_hook) != 0) {
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free (xsoftc.xpt_config_hook, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
printf("xpt_init: config_intrhook_establish failed "
"- failing attach\n");
}
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
/* fire up rescan thread */
Rename the kthread_xxx (e.g. kthread_create()) calls to kproc_xxx as they actually make whole processes. Thos makes way for us to add REAL kthread_create() and friends that actually make theads. it turns out that most of these calls actually end up being moved back to the thread version when it's added. but we need to make this cosmetic change first. I'd LOVE to do this rename in 7.0 so that we can eventually MFC the new kthread_xxx() calls.
2007-10-20 23:23:23 +00:00
if (kproc_create(xpt_scanner_thread, NULL, NULL, 0, 0, "xpt_thrd")) {
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
printf("xpt_init: failed to create rescan thread\n");
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Install our software interrupt handlers */
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
swi_add(NULL, "cambio", camisr, NULL, SWI_CAMBIO, INTR_MPSAFE, &cambio_ih);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
return (0);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
static cam_status
xptregister(struct cam_periph *periph, void *arg)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
struct cam_sim *xpt_sim;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (periph == NULL) {
printf("xptregister: periph was NULL!!\n");
return(CAM_REQ_CMP_ERR);
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xpt_sim = (struct cam_sim *)arg;
xpt_sim->softc = periph;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_periph = periph;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
periph->softc = NULL;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(CAM_REQ_CMP);
}
int32_t
xpt_add_periph(struct cam_periph *periph)
{
struct cam_ed *device;
int32_t status;
struct periph_list *periph_head;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(periph->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
device = periph->path->device;
periph_head = &device->periphs;
status = CAM_REQ_CMP;
if (device != NULL) {
/*
* Make room for this peripheral
* so it will fit in the queue
* when it's scheduled to run
*/
status = camq_resize(&device->drvq,
device->drvq.array_size + 1);
device->generation++;
SLIST_INSERT_HEAD(periph_head, periph, periph_links);
}
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
xsoftc.xpt_generation++;
mtx_unlock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return (status);
}
void
xpt_remove_periph(struct cam_periph *periph)
{
struct cam_ed *device;
Revert r186186 for now; it breaks stuff. Approved by: rwatson (mentor)
2008-12-17 10:49:03 +00:00
mtx_assert(periph->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
device = periph->path->device;
if (device != NULL) {
struct periph_list *periph_head;
periph_head = &device->periphs;
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Release the slot for this peripheral */
camq_resize(&device->drvq, device->drvq.array_size - 1);
device->generation++;
Back out the previous change to the queue(3) interface. It was not discussed and should probably not happen. Requested by: msmith and others
2000-05-26 02:09:24 +00:00
SLIST_REMOVE(periph_head, periph, cam_periph, periph_links);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
xsoftc.xpt_generation++;
mtx_unlock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
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 ccb_pathinq cpi;
struct ccb_trans_settings cts;
struct cam_path *path;
u_int speed;
u_int freq;
u_int mb;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(periph->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
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->path;
/*
* To ensure that this is printed in one piece,
* mask out CAM interrupts.
*/
printf("%s%d at %s%d bus %d target %d lun %d\n",
periph->periph_name, periph->unit_number,
path->bus->sim->sim_name,
path->bus->sim->unit_number,
path->bus->sim->bus_id,
path->target->target_id,
path->device->lun_id);
printf("%s%d: ", periph->periph_name, periph->unit_number);
scsi_print_inquiry(&path->device->inq_data);
cleanup some of the XPORT_FC code
2001-06-24 18:19:24 +00:00
if (bootverbose && 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);
}
xpt_setup_ccb(&cts.ccb_h, path, /*priority*/1);
cts.ccb_h.func_code = XPT_GET_TRAN_SETTINGS;
cts.type = CTS_TYPE_CURRENT_SETTINGS;
xpt_action((union ccb*)&cts);
XPT_GET_TRAN_SETTINGS can fail- so check for an error and don't blindly assume it succeeded.
2006-11-07 23:06:00 +00:00
if ((cts.ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
return;
}
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
/* Ask the SIM for its base transfer speed */
xpt_setup_ccb(&cpi.ccb_h, path, /*priority*/1);
cpi.ccb_h.func_code = XPT_PATH_INQ;
xpt_action((union ccb *)&cpi);
speed = cpi.base_transfer_speed;
freq = 0;
cleanup some of the XPORT_FC code
2001-06-24 18:19:24 +00:00
if (cts.ccb_h.status == CAM_REQ_CMP && cts.transport == XPORT_SPI) {
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 ccb_trans_settings_spi *spi;
spi = &cts.xport_specific.spi;
if ((spi->valid & CTS_SPI_VALID_SYNC_OFFSET) != 0
&& spi->sync_offset != 0) {
freq = scsi_calc_syncsrate(spi->sync_period);
speed = freq;
}
if ((spi->valid & CTS_SPI_VALID_BUS_WIDTH) != 0)
speed *= (0x01 << spi->bus_width);
}
first blush at some FC path inquiry settings
2001-06-04 18:23:49 +00:00
cleanup some of the XPORT_FC code
2001-06-24 18:19:24 +00:00
if (cts.ccb_h.status == CAM_REQ_CMP && cts.transport == XPORT_FC) {
struct ccb_trans_settings_fc *fc = &cts.xport_specific.fc;
if (fc->valid & CTS_FC_VALID_SPEED) {
speed = fc->bitrate;
}
first blush at some FC path inquiry settings
2001-06-04 18:23:49 +00:00
}
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
CAM_NEW_TRAN_CODE placeholders for SAS support. Reviewed by: nate, ken, justin MFC after: 1 month
2006-01-20 19:21:38 +00:00
if (cts.ccb_h.status == CAM_REQ_CMP && cts.transport == XPORT_SAS) {
struct ccb_trans_settings_sas *sas = &cts.xport_specific.sas;
if (sas->valid & CTS_SAS_VALID_SPEED) {
speed = sas->bitrate;
}
}
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
mb = speed / 1000;
if (mb > 0)
printf("%s%d: %d.%03dMB/s transfers",
periph->periph_name, periph->unit_number,
mb, speed % 1000);
else
printf("%s%d: %dKB/s transfers", periph->periph_name,
periph->unit_number, speed);
/* Report additional information about SPI connections */
cleanup some of the XPORT_FC code
2001-06-24 18:19:24 +00:00
if (cts.ccb_h.status == CAM_REQ_CMP && cts.transport == XPORT_SPI) {
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 ccb_trans_settings_spi *spi;
spi = &cts.xport_specific.spi;
if (freq != 0) {
printf(" (%d.%03dMHz%s, offset %d", freq / 1000,
freq % 1000,
(spi->ppr_options & MSG_EXT_PPR_DT_REQ) != 0
? " DT" : "",
spi->sync_offset);
}
if ((spi->valid & CTS_SPI_VALID_BUS_WIDTH) != 0
&& spi->bus_width > 0) {
if (freq != 0) {
printf(", ");
} else {
printf(" (");
}
printf("%dbit)", 8 * (0x01 << spi->bus_width));
} else if (freq != 0) {
printf(")");
}
}
cleanup some of the XPORT_FC code
2001-06-24 18:19:24 +00:00
if (cts.ccb_h.status == CAM_REQ_CMP && cts.transport == XPORT_FC) {
first blush at some FC path inquiry settings
2001-06-04 18:23:49 +00:00
struct ccb_trans_settings_fc *fc;
fc = &cts.xport_specific.fc;
cleanup some of the XPORT_FC code
2001-06-24 18:19:24 +00:00
if (fc->valid & CTS_FC_VALID_WWNN)
printf(" WWNN 0x%llx", (long long) fc->wwnn);
if (fc->valid & CTS_FC_VALID_WWPN)
printf(" WWPN 0x%llx", (long long) fc->wwpn);
if (fc->valid & CTS_FC_VALID_PORT)
printf(" PortID 0x%x", fc->port);
first blush at some FC path inquiry settings
2001-06-04 18:23:49 +00:00
}
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) {
Make cam_xpt's pronouncements match camcontrol (Tagged -> Command) Queueing.
2006-12-10 05:37:48 +00:00
printf("\n%s%d: Command Queueing Enabled",
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);
}
printf("\n");
/*
* We only want to print the caller's announce string if they've
* passed one in..
*/
if (announce_string != NULL)
printf("%s%d: %s\n", periph->periph_name,
periph->unit_number, announce_string);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct cam_eb *bus)
{
dev_match_ret retval;
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;
/*
style(9)
2008-01-02 01:45:31 +00:00
* If we get to this point, the user definitely wants
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* 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,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct cam_ed *device)
{
dev_match_ret retval;
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.
*/
Compare the *number* of patterns to zero, not the *pointer* to the patterns. (These lines are correct the other two times they appear.) Reported by: "Ted Unangst" <tedu@coverity.com> Approved by: rwatson (mentor), ken (scsi)
2004-02-28 12:59:56 +00:00
if ((patterns == NULL) || (num_patterns == 0))
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(DM_RET_DESCEND | DM_RET_COPY);
for (i = 0; i < num_patterns; i++) {
struct device_match_pattern *cur_pattern;
/*
* 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;
/*
* If they want to match any device node, we give them any
* device node.
*/
if (cur_pattern->flags == DEV_MATCH_ANY) {
/* set the copy flag */
retval |= DM_RET_COPY;
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* 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 == 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,
(caddr_t)&cur_pattern->inq_pat,
1, sizeof(cur_pattern->inq_pat),
scsi_static_inquiry_match) == NULL))
continue;
/*
style(9)
2008-01-02 01:45:31 +00:00
* If we get to this point, the user definitely wants
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* 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,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct cam_periph *periph)
{
dev_match_ret retval;
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;
/*
style(9)
2008-01-02 01:45:31 +00:00
* If we get to this point, the user definitely wants
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* 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;
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));
style(9)
2008-01-02 01:45:31 +00:00
cdm->pos.position_type =
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
CAM_DEV_POS_EDT | CAM_DEV_POS_BUS;
cdm->pos.cookie.bus = bus;
cdm->pos.generations[CAM_BUS_GENERATION]=
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xsoftc.bus_generation;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
strncpy(cdm->matches[j].result.bus_result.dev_name,
bus->sim->sim_name, DEV_IDLEN);
}
/*
* If the user is only interested in busses, 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.
*/
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
&& (bus == cdm->pos.cookie.bus)
&& (cdm->pos.position_type & CAM_DEV_POS_TARGET)
&& (cdm->pos.generations[CAM_TARGET_GENERATION] != 0)
&& (cdm->pos.generations[CAM_TARGET_GENERATION] !=
bus->generation)) {
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return(0);
}
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))
return(xpttargettraverse(bus,
(struct cam_et *)cdm->pos.cookie.target,
xptedttargetfunc, arg));
else
return(xpttargettraverse(bus, NULL, xptedttargetfunc, arg));
}
static int
xptedttargetfunc(struct cam_et *target, void *arg)
{
struct ccb_dev_match *cdm;
cdm = (struct ccb_dev_match *)arg;
/*
* If there is a device list generation recorded, check it to
* make sure the device list hasn't changed.
*/
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
&& (cdm->pos.cookie.bus == target->bus)
&& (cdm->pos.position_type & CAM_DEV_POS_TARGET)
&& (cdm->pos.cookie.target == target)
&& (cdm->pos.position_type & CAM_DEV_POS_DEVICE)
&& (cdm->pos.generations[CAM_DEV_GENERATION] != 0)
&& (cdm->pos.generations[CAM_DEV_GENERATION] !=
target->generation)) {
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return(0);
}
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
&& (cdm->pos.cookie.bus == target->bus)
&& (cdm->pos.position_type & CAM_DEV_POS_TARGET)
&& (cdm->pos.cookie.target == target)
&& (cdm->pos.position_type & CAM_DEV_POS_DEVICE)
&& (cdm->pos.cookie.device != NULL))
return(xptdevicetraverse(target,
(struct cam_ed *)cdm->pos.cookie.device,
xptedtdevicefunc, arg));
else
return(xptdevicetraverse(target, NULL, xptedtdevicefunc, arg));
}
static int
xptedtdevicefunc(struct cam_ed *device, void *arg)
{
struct ccb_dev_match *cdm;
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_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));
style(9)
2008-01-02 01:45:31 +00:00
cdm->pos.position_type =
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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]=
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xsoftc.bus_generation;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
cdm->pos.cookie.target = device->target;
cdm->pos.generations[CAM_TARGET_GENERATION] =
device->target->bus->generation;
cdm->pos.cookie.device = device;
style(9)
2008-01-02 01:45:31 +00:00
cdm->pos.generations[CAM_DEV_GENERATION] =
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
bcopy(&device->inq_data,
&cdm->matches[j].result.device_result.inq_data,
sizeof(struct scsi_inquiry_data));
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;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
/*
* 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.
*/
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
&& (device->target->bus == cdm->pos.cookie.bus)
&& (cdm->pos.position_type & CAM_DEV_POS_TARGET)
&& (device->target == cdm->pos.cookie.target)
&& (cdm->pos.position_type & CAM_DEV_POS_DEVICE)
&& (device == cdm->pos.cookie.device)
&& (cdm->pos.position_type & CAM_DEV_POS_PERIPH)
&& (cdm->pos.generations[CAM_PERIPH_GENERATION] != 0)
&& (cdm->pos.generations[CAM_PERIPH_GENERATION] !=
device->generation)){
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return(0);
}
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
&& (cdm->pos.cookie.bus == device->target->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)
&& (cdm->pos.cookie.periph != NULL))
return(xptperiphtraverse(device,
(struct cam_periph *)cdm->pos.cookie.periph,
xptedtperiphfunc, arg));
else
return(xptperiphtraverse(device, NULL, 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;
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));
style(9)
2008-01-02 01:45:31 +00:00
cdm->pos.position_type =
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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]=
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xsoftc.bus_generation;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
cdm->pos.cookie.target = periph->path->target;
cdm->pos.generations[CAM_TARGET_GENERATION] =
periph->path->bus->generation;
cdm->pos.cookie.device = periph->path->device;
style(9)
2008-01-02 01:45:31 +00:00
cdm->pos.generations[CAM_DEV_GENERATION] =
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
strncpy(cdm->matches[j].result.periph_result.periph_name,
periph->periph_name, DEV_IDLEN);
}
return(1);
}
static int
xptedtmatch(struct ccb_dev_match *cdm)
{
int ret;
cdm->num_matches = 0;
/*
* Check the bus list generation. If it has changed, the user
* needs to reset everything and start over.
*/
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
&& (cdm->pos.generations[CAM_BUS_GENERATION] != 0)
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
&& (cdm->pos.generations[CAM_BUS_GENERATION] != xsoftc.bus_generation)) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return(0);
}
if ((cdm->pos.position_type & CAM_DEV_POS_BUS)
&& (cdm->pos.cookie.bus != NULL))
ret = xptbustraverse((struct cam_eb *)cdm->pos.cookie.bus,
xptedtbusfunc, cdm);
else
ret = xptbustraverse(NULL, 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)
{
struct ccb_dev_match *cdm;
cdm = (struct ccb_dev_match *)arg;
if ((cdm->pos.position_type & CAM_DEV_POS_PDPTR)
&& (cdm->pos.cookie.pdrv == pdrv)
&& (cdm->pos.position_type & CAM_DEV_POS_PERIPH)
&& (cdm->pos.generations[CAM_PERIPH_GENERATION] != 0)
&& (cdm->pos.generations[CAM_PERIPH_GENERATION] !=
(*pdrv)->generation)) {
cdm->status = CAM_DEV_MATCH_LIST_CHANGED;
return(0);
}
if ((cdm->pos.position_type & CAM_DEV_POS_PDPTR)
&& (cdm->pos.cookie.pdrv == pdrv)
&& (cdm->pos.position_type & CAM_DEV_POS_PERIPH)
&& (cdm->pos.cookie.periph != NULL))
return(xptpdperiphtraverse(pdrv,
(struct cam_periph *)cdm->pos.cookie.periph,
xptplistperiphfunc, arg));
else
return(xptpdperiphtraverse(pdrv, NULL,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;
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));
style(9)
2008-01-02 01:45:31 +00:00
cdm->pos.position_type =
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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.
*/
Change the peripheral driver list from a linker set to module driven driver registration. This should allow things like da, sa, cd etc to be in seperate KLD's to the cam core and make them preloadable.
2001-02-07 07:05:59 +00:00
for (pdrv = periph_drivers; *pdrv != NULL; pdrv++) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (strcmp((*pdrv)->driver_name,
periph->periph_name) == 0)
break;
}
Check the correct periph pointer after search for it. Submitted by: Coverity Prevent analysis tool
2005-02-20 23:51:20 +00:00
if (*pdrv == NULL) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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 = -1;
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 = -1;
cdm->matches[j].result.periph_result.unit_number =
periph->unit_number;
strncpy(cdm->matches[j].result.periph_result.periph_name,
periph->periph_name, DEV_IDLEN);
}
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 busses 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(
cam_periph.c: Move handling of CAM_AUTOSENSE_FAIL into block dealing with all other scsi status errors. cam_queue.c: cam_queue.h: Fix 'off by one' heap bug in a more efficient manner. Since heap algorithms like to deal with indexes started from 1, offset our heap array pointer at allocation time to make this so for a C environment. This makes the implementation of the algorithm a bit more efficient. cam_xpt.c: Use macros for accessing the head of the heap so that code is isolated from implementation details of the heap.
1999-04-19 21:26:08 +00:00
(struct periph_driver **)cdm->pos.cookie.pdrv,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
for (bus = (start_bus ? start_bus : TAILQ_FIRST(&xsoftc.xpt_busses));
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
bus != NULL;
bus = next_bus) {
next_bus = TAILQ_NEXT(bus, links);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_unlock(&xsoftc.xpt_topo_lock);
Inline cam_periph_lock|unlock to make debugging easier. Use more CAM_SIM_LOCK() more uniformly.
2007-04-19 22:46:26 +00:00
CAM_SIM_LOCK(bus->sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
retval = tr_func(bus, arg);
Inline cam_periph_lock|unlock to make debugging easier. Use more CAM_SIM_LOCK() more uniformly.
2007-04-19 22:46:26 +00:00
CAM_SIM_UNLOCK(bus->sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (retval == 0)
return(retval);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_unlock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(retval);
}
Make "kldunload atapicam" return EBUSY instead of deadlocking when a device created by atapicam is being kept opened or mounted. This is probably just a temporary solution until we invent something better. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation Reported by: Jaakko Heinonen
2009-01-08 17:26:51 +00:00
int
xpt_sim_opened(struct cam_sim *sim)
{
struct cam_eb *bus;
struct cam_et *target;
struct cam_ed *device;
struct cam_periph *periph;
KASSERT(sim->refcount >= 1, ("sim->refcount >= 1"));
mtx_assert(sim->mtx, MA_OWNED);
mtx_lock(&xsoftc.xpt_topo_lock);
TAILQ_FOREACH(bus, &xsoftc.xpt_busses, links) {
if (bus->sim != sim)
continue;
TAILQ_FOREACH(target, &bus->et_entries, links) {
TAILQ_FOREACH(device, &target->ed_entries, links) {
SLIST_FOREACH(periph, &device->periphs,
periph_links) {
if (periph->refcount > 0) {
mtx_unlock(&xsoftc.xpt_topo_lock);
return (1);
}
}
}
}
}
mtx_unlock(&xsoftc.xpt_topo_lock);
return (0);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
for (target = (start_target ? start_target :
TAILQ_FIRST(&bus->et_entries));
target != NULL; target = next_target) {
next_target = TAILQ_NEXT(target, links);
retval = tr_func(target, arg);
if (retval == 0)
return(retval);
}
return(retval);
}
static int
xptdevicetraverse(struct cam_et *target, struct cam_ed *start_device,
xpt_devicefunc_t *tr_func, void *arg)
{
struct cam_ed *device, *next_device;
int retval;
retval = 1;
for (device = (start_device ? start_device :
TAILQ_FIRST(&target->ed_entries));
device != NULL;
device = next_device) {
next_device = TAILQ_NEXT(device, links);
retval = tr_func(device, arg);
if (retval == 0)
return(retval);
}
return(retval);
}
static int
xptperiphtraverse(struct cam_ed *device, struct cam_periph *start_periph,
xpt_periphfunc_t *tr_func, void *arg)
{
struct cam_periph *periph, *next_periph;
int retval;
retval = 1;
for (periph = (start_periph ? start_periph :
SLIST_FIRST(&device->periphs));
periph != NULL;
periph = next_periph) {
next_periph = SLIST_NEXT(periph, periph_links);
retval = tr_func(periph, arg);
if (retval == 0)
return(retval);
}
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.
*/
Change the peripheral driver list from a linker set to module driven driver registration. This should allow things like da, sa, cd etc to be in seperate KLD's to the cam core and make them preloadable.
2001-02-07 07:05:59 +00:00
for (pdrv = (start_pdrv ? start_pdrv : periph_drivers);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
*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;
for (periph = (start_periph ? start_periph :
TAILQ_FIRST(&(*pdrv)->units)); periph != NULL;
periph = next_periph) {
next_periph = TAILQ_NEXT(periph, unit_links);
retval = tr_func(periph, arg);
if (retval == 0)
return(retval);
}
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 async_node *cur_entry;
cur_entry = (struct async_node *)arg;
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
/*
* 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);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_compile_path(&path,
NULL,
device->target->bus->path_id,
device->target->target_id,
device->lun_id);
xpt_setup_ccb(&cgd.ccb_h, &path, /*priority*/1);
cgd.ccb_h.func_code = XPT_GDEV_TYPE;
xpt_action((union ccb *)&cgd);
cur_entry->callback(cur_entry->callback_arg,
AC_FOUND_DEVICE,
&path, &cgd);
xpt_release_path(&path);
return(1);
}
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static int
xptsetasyncbusfunc(struct cam_eb *bus, void *arg)
{
struct cam_path path;
struct ccb_pathinq cpi;
struct async_node *cur_entry;
cur_entry = (struct async_node *)arg;
xpt_compile_path(&path, /*periph*/NULL,
bus->sim->path_id,
CAM_TARGET_WILDCARD,
CAM_LUN_WILDCARD);
xpt_setup_ccb(&cpi.ccb_h, &path, /*priority*/1);
cpi.ccb_h.func_code = XPT_PATH_INQ;
xpt_action((union ccb *)&cpi);
cur_entry->callback(cur_entry->callback_arg,
AC_PATH_REGISTERED,
&path, &cpi);
xpt_release_path(&path);
return(1);
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
static void
xpt_action_sasync_cb(void *context, int pending)
{
struct async_node *cur_entry;
For the XPT_SASYNC_CB operation, only decouple the broadcast to the bus and device lists instead of decoupling the whole operation. This avoids problems with SIMs going away.
2007-04-16 19:55:36 +00:00
struct xpt_task *task;
uint32_t added;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
task = (struct xpt_task *)context;
For the XPT_SASYNC_CB operation, only decouple the broadcast to the bus and device lists instead of decoupling the whole operation. This avoids problems with SIMs going away.
2007-04-16 19:55:36 +00:00
cur_entry = (struct async_node *)task->data1;
added = task->data2;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if ((added & AC_FOUND_DEVICE) != 0) {
/*
* Get this peripheral up to date with all
* the currently existing devices.
*/
xpt_for_all_devices(xptsetasyncfunc, cur_entry);
}
if ((added & AC_PATH_REGISTERED) != 0) {
/*
* Get this peripheral up to date with all
* the currently existing busses.
*/
xpt_for_all_busses(xptsetasyncbusfunc, cur_entry);
Fix whitespace.
2008-08-01 15:13:56 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
free(task, M_CAMXPT);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
void
xpt_action(union ccb *start_ccb)
{
Run xpt_action at splsoftcam() so that it cannot be re-entered by our SWI. The XPT doesn't have a problem with this itself, but some controllers drivers may have been caught off guard by the old behavior. XPT_CONT_TARGET_IO is also a valid ccb type for cam_periph_unmapmem.
1999-05-25 20:17:29 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
CAM_DEBUG(start_ccb->ccb_h.path, CAM_DEBUG_TRACE, ("xpt_action\n"));
start_ccb->ccb_h.status = CAM_REQ_INPROG;
switch (start_ccb->ccb_h.func_code) {
case XPT_SCSI_IO:
Add a new CAM debugging mode, CAM_DEBUG_CDB. This causes the kernel to print out a one line description/dump of every SCSI CDB sent to a particular debugging target or targets. This is a good bit more useful than the other debugging modes, I think. Change some things in LINT to note the availability of this new option. Fix an erroneous argument to scsi_cdb_string() in scsi_all.c Reviewed by: gibbs
1998-10-02 21:00:58 +00:00
{
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;
Add a new CAM debugging mode, CAM_DEBUG_CDB. This causes the kernel to print out a one line description/dump of every SCSI CDB sent to a particular debugging target or targets. This is a good bit more useful than the other debugging modes, I think. Change some things in LINT to note the availability of this new option. Fix an erroneous argument to scsi_cdb_string() in scsi_all.c Reviewed by: gibbs
1998-10-02 21:00:58 +00:00
#ifdef CAMDEBUG
char cdb_str[(SCSI_MAX_CDBLEN * 3) + 1];
struct cam_path *path;
path = start_ccb->ccb_h.path;
#endif
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* 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.
*/
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
device = start_ccb->ccb_h.path->device;
if (device->protocol_version <= SCSI_REV_2
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
&& 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;
Add a new CAM debugging mode, CAM_DEBUG_CDB. This causes the kernel to print out a one line description/dump of every SCSI CDB sent to a particular debugging target or targets. This is a good bit more useful than the other debugging modes, I think. Change some things in LINT to note the availability of this new option. Fix an erroneous argument to scsi_cdb_string() in scsi_all.c Reviewed by: gibbs
1998-10-02 21:00:58 +00:00
CAM_DEBUG(path, CAM_DEBUG_CDB,("%s. CDB: %s\n",
scsi_op_desc(start_ccb->csio.cdb_io.cdb_bytes[0],
&path->device->inq_data),
scsi_cdb_string(start_ccb->csio.cdb_io.cdb_bytes,
Fix several potential buffer overrun conditions. These changes have been tested both in the kernel and in userland. Also, fix a couple of printf warnings that show up when CAMDEBUG is defined. Reviewed by: imp Partially submitted by: imp
1998-10-15 19:08:58 +00:00
cdb_str, sizeof(cdb_str))));
Add a new CAM debugging mode, CAM_DEBUG_CDB. This causes the kernel to print out a one line description/dump of every SCSI CDB sent to a particular debugging target or targets. This is a good bit more useful than the other debugging modes, I think. Change some things in LINT to note the availability of this new option. Fix an erroneous argument to scsi_cdb_string() in scsi_all.c Reviewed by: gibbs
1998-10-02 21:00:58 +00:00
}
Add /* FALLTHROUGH */ Move /* FALLTHROUGH */ to correct location. Remove unused variable(s). Found by: FlexeLint
2003-05-31 20:46:21 +00:00
/* FALLTHROUGH */
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
case XPT_TARGET_IO:
case XPT_CONT_TARGET_IO:
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
start_ccb->csio.sense_resid = 0;
start_ccb->csio.resid = 0;
/* FALLTHROUGH */
Fix Bus Device Resets from userland, and specifically from camcontrol. camcontrol was setup to use the old scheme of going through the xpt(4) device, which never worked properly (and has been disabled for a while). camcontrol now sends BDRs through the pass(4) device, and XPT_RESET_DEV CCBs are put on the device queue in the transport layer, as they should be. Submitted by: luoqi Reviewed by: ken
1999-10-31 02:39:44 +00:00
case XPT_RESET_DEV:
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
case XPT_ENG_EXEC:
{
struct cam_path *path;
int runq;
path = start_ccb->ccb_h.path;
cam_ccbq_insert_ccb(&path->device->ccbq, start_ccb);
if (path->device->qfrozen_cnt == 0)
runq = xpt_schedule_dev_sendq(path->bus, path->device);
else
runq = 0;
if (runq != 0)
xpt_run_dev_sendq(path->bus);
break;
}
case XPT_SET_TRAN_SETTINGS:
{
xpt_set_transfer_settings(&start_ccb->cts,
Correct problems with xpt_set_transfer_settings and async transfer negotiation changes with wildcarded paths.
1998-09-25 22:35:56 +00:00
start_ccb->ccb_h.path->device,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*async_update*/FALSE);
break;
}
case XPT_CALC_GEOMETRY:
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
{
struct cam_sim *sim;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* 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;
}
#ifdef PC98
/*
* In a PC-98 system, geometry translation depens on
* the "real" device geometry obtained from mode page 4.
* SCSI geometry translation is performed in the
* initialization routine of the SCSI BIOS and the result
* stored in host memory. If the translation is available
* in host memory, use it. If not, rely on the default
* translation the device driver performs.
*/
if (scsi_da_bios_params(&start_ccb->ccg) != 0) {
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
#endif
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
sim = start_ccb->ccb_h.path->bus->sim;
(*(sim->sim_action))(sim, start_ccb);
break;
}
Wire up the XPT_ABORT and XPT_RESET_DEV ccb function codes so they can be delivered to controller drivers. Adjust for changes to the ccb_hdr list types in cam_queue.h
1998-12-15 08:13:10 +00:00
case XPT_ABORT:
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
{
union ccb* abort_ccb;
abort_ccb = start_ccb->cab.abort_ccb;
if (XPT_FC_IS_DEV_QUEUED(abort_ccb)) {
if (abort_ccb->ccb_h.pinfo.index >= 0) {
struct cam_ccbq *ccbq;
ccbq = &abort_ccb->ccb_h.path->device->ccbq;
cam_ccbq_remove_ccb(ccbq, abort_ccb);
abort_ccb->ccb_h.status =
CAM_REQ_ABORTED|CAM_DEV_QFRZN;
xpt_freeze_devq(abort_ccb->ccb_h.path, 1);
xpt_done(abort_ccb);
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
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;
}
style(9)
2008-01-02 01:45:31 +00:00
}
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +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.
*/
}
Add /* FALLTHROUGH */ Move /* FALLTHROUGH */ to correct location. Remove unused variable(s). Found by: FlexeLint
2003-05-31 20:46:21 +00:00
/* FALLTHROUGH */
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
case XPT_ACCEPT_TARGET_IO:
case XPT_EN_LUN:
case XPT_IMMED_NOTIFY:
case XPT_NOTIFY_ACK:
case XPT_GET_TRAN_SETTINGS:
case XPT_RESET_BUS:
{
struct cam_sim *sim;
sim = start_ccb->ccb_h.path->bus->sim;
(*(sim->sim_action))(sim, start_ccb);
break;
}
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
case XPT_PATH_INQ:
{
struct cam_sim *sim;
sim = start_ccb->ccb_h.path->bus->sim;
(*(sim->sim_action))(sim, start_ccb);
break;
}
case XPT_PATH_STATS:
start_ccb->cpis.last_reset =
start_ccb->ccb_h.path->bus->last_reset;
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
case XPT_GDEV_TYPE:
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
{
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
struct cam_ed *dev;
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
dev = start_ccb->ccb_h.path->device;
if ((dev->flags & CAM_DEV_UNCONFIGURED) != 0) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
start_ccb->ccb_h.status = CAM_DEV_NOT_THERE;
} else {
struct ccb_getdev *cgd;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
struct cam_eb *bus;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct cam_et *tar;
cgd = &start_ccb->cgd;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
bus = cgd->ccb_h.path->bus;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
tar = cgd->ccb_h.path->target;
cgd->inq_data = dev->inq_data;
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);
}
style(9)
2008-01-02 01:45:31 +00:00
break;
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
}
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
case XPT_GDEV_STATS:
{
struct cam_ed *dev;
dev = start_ccb->ccb_h.path->device;
if ((dev->flags & CAM_DEV_UNCONFIGURED) != 0) {
start_ccb->ccb_h.status = CAM_DEV_NOT_THERE;
} else {
struct ccb_getdevstats *cgds;
struct cam_eb *bus;
struct cam_et *tar;
cgds = &start_ccb->cgds;
bus = cgds->ccb_h.path->bus;
tar = cgds->ccb_h.path->target;
cgds->dev_openings = dev->ccbq.dev_openings;
cgds->dev_active = dev->ccbq.dev_active;
cgds->devq_openings = dev->ccbq.devq_openings;
cgds->devq_queued = dev->ccbq.queue.entries;
cgds->held = dev->ccbq.held;
cgds->last_reset = tar->last_reset;
Rely on the statistics in XPT_GDEV_STATS instead of the versions still retained in XPT_GDEV_TYPE for binary compatibility. Mark the legacy structure values for removal when we bump the major CAM revision.
1999-05-23 18:57:29 +00:00
cgds->maxtags = dev->quirk->maxtags;
cgds->mintags = dev->quirk->mintags;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
if (timevalcmp(&tar->last_reset, &bus->last_reset, <))
cgds->last_reset = bus->last_reset;
cgds->ccb_h.status = CAM_REQ_CMP;
}
break;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct cam_ed *device;
int found;
found = 0;
/*
* Don't want anyone mucking with our data.
*/
device = start_ccb->ccb_h.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
style(9)
2008-01-02 01:45:31 +00:00
* list has changed, and therefore they need to start over
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* from the beginning.
*/
style(9)
2008-01-02 01:45:31 +00:00
if ((cgdl->index != 0) &&
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
(cgdl->generation != device->generation)) {
cgdl->status = CAM_GDEVLIST_LIST_CHANGED;
break;
}
/*
style(9)
2008-01-02 01:45:31 +00:00
* Traverse the list of peripherals and attempt to find
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* the requested peripheral.
*/
Mechanical change to use <sys/queue.h> macro API instead of fondling implementation details. Created with: sed(1) Reviewed by: md5(1)
2001-02-04 13:13:25 +00:00
for (nperiph = SLIST_FIRST(periph_head), i = 0;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
(nperiph != NULL) && (i <= cgdl->index);
Mechanical change to use <sys/queue.h> macro API instead of fondling implementation details. Created with: sed(1) Reviewed by: md5(1)
2001-02-04 13:13:25 +00:00
nperiph = SLIST_NEXT(nperiph, periph_links), i++) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (i == cgdl->index) {
strncpy(cgdl->periph_name,
nperiph->periph_name,
DEV_IDLEN);
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 busses, 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;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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:
Add /* FALLTHROUGH */ Move /* FALLTHROUGH */ to correct location. Remove unused variable(s). Found by: FlexeLint
2003-05-31 20:46:21 +00:00
xptedtmatch(cdm);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
case CAM_DEV_POS_PDRV:
Add /* FALLTHROUGH */ Move /* FALLTHROUGH */ to correct location. Remove unused variable(s). Found by: FlexeLint
2003-05-31 20:46:21 +00:00
xptperiphlistmatch(cdm);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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:
{
For the XPT_SASYNC_CB operation, only decouple the broadcast to the bus and device lists instead of decoupling the whole operation. This avoids problems with SIMs going away.
2007-04-16 19:55:36 +00:00
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 = &csa->ccb_h.path->device->asyncs;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
For the XPT_SASYNC_CB operation, only decouple the broadcast to the bus and device lists instead of decoupling the whole operation. This avoids problems with SIMs going away.
2007-04-16 19:55:36 +00:00
* If there is already an entry for us, simply
* update it.
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
*/
For the XPT_SASYNC_CB operation, only decouple the broadcast to the bus and device lists instead of decoupling the whole operation. This avoids problems with SIMs going away.
2007-04-16 19:55:36 +00:00
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);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
For the XPT_SASYNC_CB operation, only decouple the broadcast to the bus and device lists instead of decoupling the whole operation. This avoids problems with SIMs going away.
2007-04-16 19:55:36 +00:00
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);
csa->ccb_h.path->device->refcount--;
free(cur_entry, M_CAMXPT);
} else {
cur_entry->event_enable = csa->event_enable;
}
} 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->callback_arg = csa->callback_arg;
cur_entry->callback = csa->callback;
SLIST_INSERT_HEAD(async_head, cur_entry, links);
csa->ccb_h.path->device->refcount++;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
For the XPT_SASYNC_CB operation, only decouple the broadcast to the bus and device lists instead of decoupling the whole operation. This avoids problems with SIMs going away.
2007-04-16 19:55:36 +00:00
/*
* Need to decouple this operation via a taqskqueue so that
* the locking doesn't become a mess.
*/
if ((added & (AC_FOUND_DEVICE | AC_PATH_REGISTERED)) != 0) {
struct xpt_task *task;
task = malloc(sizeof(struct xpt_task), M_CAMXPT,
M_NOWAIT);
if (task == NULL) {
csa->ccb_h.status = CAM_RESRC_UNAVAIL;
break;
}
TASK_INIT(&task->task, 0, xpt_action_sasync_cb, task);
task->data1 = cur_entry;
task->data2 = added;
taskqueue_enqueue(taskqueue_thread, &task->task);
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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 = crs->ccb_h.path->device;
if (dev == NULL) {
crs->ccb_h.status = CAM_DEV_NOT_THERE;
break;
}
if ((crs->release_flags & RELSIM_ADJUST_OPENINGS) != 0) {
Handle some of the inquiry flags that have come into usage as of SPC2r20. Specifically, handle the BQueue flag which will indicate that a device supports the Basic Queueing model (no Head of Queue or Ordered tags). When this flag is set, SID_CmdQueue is clear. This has causes FreeBSD to assume that the device did not support tagged operations. MFC after: 1 month
2006-05-30 22:44:00 +00:00
if (INQ_DATA_TQ_ENABLED(&dev->inq_data)) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Don't ever go below one opening */
if (crs->openings > 0) {
xpt_dev_ccbq_resize(crs->ccb_h.path,
crs->openings);
People who wanted the "tagged opening" message can boot -v. The rest of us, who appear to have complained incessantly about this message since it appeared, can now stop.
1999-03-11 10:48:02 +00:00
if (bootverbose) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(crs->ccb_h.path,
"tagged openings now %d\n",
crs->openings);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
}
}
}
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;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
callout_stop(&dev->callout);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} else {
start_ccb->ccb_h.flags |= CAM_DEV_QFREEZE;
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
callout_reset(&dev->callout,
(crs->release_timeout * hz) / 1000,
xpt_release_devq_timeout, dev);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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 {
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +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 {
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
dev->flags |= CAM_DEV_REL_ON_QUEUE_EMPTY;
start_ccb->ccb_h.flags |= CAM_DEV_QFREEZE;
}
}
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if ((start_ccb->ccb_h.flags & CAM_DEV_QFREEZE) == 0) {
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(crs->ccb_h.path, /*count*/1,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*run_queue*/TRUE);
}
start_ccb->crs.qfrozen_cnt = dev->qfrozen_cnt;
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
}
case XPT_SCAN_BUS:
xpt_scan_bus(start_ccb->ccb_h.path->periph, start_ccb);
break;
case XPT_SCAN_LUN:
xpt_scan_lun(start_ccb->ccb_h.path->periph,
start_ccb->ccb_h.path, start_ccb->crcn.flags,
start_ccb);
break;
case XPT_DEBUG: {
#ifdef CAMDEBUG
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
#ifdef CAM_DEBUG_DELAY
cam_debug_delay = CAM_DEBUG_DELAY;
#endif
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
cam_dflags = start_ccb->cdbg.flags;
if (cam_dpath != NULL) {
xpt_free_path(cam_dpath);
cam_dpath = NULL;
}
if (cam_dflags != CAM_DEBUG_NONE) {
if (xpt_create_path(&cam_dpath, xpt_periph,
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;
cam_dflags = CAM_DEBUG_NONE;
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
} else {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
start_ccb->ccb_h.status = CAM_REQ_CMP;
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(cam_dpath, "debugging flags now %x\n",
cam_dflags);
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} else {
cam_dpath = NULL;
start_ccb->ccb_h.status = CAM_REQ_CMP;
}
#else /* !CAMDEBUG */
start_ccb->ccb_h.status = CAM_FUNC_NOTAVAIL;
#endif /* CAMDEBUG */
break;
}
case XPT_NOOP:
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
if ((start_ccb->ccb_h.flags & CAM_DEV_QFREEZE) != 0)
xpt_freeze_devq(start_ccb->ccb_h.path, 1);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
start_ccb->ccb_h.status = CAM_REQ_CMP;
break;
default:
case XPT_SDEV_TYPE:
case XPT_TERM_IO:
case XPT_ENG_INQ:
/* XXX Implement */
start_ccb->ccb_h.status = CAM_PROVIDE_FAIL;
break;
}
}
void
xpt_polled_action(union ccb *start_ccb)
{
u_int32_t timeout;
style(9)
2008-01-02 01:45:31 +00:00
struct cam_sim *sim;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct cam_devq *devq;
struct cam_ed *dev;
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
timeout = start_ccb->ccb_h.timeout;
sim = start_ccb->ccb_h.path->bus->sim;
devq = sim->devq;
dev = start_ccb->ccb_h.path->device;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(sim->mtx, MA_OWNED);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Steal an opening so that no other queued requests
* can get it before us while we simulate interrupts.
*/
dev->ccbq.devq_openings--;
style(9)
2008-01-02 01:45:31 +00:00
dev->ccbq.dev_openings--;
When deregistering a bus, attempt to flush out all outstanding operations before returning. Point the bus at a dummy cam_sim structure so that any CCBs will complete immediately with a CAM_DEV_NOT_THERE status, and ensure that any xpt_schedule() calls on the bus's devices will immediately call the peripheral's periph_start() routine. Also repeat the async messages because devices that were part of the way through being probed may appear after the original AC_LOST_DEVICE was sent, and would otherwise never go away. These changes make it possible to deregister a bus and free the SIM at most stages during bus probing without the usual crashes in camisr(). In particular, plugging in a umass device and then unplugging it as soon as the first probe messages appeared would almost always result in a crash. Now the device just goes away with a few CAM errors and all references to the CAM bus, target and device are dropped correctly.
2006-01-11 02:06:08 +00:00
while(((devq != NULL && devq->send_openings <= 0) ||
dev->ccbq.dev_openings < 0) && (--timeout > 0)) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
DELAY(1000);
(*(sim->sim_poll))(sim);
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
camisr_runqueue(&sim->sim_doneq);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
dev->ccbq.devq_openings++;
dev->ccbq.dev_openings++;
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (timeout != 0) {
xpt_action(start_ccb);
while(--timeout > 0) {
(*(sim->sim_poll))(sim);
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
camisr_runqueue(&sim->sim_doneq);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if ((start_ccb->ccb_h.status & CAM_STATUS_MASK)
!= CAM_REQ_INPROG)
break;
DELAY(1000);
}
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;
}
} else {
start_ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
}
}
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Schedule a peripheral driver to receive a ccb when it's
* target device has space for more transactions.
*/
void
xpt_schedule(struct cam_periph *perph, u_int32_t new_priority)
{
struct cam_ed *device;
int runq;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(perph->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
CAM_DEBUG(perph->path, CAM_DEBUG_TRACE, ("xpt_schedule\n"));
device = perph->path->device;
if (periph_is_queued(perph)) {
/* Simply reorder based on new priority */
CAM_DEBUG(perph->path, CAM_DEBUG_SUBTRACE,
(" change priority to %d\n", new_priority));
if (new_priority < perph->pinfo.priority) {
camq_change_priority(&device->drvq,
perph->pinfo.index,
new_priority);
}
runq = 0;
} else {
/* New entry on the queue */
CAM_DEBUG(perph->path, CAM_DEBUG_SUBTRACE,
(" added periph to queue\n"));
perph->pinfo.priority = new_priority;
Remove camq_regen(). We already perform modular comparisons for generation counts, so no further steps to deal with generation count wrap are required. Fix an off by one problem in the camq heap code.
1999-04-07 22:57:48 +00:00
perph->pinfo.generation = ++device->drvq.generation;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
camq_insert(&device->drvq, &perph->pinfo);
runq = xpt_schedule_dev_allocq(perph->path->bus, device);
}
if (runq != 0) {
CAM_DEBUG(perph->path, CAM_DEBUG_SUBTRACE,
(" calling xpt_run_devq\n"));
xpt_run_dev_allocq(perph->path->bus);
}
}
/*
* 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
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
* to run the queue.
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +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;
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT, ("xpt_schedule_dev\n"));
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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);
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
("changed priority to %d\n",
new_priority));
}
retval = 0;
} else {
/* New entry on the queue */
if (new_priority < old_priority)
pinfo->priority = new_priority;
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
("Inserting onto queue\n"));
Remove camq_regen(). We already perform modular comparisons for generation counts, so no further steps to deal with generation count wrap are required. Fix an off by one problem in the camq heap code.
1999-04-07 22:57:48 +00:00
pinfo->generation = ++queue->generation;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
camq_insert(queue, pinfo);
retval = 1;
}
return (retval);
}
static void
xpt_run_dev_allocq(struct cam_eb *bus)
{
struct cam_devq *devq;
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT, ("xpt_run_dev_allocq\n"));
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
devq = bus->sim->devq;
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
(" qfrozen_cnt == 0x%x, entries == %d, "
"openings == %d, active == %d\n",
devq->alloc_queue.qfrozen_cnt,
devq->alloc_queue.entries,
devq->alloc_openings,
devq->alloc_active));
devq->alloc_queue.qfrozen_cnt++;
while ((devq->alloc_queue.entries > 0)
&& (devq->alloc_openings > 0)
&& (devq->alloc_queue.qfrozen_cnt <= 1)) {
struct cam_ed_qinfo *qinfo;
struct cam_ed *device;
union ccb *work_ccb;
struct cam_periph *drv;
struct camq *drvq;
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
qinfo = (struct cam_ed_qinfo *)camq_remove(&devq->alloc_queue,
cam_periph.c: Move handling of CAM_AUTOSENSE_FAIL into block dealing with all other scsi status errors. cam_queue.c: cam_queue.h: Fix 'off by one' heap bug in a more efficient manner. Since heap algorithms like to deal with indexes started from 1, offset our heap array pointer at allocation time to make this so for a C environment. This makes the implementation of the algorithm a bit more efficient. cam_xpt.c: Use macros for accessing the head of the heap so that code is isolated from implementation details of the heap.
1999-04-19 21:26:08 +00:00
CAMQ_HEAD);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
device = qinfo->device;
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
Fix several potential buffer overrun conditions. These changes have been tested both in the kernel and in userland. Also, fix a couple of printf warnings that show up when CAMDEBUG is defined. Reviewed by: imp Partially submitted by: imp
1998-10-15 19:08:58 +00:00
("running device %p\n", device));
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
drvq = &device->drvq;
#ifdef CAMDEBUG
if (drvq->entries <= 0) {
panic("xpt_run_dev_allocq: "
"Device on queue without any work to do");
}
#endif
if ((work_ccb = xpt_get_ccb(device)) != NULL) {
devq->alloc_openings--;
devq->alloc_active++;
cam_periph.c: Move handling of CAM_AUTOSENSE_FAIL into block dealing with all other scsi status errors. cam_queue.c: cam_queue.h: Fix 'off by one' heap bug in a more efficient manner. Since heap algorithms like to deal with indexes started from 1, offset our heap array pointer at allocation time to make this so for a C environment. This makes the implementation of the algorithm a bit more efficient. cam_xpt.c: Use macros for accessing the head of the heap so that code is isolated from implementation details of the heap.
1999-04-19 21:26:08 +00:00
drv = (struct cam_periph*)camq_remove(drvq, CAMQ_HEAD);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_setup_ccb(&work_ccb->ccb_h, drv->path,
drv->pinfo.priority);
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
("calling periph start\n"));
drv->periph_start(drv, work_ccb);
} else {
/*
* Malloc failure in alloc_ccb
*/
/*
* XXX add us to a list to be run from free_ccb
* if we don't have any ccbs active on this
* device queue otherwise we may never get run
* again.
*/
break;
}
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (drvq->entries > 0) {
/* We have more work. Attempt to reschedule */
xpt_schedule_dev_allocq(bus, device);
}
}
devq->alloc_queue.qfrozen_cnt--;
}
static void
xpt_run_dev_sendq(struct cam_eb *bus)
{
struct cam_devq *devq;
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT, ("xpt_run_dev_sendq\n"));
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
devq = bus->sim->devq;
devq->send_queue.qfrozen_cnt++;
while ((devq->send_queue.entries > 0)
&& (devq->send_openings > 0)) {
struct cam_ed_qinfo *qinfo;
struct cam_ed *device;
union ccb *work_ccb;
struct cam_sim *sim;
if (devq->send_queue.qfrozen_cnt > 1) {
break;
}
qinfo = (struct cam_ed_qinfo *)camq_remove(&devq->send_queue,
cam_periph.c: Move handling of CAM_AUTOSENSE_FAIL into block dealing with all other scsi status errors. cam_queue.c: cam_queue.h: Fix 'off by one' heap bug in a more efficient manner. Since heap algorithms like to deal with indexes started from 1, offset our heap array pointer at allocation time to make this so for a C environment. This makes the implementation of the algorithm a bit more efficient. cam_xpt.c: Use macros for accessing the head of the heap so that code is isolated from implementation details of the heap.
1999-04-19 21:26:08 +00:00
CAMQ_HEAD);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
device = qinfo->device;
/*
* If the device has been "frozen", don't attempt
* to run it.
*/
if (device->qfrozen_cnt > 0) {
continue;
}
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT,
Fix several potential buffer overrun conditions. These changes have been tested both in the kernel and in userland. Also, fix a couple of printf warnings that show up when CAMDEBUG is defined. Reviewed by: imp Partially submitted by: imp
1998-10-15 19:08:58 +00:00
("running device %p\n", device));
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
cam_periph.c: Move handling of CAM_AUTOSENSE_FAIL into block dealing with all other scsi status errors. cam_queue.c: cam_queue.h: Fix 'off by one' heap bug in a more efficient manner. Since heap algorithms like to deal with indexes started from 1, offset our heap array pointer at allocation time to make this so for a C environment. This makes the implementation of the algorithm a bit more efficient. cam_xpt.c: Use macros for accessing the head of the heap so that code is isolated from implementation details of the heap.
1999-04-19 21:26:08 +00:00
work_ccb = cam_ccbq_peek_ccb(&device->ccbq, CAMQ_HEAD);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (work_ccb == NULL) {
Add \n to various printfs. grep shows these should be the last ones.
2002-08-27 18:43:14 +00:00
printf("device on run queue with no ccbs???\n");
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
continue;
}
if ((work_ccb->ccb_h.flags & CAM_HIGH_POWER) != 0) {
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_lock);
if (xsoftc.num_highpower <= 0) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* We got a high power command, but we
* don't have any available slots. Freeze
* the device queue until we have a slot
* available.
*/
device->qfrozen_cnt++;
style(9)
2008-01-02 01:45:31 +00:00
STAILQ_INSERT_TAIL(&xsoftc.highpowerq,
&work_ccb->ccb_h,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_links.stqe);
Add a missing mutex unlock. Reported by: Michael Jung
2007-12-16 07:49:44 +00:00
mtx_unlock(&xsoftc.xpt_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
continue;
} else {
/*
* Consume a high power slot while
* this ccb runs.
*/
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xsoftc.num_highpower--;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_unlock(&xsoftc.xpt_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
devq->active_dev = device;
cam_ccbq_remove_ccb(&device->ccbq, work_ccb);
cam_ccbq_send_ccb(&device->ccbq, work_ccb);
devq->send_openings--;
style(9)
2008-01-02 01:45:31 +00:00
devq->send_active++;
cam_periph.c: Move handling of CAM_AUTOSENSE_FAIL into block dealing with all other scsi status errors. cam_queue.c: cam_queue.h: Fix 'off by one' heap bug in a more efficient manner. Since heap algorithms like to deal with indexes started from 1, offset our heap array pointer at allocation time to make this so for a C environment. This makes the implementation of the algorithm a bit more efficient. cam_xpt.c: Use macros for accessing the head of the heap so that code is isolated from implementation details of the heap.
1999-04-19 21:26:08 +00:00
if (device->ccbq.queue.entries > 0)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_schedule_dev_sendq(bus, device);
if (work_ccb && (work_ccb->ccb_h.flags & CAM_DEV_QFREEZE) != 0){
/*
* The client wants to freeze the queue
* after this CCB is sent.
*/
device->qfrozen_cnt++;
}
Whoops- forgot to commit this other pearl from Justin- only set or clear CAM_TAG_ACTION_VALID if this is an XPT_SCSI_IO CCB (otherwise, the peripheral driver knows best...).. Obtained from:gibbs@freebsd.org
2000-07-17 00:43:47 +00:00
/* 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;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Device queues can be shared among multiple sim instances
* that reside on different busses. Use the SIM in the queue
* CCB's path, rather than the one in the bus that was passed
* into this function.
*/
sim = work_ccb->ccb_h.path->bus->sim;
(*(sim->sim_action))(sim, work_ccb);
devq->active_dev = NULL;
}
devq->send_queue.qfrozen_cnt--;
}
/*
* This function merges stuff from the slave ccb into the master ccb, while
* keeping important fields in the master ccb constant.
*/
void
xpt_merge_ccb(union ccb *master_ccb, union ccb *slave_ccb)
{
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Pull fields that are valid for peripheral drivers to set
* into the master CCB along with the CCB "payload".
*/
master_ccb->ccb_h.retry_count = slave_ccb->ccb_h.retry_count;
master_ccb->ccb_h.func_code = slave_ccb->ccb_h.func_code;
master_ccb->ccb_h.timeout = slave_ccb->ccb_h.timeout;
master_ccb->ccb_h.flags = slave_ccb->ccb_h.flags;
bcopy(&(&slave_ccb->ccb_h)[1], &(&master_ccb->ccb_h)[1],
sizeof(union ccb) - sizeof(struct ccb_hdr));
}
void
xpt_setup_ccb(struct ccb_hdr *ccb_h, struct cam_path *path, u_int32_t priority)
{
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
Remove camq_regen(). We already perform modular comparisons for generation counts, so no further steps to deal with generation count wrap are required. Fix an off by one problem in the camq heap code.
1999-04-07 22:57:48 +00:00
ccb_h->pinfo.generation = ++path->device->ccbq.queue.generation;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} else {
ccb_h->target_lun = CAM_TARGET_WILDCARD;
}
ccb_h->pinfo.index = CAM_UNQUEUED_INDEX;
ccb_h->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;
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
path = (struct cam_path *)malloc(sizeof(*path), M_CAMXPT, M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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) {
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
free(path, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
path = NULL;
}
*new_path_ptr = path;
return (status);
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
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)
{
struct cam_path *path;
struct cam_eb *bus = NULL;
cam_status status;
int need_unlock = 0;
path = (struct cam_path *)malloc(sizeof(*path), M_CAMXPT, M_WAITOK);
if (path_id != CAM_BUS_WILDCARD) {
bus = xpt_find_bus(path_id);
if (bus != NULL) {
need_unlock = 1;
Inline cam_periph_lock|unlock to make debugging easier. Use more CAM_SIM_LOCK() more uniformly.
2007-04-19 22:46:26 +00:00
CAM_SIM_LOCK(bus->sim);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
}
}
status = xpt_compile_path(path, periph, path_id, target_id, lun_id);
if (need_unlock)
Inline cam_periph_lock|unlock to make debugging easier. Use more CAM_SIM_LOCK() more uniformly.
2007-04-19 22:46:26 +00:00
CAM_SIM_UNLOCK(bus->sim);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if (status != CAM_REQ_CMP) {
free(path, M_CAMXPT);
path = NULL;
}
*new_path_ptr = path;
return (status);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static 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 */
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
/*
* We will potentially modify the EDT, so block interrupts
* that may attempt to create cam paths.
*/
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
bus = xpt_find_bus(path_id);
if (bus == NULL) {
status = CAM_PATH_INVALID;
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
} else {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
target = xpt_find_target(bus, target_id);
if (target == NULL) {
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
/* Create one */
struct cam_et *new_target;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
new_target = xpt_alloc_target(bus, target_id);
if (new_target == NULL) {
status = CAM_RESRC_UNAVAIL;
} else {
target = new_target;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
}
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
if (target != NULL) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
device = xpt_find_device(target, lun_id);
if (device == NULL) {
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
/* Create one */
struct cam_ed *new_device;
new_device = xpt_alloc_device(bus,
target,
lun_id);
if (new_device == NULL) {
status = CAM_RESRC_UNAVAIL;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} else {
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
device = new_device;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
}
}
}
/*
* 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(bus, target, device);
if (target != NULL)
xpt_release_target(bus, target);
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
if (bus != NULL)
xpt_release_bus(bus);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
return (status);
}
static void
xpt_release_path(struct cam_path *path)
{
CAM_DEBUG(path, CAM_DEBUG_TRACE, ("xpt_release_path\n"));
Release bus entries in the EDT when a path is released. Null out path entries when the path is released to catch bugs. Free our configuration hook when we're done with it.
1999-04-23 23:25:48 +00:00
if (path->device != NULL) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_release_device(path->bus, path->target, path->device);
Release bus entries in the EDT when a path is released. Null out path entries when the path is released to catch bugs. Free our configuration hook when we're done with it.
1999-04-23 23:25:48 +00:00
path->device = NULL;
}
if (path->target != NULL) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_release_target(path->bus, path->target);
Release bus entries in the EDT when a path is released. Null out path entries when the path is released to catch bugs. Free our configuration hook when we're done with it.
1999-04-23 23:25:48 +00:00
path->target = NULL;
}
if (path->bus != NULL) {
xpt_release_bus(path->bus);
path->bus = NULL;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
void
xpt_free_path(struct cam_path *path)
{
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
CAM_DEBUG(path, CAM_DEBUG_TRACE, ("xpt_free_path\n"));
xpt_release_path(path);
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
free(path, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
/*
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
* Return -1 for failure, 0 for exact match, 1 for match with wildcards
* in path1, 2 for match with wildcards in path2.
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
*/
int
xpt_path_comp(struct cam_path *path1, struct cam_path *path2)
{
int retval = 0;
if (path1->bus != path2->bus) {
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
if (path1->bus->path_id == CAM_BUS_WILDCARD)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
retval = 1;
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
else if (path2->bus->path_id == CAM_BUS_WILDCARD)
retval = 2;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
else
return (-1);
}
if (path1->target != path2->target) {
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
if (path1->target->target_id == CAM_TARGET_WILDCARD) {
if (retval == 0)
retval = 1;
} else if (path2->target->target_id == CAM_TARGET_WILDCARD)
retval = 2;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
else
return (-1);
}
if (path1->device != path2->device) {
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
if (path1->device->lun_id == CAM_LUN_WILDCARD) {
if (retval == 0)
retval = 1;
} else if (path2->device->lun_id == CAM_LUN_WILDCARD)
retval = 2;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
else
return (-1);
}
return (retval);
}
void
xpt_print_path(struct cam_path *path)
{
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (path == NULL)
printf("(nopath): ");
else {
if (path->periph != NULL)
printf("(%s%d:", path->periph->periph_name,
path->periph->unit_number);
else
printf("(noperiph:");
if (path->bus != NULL)
printf("%s%d:%d:", path->bus->sim->sim_name,
path->bus->sim->unit_number,
path->bus->sim->bus_id);
else
printf("nobus:");
if (path->target != NULL)
printf("%d:", path->target->target_id);
else
printf("X:");
if (path->device != NULL)
printf("%d): ", path->device->lun_id);
else
printf("X): ");
}
}
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
void
xpt_print(struct cam_path *path, const char *fmt, ...)
{
va_list ap;
xpt_print_path(path);
va_start(ap, fmt);
vprintf(fmt, ap);
va_end(ap);
}
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;
Guard against NULL pointer dereference. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation Found with: Coverity Prevent(tm) CID: 1847
2009-01-23 21:05:16 +00:00
#ifdef INVARIANTS
if (path != NULL && path->bus != NULL && path->bus->sim != NULL)
mtx_assert(path->bus->sim->mtx, MA_OWNED);
#endif
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
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);
if (path == NULL)
sbuf_printf(&sb, "(nopath): ");
else {
if (path->periph != NULL)
sbuf_printf(&sb, "(%s%d:", path->periph->periph_name,
path->periph->unit_number);
else
sbuf_printf(&sb, "(noperiph:");
if (path->bus != NULL)
sbuf_printf(&sb, "%s%d:%d:", path->bus->sim->sim_name,
path->bus->sim->unit_number,
path->bus->sim->bus_id);
else
sbuf_printf(&sb, "nobus:");
if (path->target != NULL)
sbuf_printf(&sb, "%d:", path->target->target_id);
else
sbuf_printf(&sb, "X:");
if (path->device != NULL)
sbuf_printf(&sb, "%d): ", path->device->lun_id);
else
sbuf_printf(&sb, "X): ");
}
sbuf_finish(&sb);
return(sbuf_len(&sb));
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
path_id_t
xpt_path_path_id(struct cam_path *path)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(path->bus->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(path->bus->path_id);
}
target_id_t
xpt_path_target_id(struct cam_path *path)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(path->bus->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(path->bus->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (path->device != NULL)
return (path->device->lun_id);
else
return (CAM_LUN_WILDCARD);
}
struct cam_sim *
xpt_path_sim(struct cam_path *path)
{
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return (path->bus->sim);
}
struct cam_periph*
xpt_path_periph(struct cam_path *path)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(path->bus->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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_path *path;
struct cam_ed *device;
struct cam_eb *bus;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
struct cam_sim *sim;
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
use CAM_DEBUG_XPT to track XPT; correct a misspelling
1998-12-06 00:06:48 +00:00
CAM_DEBUG_PRINT(CAM_DEBUG_XPT, ("xpt_release_ccb\n"));
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
path = free_ccb->ccb_h.path;
device = path->device;
bus = path->bus;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
sim = bus->sim;
mtx_assert(sim->mtx, MA_OWNED);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
cam_ccbq_release_opening(&device->ccbq);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if (sim->ccb_count > sim->max_ccbs) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_free_ccb(free_ccb);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
sim->ccb_count--;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} else {
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
SLIST_INSERT_HEAD(&sim->ccb_freeq, &free_ccb->ccb_h,
xpt_links.sle);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if (sim->devq == NULL) {
When deregistering a bus, attempt to flush out all outstanding operations before returning. Point the bus at a dummy cam_sim structure so that any CCBs will complete immediately with a CAM_DEV_NOT_THERE status, and ensure that any xpt_schedule() calls on the bus's devices will immediately call the peripheral's periph_start() routine. Also repeat the async messages because devices that were part of the way through being probed may appear after the original AC_LOST_DEVICE was sent, and would otherwise never go away. These changes make it possible to deregister a bus and free the SIM at most stages during bus probing without the usual crashes in camisr(). In particular, plugging in a umass device and then unplugging it as soon as the first probe messages appeared would almost always result in a crash. Now the device just goes away with a few CAM errors and all references to the CAM bus, target and device are dropped correctly.
2006-01-11 02:06:08 +00:00
return;
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
sim->devq->alloc_openings++;
sim->devq->alloc_active--;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* XXX Turn this into an inline function - xpt_run_device?? */
if ((device_is_alloc_queued(device) == 0)
&& (device->drvq.entries > 0)) {
xpt_schedule_dev_allocq(bus, device);
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if (dev_allocq_is_runnable(sim->devq))
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_run_dev_allocq(bus);
}
/* Functions accessed by SIM drivers */
/*
* 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 busses and assigns
* it a path_id. The path_id may be influenced by "hard wiring"
* information specified by the user. Once interrupt services are
Make "kldunload atapicam" return EBUSY instead of deadlocking when a device created by atapicam is being kept opened or mounted. This is probably just a temporary solution until we invent something better. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation Reported by: Jaakko Heinonen
2009-01-08 17:26:51 +00:00
* available, the bus will be probed.
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
*/
int32_t
Prepare for future integration between CAM and newbus. xpt_bus_register now takes a device_t to be the parent of the bus that is being created. Most SIMs have been updated with a reasonable argument, but a few exceptions just pass NULL for now. This argument isn't used yet and the newbus integration likely won't be ready until after 7.0-RELEASE.
2007-06-17 05:55:54 +00:00
xpt_bus_register(struct cam_sim *sim, device_t parent, u_int32_t bus)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
struct cam_eb *new_bus;
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
struct cam_eb *old_bus;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct ccb_pathinq cpi;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
sim->bus_id = bus;
new_bus = (struct cam_eb *)malloc(sizeof(*new_bus),
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
M_CAMXPT, M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (new_bus == NULL) {
/* Couldn't satisfy request */
return (CAM_RESRC_UNAVAIL);
}
if (strcmp(sim->sim_name, "xpt") != 0) {
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
sim->path_id =
xptpathid(sim->sim_name, sim->unit_number, sim->bus_id);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
TAILQ_INIT(&new_bus->et_entries);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
new_bus->path_id = sim->path_id;
Add SIM refcounting. This is slightly different from what DragonFly does - in DragonFly, it's cam_sim_release() what actually frees the SIM; cam_sim_free does nothing more than calling cam_sim_release(). Here, we drain in cam_sim_free, waiting for refcount to drop to zero. We cannot do the same think DragonFly does, because after cam_sim_free returns, client would destroy the sim->mtx, and CAM would trip over an initialized mutex. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation
2008-12-16 16:57:33 +00:00
cam_sim_hold(sim);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
new_bus->sim = sim;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
timevalclear(&new_bus->last_reset);
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
new_bus->flags = 0;
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
new_bus->refcount = 1; /* Held until a bus_deregister event */
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
new_bus->generation = 0;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
old_bus = TAILQ_FIRST(&xsoftc.xpt_busses);
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
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
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
TAILQ_INSERT_TAIL(&xsoftc.xpt_busses, new_bus, links);
xsoftc.bus_generation++;
mtx_unlock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Notify interested parties */
if (sim->path_id != CAM_XPT_PATH_ID) {
struct cam_path path;
xpt_compile_path(&path, /*periph*/NULL, sim->path_id,
CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD);
xpt_setup_ccb(&cpi.ccb_h, &path, /*priority*/1);
cpi.ccb_h.func_code = XPT_PATH_INQ;
xpt_action((union ccb *)&cpi);
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
xpt_async(AC_PATH_REGISTERED, &path, &cpi);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_release_path(&path);
}
return (CAM_SUCCESS);
}
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
int32_t
xpt_bus_deregister(path_id_t pathid)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
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);
When deregistering a bus, attempt to flush out all outstanding operations before returning. Point the bus at a dummy cam_sim structure so that any CCBs will complete immediately with a CAM_DEV_NOT_THERE status, and ensure that any xpt_schedule() calls on the bus's devices will immediately call the peripheral's periph_start() routine. Also repeat the async messages because devices that were part of the way through being probed may appear after the original AC_LOST_DEVICE was sent, and would otherwise never go away. These changes make it possible to deregister a bus and free the SIM at most stages during bus probing without the usual crashes in camisr(). In particular, plugging in a umass device and then unplugging it as soon as the first probe messages appeared would almost always result in a crash. Now the device just goes away with a few CAM errors and all references to the CAM bus, target and device are dropped correctly.
2006-01-11 02:06:08 +00:00
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
/* 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;
Hints overhaul: - Replace some very poorly thought out API hacks that should have been fixed a long while ago. - Provide some much more flexible search functions (resource_find_*()) - Use strings for storage instead of an outgrowth of the rather inconvenient temporary ioconf table from config(). We already had a fallback to using strings before malloc/vm was running anyway.
2001-06-12 09:40:04 +00:00
const char *strval;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
pathid = 0;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
bus = TAILQ_FIRST(&xsoftc.xpt_busses);
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
retry:
/* Find an unoccupied pathid */
Add an xpt_rescan function and a thread that will field rescan requests. The purpose of this is to allow a SIM (or other entities) to request a bus rescan and have it then fielded in a different (process) context from the caller. There are probably better ways to accomplish this, but it's a very small change that helps solve a number of problems. Reviewed by: Justin, Ken and Scott. MFC after: 2 weeks
2007-02-23 05:47:36 +00:00
while (bus != NULL && bus->path_id <= pathid) {
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
if (bus->path_id == pathid)
pathid++;
bus = TAILQ_NEXT(bus, links);
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_unlock(&xsoftc.xpt_topo_lock);
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
/*
* Ensure that this pathid is not reserved for
* a bus that may be registered in the future.
*/
A minimal update to the device wiring code so that it looks at the dynamic resource tables rather than relying on a duplicated cam-specific table generated by config(8) in ioconf.c. This was a major holdup to getting loader / userconfig driven configuration of scsi wiring.
2000-01-23 18:03:38 +00:00
if (resource_string_value("scbus", pathid, "at", &strval) == 0) {
++pathid;
/* Start the search over */
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
A minimal update to the device wiring code so that it looks at the dynamic resource tables rather than relying on a duplicated cam-specific table generated by config(8) in ioconf.c. This was a major holdup to getting loader / userconfig driven configuration of scsi wiring.
2000-01-23 18:03:38 +00:00
goto retry;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
return (pathid);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
static path_id_t
xptpathid(const char *sim_name, int sim_unit, int sim_bus)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
path_id_t pathid;
A minimal update to the device wiring code so that it looks at the dynamic resource tables rather than relying on a duplicated cam-specific table generated by config(8) in ioconf.c. This was a major holdup to getting loader / userconfig driven configuration of scsi wiring.
2000-01-23 18:03:38 +00:00
int i, dunit, val;
Slight cleanup - this is logically equivalent code but means one less use of the evil resource_locate() function.
2000-10-15 10:17:55 +00:00
char buf[32];
Hints overhaul: - Replace some very poorly thought out API hacks that should have been fixed a long while ago. - Provide some much more flexible search functions (resource_find_*()) - Use strings for storage instead of an outgrowth of the rather inconvenient temporary ioconf table from config(). We already had a fallback to using strings before malloc/vm was running anyway.
2001-06-12 09:40:04 +00:00
const char *dname;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
pathid = CAM_XPT_PATH_ID;
A minimal update to the device wiring code so that it looks at the dynamic resource tables rather than relying on a duplicated cam-specific table generated by config(8) in ioconf.c. This was a major holdup to getting loader / userconfig driven configuration of scsi wiring.
2000-01-23 18:03:38 +00:00
snprintf(buf, sizeof(buf), "%s%d", sim_name, sim_unit);
Hints overhaul: - Replace some very poorly thought out API hacks that should have been fixed a long while ago. - Provide some much more flexible search functions (resource_find_*()) - Use strings for storage instead of an outgrowth of the rather inconvenient temporary ioconf table from config(). We already had a fallback to using strings before malloc/vm was running anyway.
2001-06-12 09:40:04 +00:00
i = 0;
while ((resource_find_match(&i, &dname, &dunit, "at", buf)) == 0) {
if (strcmp(dname, "scbus")) {
Slight cleanup - this is logically equivalent code but means one less use of the evil resource_locate() function.
2000-10-15 10:17:55 +00:00
/* Avoid a bit of foot shooting. */
continue;
}
A minimal update to the device wiring code so that it looks at the dynamic resource tables rather than relying on a duplicated cam-specific table generated by config(8) in ioconf.c. This was a major holdup to getting loader / userconfig driven configuration of scsi wiring.
2000-01-23 18:03:38 +00:00
if (dunit < 0) /* unwired?! */
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
continue;
A minimal update to the device wiring code so that it looks at the dynamic resource tables rather than relying on a duplicated cam-specific table generated by config(8) in ioconf.c. This was a major holdup to getting loader / userconfig driven configuration of scsi wiring.
2000-01-23 18:03:38 +00:00
if (resource_int_value("scbus", dunit, "bus", &val) == 0) {
if (sim_bus == val) {
pathid = dunit;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
}
A minimal update to the device wiring code so that it looks at the dynamic resource tables rather than relying on a duplicated cam-specific table generated by config(8) in ioconf.c. This was a major holdup to getting loader / userconfig driven configuration of scsi wiring.
2000-01-23 18:03:38 +00:00
} 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;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
}
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
if (pathid == CAM_XPT_PATH_ID)
pathid = xptnextfreepathid();
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return (pathid);
}
void
xpt_async(u_int32_t async_code, struct cam_path *path, void *async_arg)
{
struct cam_eb *bus;
struct cam_et *target, *next_target;
struct cam_ed *device, *next_device;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(path->bus->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
CAM_DEBUG(path, CAM_DEBUG_TRACE, ("xpt_async\n"));
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
/*
* Most async events come from a CAM interrupt context. In
* a few cases, the error recovery code at the peripheral layer,
* which may run from our SWI or a process context, may signal
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
* deferred events with a call to xpt_async.
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
*/
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
bus = path->bus;
style(9)
2008-01-02 01:45:31 +00:00
if (async_code == AC_BUS_RESET) {
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
/* Update our notion of when the last reset occurred */
microtime(&bus->last_reset);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
for (target = TAILQ_FIRST(&bus->et_entries);
target != NULL;
target = next_target) {
next_target = TAILQ_NEXT(target, links);
if (path->target != target
Fix async notifications for listners registered to wildcard nodes. For example, a client registered to receive specific events for bus 0, target *, lun *, was not receiving notifications. Reviewed by: ken@FreeBSD.org
2000-11-06 20:12:07 +00:00
&& path->target->target_id != CAM_TARGET_WILDCARD
&& target->target_id != CAM_TARGET_WILDCARD)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
continue;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
if (async_code == AC_SENT_BDR) {
/* Update our notion of when the last reset occurred */
microtime(&path->target->last_reset);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
for (device = TAILQ_FIRST(&target->ed_entries);
device != NULL;
device = next_device) {
next_device = TAILQ_NEXT(device, links);
style(9)
2008-01-02 01:45:31 +00:00
if (path->device != device
Fix async notifications for listners registered to wildcard nodes. For example, a client registered to receive specific events for bus 0, target *, lun *, was not receiving notifications. Reviewed by: ken@FreeBSD.org
2000-11-06 20:12:07 +00:00
&& path->device->lun_id != CAM_LUN_WILDCARD
&& device->lun_id != CAM_LUN_WILDCARD)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
continue;
Fix async notifications for listners registered to wildcard nodes. For example, a client registered to receive specific events for bus 0, target *, lun *, was not receiving notifications. Reviewed by: ken@FreeBSD.org
2000-11-06 20:12:07 +00:00
xpt_dev_async(async_code, bus, target,
device, async_arg);
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
Fix async notifications for listners registered to wildcard nodes. For example, a client registered to receive specific events for bus 0, target *, lun *, was not receiving notifications. Reviewed by: ken@FreeBSD.org
2000-11-06 20:12:07 +00:00
xpt_async_bcast(&device->asyncs, async_code,
path, async_arg);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
}
style(9)
2008-01-02 01:45:31 +00:00
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
/*
* If this wasn't a fully wildcarded async, tell all
* clients that want all async events.
*/
if (bus != xpt_periph->path->bus)
xpt_async_bcast(&xpt_periph->path->device->asyncs, async_code,
path, async_arg);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
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;
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);
if ((cur_entry->event_enable & async_code) != 0)
cur_entry->callback(cur_entry->callback_arg,
async_code, path,
async_arg);
cur_entry = next_entry;
}
}
Fix async notifications for listners registered to wildcard nodes. For example, a client registered to receive specific events for bus 0, target *, lun *, was not receiving notifications. Reviewed by: ken@FreeBSD.org
2000-11-06 20:12:07 +00:00
/*
* Handle any per-device event notifications that require action by the XPT.
*/
static void
xpt_dev_async(u_int32_t async_code, struct cam_eb *bus, struct cam_et *target,
struct cam_ed *device, void *async_arg)
{
cam_status status;
struct cam_path newpath;
/*
* We only need to handle events for real devices.
*/
if (target->target_id == CAM_TARGET_WILDCARD
|| device->lun_id == CAM_LUN_WILDCARD)
return;
/*
* We need our own path with wildcards expanded to
* handle certain types of events.
*/
if ((async_code == AC_SENT_BDR)
|| (async_code == AC_BUS_RESET)
|| (async_code == AC_INQ_CHANGED))
status = xpt_compile_path(&newpath, NULL,
bus->path_id,
target->target_id,
device->lun_id);
else
status = CAM_REQ_CMP_ERR;
if (status == CAM_REQ_CMP) {
/*
* Allow transfer negotiation to occur in a
* tag free environment.
*/
if (async_code == AC_SENT_BDR
|| async_code == AC_BUS_RESET)
xpt_toggle_tags(&newpath);
if (async_code == AC_INQ_CHANGED) {
/*
* We've sent a start unit command, or
* something similar to a device that
* may have caused its inquiry data to
* change. So we re-scan the device to
* refresh the inquiry data for it.
*/
xpt_scan_lun(newpath.periph, &newpath,
CAM_EXPECT_INQ_CHANGE, NULL);
}
xpt_release_path(&newpath);
} else if (async_code == AC_LOST_DEVICE) {
device->flags |= CAM_DEV_UNCONFIGURED;
} else if (async_code == AC_TRANSFER_NEG) {
struct ccb_trans_settings *settings;
settings = (struct ccb_trans_settings *)async_arg;
xpt_set_transfer_settings(settings, device,
/*async_update*/TRUE);
}
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
u_int32_t
xpt_freeze_devq(struct cam_path *path, u_int count)
{
struct ccb_hdr *ccbh;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(path->bus->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
path->device->qfrozen_cnt += count;
/*
* Mark the last CCB in the queue as needing
* to be requeued if the driver hasn't
* changed it's state yet. This fixes a race
* where a ccb is just about to be queued to
* a controller driver when it's interrupt routine
* freezes the queue. To completly close the
* hole, controller drives must check to see
* if a ccb's status is still CAM_REQ_INPROG
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
* just before they queue
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* the CCB. See ahc_action/ahc_freeze_devq for
* an example.
*/
Wire up the XPT_ABORT and XPT_RESET_DEV ccb function codes so they can be delivered to controller drivers. Adjust for changes to the ccb_hdr list types in cam_queue.h
1998-12-15 08:13:10 +00:00
ccbh = TAILQ_LAST(&path->device->ccbq.active_ccbs, ccb_hdr_tailq);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (ccbh && ccbh->status == CAM_REQ_INPROG)
ccbh->status = CAM_REQUEUE_REQ;
return (path->device->qfrozen_cnt);
}
u_int32_t
xpt_freeze_simq(struct cam_sim *sim, u_int count)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
sim->devq->send_queue.qfrozen_cnt += count;
if (sim->devq->active_dev != NULL) {
struct ccb_hdr *ccbh;
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
ccbh = TAILQ_LAST(&sim->devq->active_dev->ccbq.active_ccbs,
Wire up the XPT_ABORT and XPT_RESET_DEV ccb function codes so they can be delivered to controller drivers. Adjust for changes to the ccb_hdr list types in cam_queue.h
1998-12-15 08:13:10 +00:00
ccb_hdr_tailq);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (ccbh && ccbh->status == CAM_REQ_INPROG)
ccbh->status = CAM_REQUEUE_REQ;
}
return (sim->devq->send_queue.qfrozen_cnt);
}
static void
xpt_release_devq_timeout(void *arg)
{
struct cam_ed *device;
device = (struct cam_ed *)arg;
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq_device(device, /*count*/1, /*run_queue*/TRUE);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
void
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(struct cam_path *path, u_int count, int run_queue)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(path->bus->sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq_device(path->device, count, run_queue);
}
static void
xpt_release_devq_device(struct cam_ed *dev, u_int count, int run_queue)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
int rundevq;
rundevq = 0;
if (dev->qfrozen_cnt > 0) {
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
count = (count > dev->qfrozen_cnt) ? dev->qfrozen_cnt : count;
dev->qfrozen_cnt -= count;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (dev->qfrozen_cnt == 0) {
/*
* 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) {
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
callout_stop(&dev->callout);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
dev->flags &= ~CAM_DEV_REL_TIMEOUT_PENDING;
}
/*
* Now that we are unfrozen schedule the
* device so any pending transactions are
* run.
*/
if ((dev->ccbq.queue.entries > 0)
&& (xpt_schedule_dev_sendq(dev->target->bus, dev))
&& (run_queue != 0)) {
rundevq = 1;
}
}
}
if (rundevq != 0)
xpt_run_dev_sendq(dev->target->bus);
}
void
xpt_release_simq(struct cam_sim *sim, int run_queue)
{
struct camq *sendq;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(sim->mtx, MA_OWNED);
Sprinkle GIANT_REQUIRED asserts around the xpt layer to aid with locking the SCSI drivers.
2003-11-09 02:22:33 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
sendq = &(sim->devq->send_queue);
if (sendq->qfrozen_cnt > 0) {
sendq->qfrozen_cnt--;
if (sendq->qfrozen_cnt == 0) {
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
struct cam_eb *bus;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +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){
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
callout_stop(&sim->callout);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
sim->flags &= ~CAM_SIM_REL_TIMEOUT_PENDING;
}
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
bus = xpt_find_bus(sim->path_id);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (run_queue) {
/*
* Now that we are unfrozen run the send queue.
*/
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
xpt_run_dev_sendq(bus);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
xpt_release_bus(bus);
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
}
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
/*
* XXX Appears to be unused.
*/
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static void
xpt_release_simq_timeout(void *arg)
{
struct cam_sim *sim;
sim = (struct cam_sim *)arg;
xpt_release_simq(sim, /* run_queue */ TRUE);
}
void
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
xpt_done(union ccb *done_ccb)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
struct cam_sim *sim;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
CAM_DEBUG(done_ccb->ccb_h.path, CAM_DEBUG_TRACE, ("xpt_done\n"));
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
if ((done_ccb->ccb_h.func_code & XPT_FC_QUEUED) != 0) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Queue up the request for handling by our SWI handler
* any of the "non-immediate" type of ccbs.
*/
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
sim = done_ccb->ccb_h.path->bus->sim;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
switch (done_ccb->ccb_h.path->periph->type) {
case CAM_PERIPH_BIO:
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
TAILQ_INSERT_TAIL(&sim->sim_doneq, &done_ccb->ccb_h,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
sim_links.tqe);
done_ccb->ccb_h.pinfo.index = CAM_DONEQ_INDEX;
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
if ((sim->flags & CAM_SIM_ON_DONEQ) == 0) {
mtx_lock(&cam_simq_lock);
TAILQ_INSERT_TAIL(&cam_simq, sim,
links);
sim->flags |= CAM_SIM_ON_DONEQ;
mtx_unlock(&cam_simq_lock);
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if ((done_ccb->ccb_h.path->periph->flags &
CAM_PERIPH_POLLED) == 0)
swi_sched(cambio_ih, 0);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
Remove the camnet swi and CAM_PERIPH_NET. It has never been used, and given that netowrk-over-scsi never really took off, there is little chance that it will ever be needed.
2004-10-05 04:22:20 +00:00
default:
panic("unknown periph type %d",
done_ccb->ccb_h.path->periph->type);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
}
}
union ccb *
Revert a driver API change to xpt_alloc_ccb that isn't necessary. Fix a couple of associated error checks.
2007-04-18 04:58:53 +00:00
xpt_alloc_ccb()
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
union ccb *new_ccb;
Zero the CCBs when mallocing them.
2007-04-19 14:45:37 +00:00
new_ccb = malloc(sizeof(*new_ccb), M_CAMXPT, M_ZERO|M_WAITOK);
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
return (new_ccb);
}
union ccb *
Revert a driver API change to xpt_alloc_ccb that isn't necessary. Fix a couple of associated error checks.
2007-04-18 04:58:53 +00:00
xpt_alloc_ccb_nowait()
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
{
union ccb *new_ccb;
Zero the CCBs when mallocing them.
2007-04-19 14:45:37 +00:00
new_ccb = malloc(sizeof(*new_ccb), M_CAMXPT, M_ZERO|M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return (new_ccb);
}
void
xpt_free_ccb(union ccb *free_ccb)
{
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
free(free_ccb, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
/* Private XPT functions */
/*
* Get a CAM control block for the caller. Charge the structure to the device
* referenced by the path. If the this device has no 'credits' then the
* device already has the maximum number of outstanding operations under way
* and we return NULL. If we don't have sufficient resources to allocate more
* ccbs, we also return NULL.
*/
static union ccb *
xpt_get_ccb(struct cam_ed *device)
{
union ccb *new_ccb;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
struct cam_sim *sim;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
sim = device->sim;
if ((new_ccb = (union ccb *)SLIST_FIRST(&sim->ccb_freeq)) == NULL) {
Revert a driver API change to xpt_alloc_ccb that isn't necessary. Fix a couple of associated error checks.
2007-04-18 04:58:53 +00:00
new_ccb = xpt_alloc_ccb_nowait();
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (new_ccb == NULL) {
return (NULL);
}
Revert a driver API change to xpt_alloc_ccb that isn't necessary. Fix a couple of associated error checks.
2007-04-18 04:58:53 +00:00
if ((sim->flags & CAM_SIM_MPSAFE) == 0)
callout_handle_init(&new_ccb->ccb_h.timeout_ch);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
SLIST_INSERT_HEAD(&sim->ccb_freeq, &new_ccb->ccb_h,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_links.sle);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
sim->ccb_count++;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
cam_ccbq_take_opening(&device->ccbq);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
SLIST_REMOVE_HEAD(&sim->ccb_freeq, xpt_links.sle);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return (new_ccb);
}
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
static void
xpt_release_bus(struct cam_eb *bus)
{
if ((--bus->refcount == 0)
&& (TAILQ_FIRST(&bus->et_entries) == NULL)) {
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
TAILQ_REMOVE(&xsoftc.xpt_busses, bus, links);
xsoftc.bus_generation++;
mtx_unlock(&xsoftc.xpt_topo_lock);
Add SIM refcounting. This is slightly different from what DragonFly does - in DragonFly, it's cam_sim_release() what actually frees the SIM; cam_sim_free does nothing more than calling cam_sim_release(). Here, we drain in cam_sim_free, waiting for refcount to drop to zero. We cannot do the same think DragonFly does, because after cam_sim_free returns, client would destroy the sim->mtx, and CAM would trip over an initialized mutex. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation
2008-12-16 16:57:33 +00:00
cam_sim_release(bus->sim);
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
free(bus, M_CAMXPT);
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
}
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static struct cam_et *
xpt_alloc_target(struct cam_eb *bus, target_id_t target_id)
{
struct cam_et *target;
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
target = (struct cam_et *)malloc(sizeof(*target), M_CAMXPT, M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (target != NULL) {
struct cam_et *cur_target;
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
TAILQ_INIT(&target->ed_entries);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
target->bus = bus;
target->target_id = target_id;
target->refcount = 1;
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
target->generation = 0;
timevalclear(&target->last_reset);
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
/*
* Hold a reference to our parent bus so it
* will not go away before we do.
*/
bus->refcount++;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* 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);
}
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
bus->generation++;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
return (target);
}
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
static void
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_release_target(struct cam_eb *bus, struct cam_et *target)
{
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if ((--target->refcount == 0)
&& (TAILQ_FIRST(&target->ed_entries) == NULL)) {
TAILQ_REMOVE(&bus->et_entries, target, links);
bus->generation++;
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
free(target, M_CAMXPT);
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
xpt_release_bus(bus);
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
static struct cam_ed *
xpt_alloc_device(struct cam_eb *bus, struct cam_et *target, lun_id_t 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
struct cam_path path;
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
struct cam_ed *device;
struct cam_devq *devq;
cam_status status;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Make space for us in the device queue on our bus */
devq = bus->sim->devq;
status = cam_devq_resize(devq, devq->alloc_queue.array_size + 1);
if (status != CAM_REQ_CMP) {
device = NULL;
} else {
device = (struct cam_ed *)malloc(sizeof(*device),
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
M_CAMXPT, M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
if (device != NULL) {
struct cam_ed *cur_device;
cam_init_pinfo(&device->alloc_ccb_entry.pinfo);
device->alloc_ccb_entry.device = device;
cam_init_pinfo(&device->send_ccb_entry.pinfo);
device->send_ccb_entry.device = device;
device->target = target;
device->lun_id = lun_id;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
device->sim = bus->sim;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Initialize our queues */
if (camq_init(&device->drvq, 0) != 0) {
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
free(device, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return (NULL);
}
if (cam_ccbq_init(&device->ccbq,
bus->sim->max_dev_openings) != 0) {
camq_fini(&device->drvq);
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
free(device, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return (NULL);
}
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
SLIST_INIT(&device->asyncs);
SLIST_INIT(&device->periphs);
device->generation = 0;
device->owner = NULL;
/*
* Take the default quirk entry until we have inquiry
* data and can determine a better quirk to use.
*/
device->quirk = &xpt_quirk_table[xpt_quirk_table_size - 1];
bzero(&device->inq_data, sizeof(device->inq_data));
device->inq_flags = 0;
device->queue_flags = 0;
device->serial_num = NULL;
device->serial_num_len = 0;
device->qfrozen_cnt = 0;
device->flags = CAM_DEV_UNCONFIGURED;
device->tag_delay_count = 0;
Restore queue depth settings across tag disable events. The system often disables tag queuing temporarily in order to allow controllers a window to safely perform transfer negotiation with non-compliant devices. Before this change, CAM would restore the queue depth to the controller specified maximum or device quirk level rather than any depth determined by reactions to QUEUE FULL/BUSY events or an explicit user setting. During device probe, initialize the flags field for XPT_SCAN_BUS. The uninitialized value often confused CAM into not bothering to issue an AC_FOUND_DEVICE async event for new devices. The reason this bug wasn't reported earlier is that CAM manually announces devices after the initial system bus scans. MFC: 3 days
2005-03-23 16:43:29 +00:00
device->tag_saved_openings = 0;
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
device->refcount = 1;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if (bus->sim->flags & CAM_SIM_MPSAFE)
callout_init_mtx(&device->callout, bus->sim->mtx, 0);
else
callout_init_mtx(&device->callout, &Giant, 0);
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
/*
* Hold a reference to our parent target so it
* will not go away before we do.
*/
target->refcount++;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* XXX should be limited by number of CCBs this bus can
* do.
*/
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
bus->sim->max_ccbs += device->ccbq.devq_openings;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Insertion sort into our target's device list */
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);
}
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
target->generation++;
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 (lun_id != CAM_LUN_WILDCARD) {
xpt_compile_path(&path,
NULL,
bus->path_id,
target->target_id,
lun_id);
xpt_devise_transport(&path);
xpt_release_path(&path);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
return (device);
}
static void
xpt_release_device(struct cam_eb *bus, struct cam_et *target,
struct cam_ed *device)
{
if ((--device->refcount == 0)
&& ((device->flags & CAM_DEV_UNCONFIGURED) != 0)) {
struct cam_devq *devq;
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
if (device->alloc_ccb_entry.pinfo.index != CAM_UNQUEUED_INDEX
|| device->send_ccb_entry.pinfo.index != CAM_UNQUEUED_INDEX)
panic("Removing device while still queued for ccbs");
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
if ((device->flags & CAM_DEV_REL_TIMEOUT_PENDING) != 0)
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
callout_stop(&device->callout);
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
TAILQ_REMOVE(&target->ed_entries, device,links);
target->generation++;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
bus->sim->max_ccbs -= device->ccbq.devq_openings;
Get rid of dead_sim. There is no way to make it work - any attempt to actually use it would panic on mtx operation, as dead_sim doesn't have a proper mutex. Even if it had a properly initialized mutex, it wouldn't have properly locked and owned one. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation
2008-12-16 16:54:51 +00:00
/* Release our slot in the devq */
devq = bus->sim->devq;
cam_devq_resize(devq, devq->alloc_queue.array_size - 1);
Fixing a memory leak in xpt_release_device(), which can be quickly (depends on how many memory you have) observed through "tar -tvf /dev/sa0." Without this patch, RELENG_5 and HEAD panics with something like: kmem_malloc(4096): kmem_map too small: 42258432 total allocated RELENG_4 doesn't panic but spews following errors: camq_init: - cannot malloc array! Reviewed by: gibbs, scottl Approved by: re (scottl) MFC after: 3 days
2005-06-24 08:09:05 +00:00
camq_fini(&device->drvq);
camq_fini(&device->ccbq.queue);
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
free(device, M_CAMXPT);
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
xpt_release_target(bus, target);
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
static u_int32_t
xpt_dev_ccbq_resize(struct cam_path *path, int newopenings)
{
int diff;
int result;
struct cam_ed *dev;
dev = path->device;
diff = newopenings - (dev->ccbq.dev_active + dev->ccbq.dev_openings);
result = cam_ccbq_resize(&dev->ccbq, newopenings);
if (result == CAM_REQ_CMP && (diff < 0)) {
dev->flags |= CAM_DEV_RESIZE_QUEUE_NEEDED;
}
Restore queue depth settings across tag disable events. The system often disables tag queuing temporarily in order to allow controllers a window to safely perform transfer negotiation with non-compliant devices. Before this change, CAM would restore the queue depth to the controller specified maximum or device quirk level rather than any depth determined by reactions to QUEUE FULL/BUSY events or an explicit user setting. During device probe, initialize the flags field for XPT_SCAN_BUS. The uninitialized value often confused CAM into not bothering to issue an AC_FOUND_DEVICE async event for new devices. The reason this bug wasn't reported earlier is that CAM manually announces devices after the initial system bus scans. MFC: 3 days
2005-03-23 16:43:29 +00:00
if ((dev->flags & CAM_DEV_TAG_AFTER_COUNT) != 0
|| (dev->inq_flags & SID_CmdQue) != 0)
dev->tag_saved_openings = newopenings;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Adjust the global limit */
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
dev->sim->max_ccbs += diff;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return (result);
}
static struct cam_eb *
xpt_find_bus(path_id_t path_id)
{
struct cam_eb *bus;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_topo_lock);
for (bus = TAILQ_FIRST(&xsoftc.xpt_busses);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
bus != NULL;
bus = TAILQ_NEXT(bus, links)) {
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
if (bus->path_id == path_id) {
bus->refcount++;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_unlock(&xsoftc.xpt_topo_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return (bus);
}
static struct cam_et *
xpt_find_target(struct cam_eb *bus, target_id_t target_id)
{
struct cam_et *target;
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;
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);
}
typedef struct {
union ccb *request_ccb;
struct ccb_pathinq *cpi;
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
int counter;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} xpt_scan_bus_info;
/*
* To start a scan, request_ccb is an XPT_SCAN_BUS ccb.
* As the scan progresses, xpt_scan_bus is used as the
* callback on completion function.
*/
static void
xpt_scan_bus(struct cam_periph *periph, union ccb *request_ccb)
{
CAM_DEBUG(request_ccb->ccb_h.path, CAM_DEBUG_TRACE,
("xpt_scan_bus\n"));
switch (request_ccb->ccb_h.func_code) {
case XPT_SCAN_BUS:
{
xpt_scan_bus_info *scan_info;
union ccb *work_ccb;
struct cam_path *path;
u_int i;
u_int max_target;
u_int initiator_id;
/* Find out the characteristics of the bus */
Revert a driver API change to xpt_alloc_ccb that isn't necessary. Fix a couple of associated error checks.
2007-04-18 04:58:53 +00:00
work_ccb = xpt_alloc_ccb_nowait();
if (work_ccb == NULL) {
request_ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
xpt_done(request_ccb);
return;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_setup_ccb(&work_ccb->ccb_h, request_ccb->ccb_h.path,
request_ccb->ccb_h.pinfo.priority);
work_ccb->ccb_h.func_code = XPT_PATH_INQ;
xpt_action(work_ccb);
if (work_ccb->ccb_h.status != CAM_REQ_CMP) {
request_ccb->ccb_h.status = work_ccb->ccb_h.status;
xpt_free_ccb(work_ccb);
xpt_done(request_ccb);
return;
}
Expand the hba_misc fied in the Path Inquiry ccb to allow a controller driver to specify that it does not provide initiator services (PIM_NOINITIATOR) and that the initial bus reset for device probing should be avoided (PIM_NOBUSRESET). Modify the XPT layer to honor these flags.
1998-12-10 04:05:49 +00:00
if ((work_ccb->cpi.hba_misc & PIM_NOINITIATOR) != 0) {
/*
* Can't scan the bus on an adapter that
* cannot perform the initiator role.
*/
request_ccb->ccb_h.status = CAM_REQ_CMP;
xpt_free_ccb(work_ccb);
xpt_done(request_ccb);
return;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/* Save some state for use while we probe for devices */
scan_info = (xpt_scan_bus_info *)
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
malloc(sizeof(xpt_scan_bus_info), M_CAMXPT, M_NOWAIT);
Protect against NULL pointer dereference. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation
2009-01-23 21:08:00 +00:00
if (scan_info == NULL) {
request_ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
xpt_done(request_ccb);
return;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
scan_info->request_ccb = request_ccb;
scan_info->cpi = &work_ccb->cpi;
/* Cache on our stack so we can work asynchronously */
max_target = scan_info->cpi->max_target;
initiator_id = scan_info->cpi->initiator_id;
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
* We can scan all targets in parallel, or do it sequentially.
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
*/
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
if (scan_info->cpi->hba_misc & PIM_SEQSCAN) {
max_target = 0;
scan_info->counter = 0;
} else {
scan_info->counter = scan_info->cpi->max_target + 1;
if (scan_info->cpi->initiator_id < scan_info->counter) {
scan_info->counter--;
}
}
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
for (i = 0; i <= max_target; i++) {
cam_status status;
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
if (i == initiator_id)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
continue;
status = xpt_create_path(&path, xpt_periph,
request_ccb->ccb_h.path_id,
i, 0);
if (status != CAM_REQ_CMP) {
printf("xpt_scan_bus: xpt_create_path failed"
" with status %#x, bus scan halted\n",
status);
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(scan_info, M_CAMXPT);
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
request_ccb->ccb_h.status = status;
xpt_free_ccb(work_ccb);
xpt_done(request_ccb);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
}
Revert a driver API change to xpt_alloc_ccb that isn't necessary. Fix a couple of associated error checks.
2007-04-18 04:58:53 +00:00
work_ccb = xpt_alloc_ccb_nowait();
if (work_ccb == NULL) {
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(scan_info, M_CAMXPT);
Revert a driver API change to xpt_alloc_ccb that isn't necessary. Fix a couple of associated error checks.
2007-04-18 04:58:53 +00:00
xpt_free_path(path);
request_ccb->ccb_h.status = CAM_RESRC_UNAVAIL;
xpt_done(request_ccb);
break;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_setup_ccb(&work_ccb->ccb_h, path,
request_ccb->ccb_h.pinfo.priority);
work_ccb->ccb_h.func_code = XPT_SCAN_LUN;
work_ccb->ccb_h.cbfcnp = xpt_scan_bus;
work_ccb->ccb_h.ppriv_ptr0 = scan_info;
work_ccb->crcn.flags = request_ccb->crcn.flags;
xpt_action(work_ccb);
}
break;
}
case XPT_SCAN_LUN:
{
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
cam_status status;
struct cam_path *path;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_scan_bus_info *scan_info;
path_id_t path_id;
target_id_t target_id;
lun_id_t lun_id;
/* Reuse the same CCB to query if a device was really found */
scan_info = (xpt_scan_bus_info *)request_ccb->ccb_h.ppriv_ptr0;
xpt_setup_ccb(&request_ccb->ccb_h, request_ccb->ccb_h.path,
request_ccb->ccb_h.pinfo.priority);
request_ccb->ccb_h.func_code = XPT_GDEV_TYPE;
path_id = request_ccb->ccb_h.path_id;
target_id = request_ccb->ccb_h.target_id;
lun_id = request_ccb->ccb_h.target_lun;
xpt_action(request_ccb);
if (request_ccb->ccb_h.status != CAM_REQ_CMP) {
struct cam_ed *device;
struct cam_et *target;
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
int phl;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* If we already probed lun 0 successfully, or
* we have additional configured luns on this
* target that might have "gone away", go onto
* the next lun.
*/
target = request_ccb->ccb_h.path->target;
Make sure we don't look at an edt entry that could go away when we try and figure out whether to probe higher than logical unit 7. Obtained from:gibbs@freebsd.org
1999-12-20 16:16:53 +00:00
/*
* We may touch devices that we don't
* hold references too, so ensure they
* don't disappear out from under us.
* The target above is referenced by the
* path in the request ccb.
*/
phl = 0;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
device = TAILQ_FIRST(&target->ed_entries);
Make sure we don't look at an edt entry that could go away when we try and figure out whether to probe higher than logical unit 7. Obtained from:gibbs@freebsd.org
1999-12-20 16:16:53 +00:00
if (device != NULL) {
Make the exploring of all luns supported by an HBA more of a tunable (until we get REPORT LUNS in place). If we're probing luns, and each probe succeeds, we keep going past lun 7 if we're a SCSI3 or better device (until we fail to probe). If we're probing luns, and a probe fails, we only keep going if we're quirked *for* it (CAM_QUIRK_HILUNS), and if we're not quirked *against* it (CAM_QUIRK_NOHILUNS), or we're a SCSI3 or better device and the tunable (kern.cam.cam_srch_hi) is set non-zero. Reviewed by: nate@rootlabs.org, gibbs@scsiguy.com, ken@kdm.com, scottl@samsco.org MFC after: 1 week
2005-09-16 01:26:17 +00:00
phl = CAN_SRCH_HI_SPARSE(device);
Make sure we don't look at an edt entry that could go away when we try and figure out whether to probe higher than logical unit 7. Obtained from:gibbs@freebsd.org
1999-12-20 16:16:53 +00:00
if (device->lun_id == 0)
device = TAILQ_NEXT(device, links);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if ((lun_id != 0) || (device != NULL)) {
Make sure we don't look at an edt entry that could go away when we try and figure out whether to probe higher than logical unit 7. Obtained from:gibbs@freebsd.org
1999-12-20 16:16:53 +00:00
if (lun_id < (CAM_SCSI2_MAXLUN-1) || phl)
Put in a limit (8) for lun width- do not exceed probing this unless a device is quirked for allowing it. This will allow SCSI cards to specify their true correct lun width (32 for Qlogic SCSI, 16 or 65536 for Qlogic Fibre Channel) without crippling the system because of f/w breakage of a lot of drives. Reviewed by: gibbs@freebsd.org
1999-08-16 17:47:39 +00:00
lun_id++;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
} else {
struct cam_ed *device;
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
device = request_ccb->ccb_h.path->device;
if ((device->quirk->quirks & CAM_QUIRK_NOLUNS) == 0) {
/* Try the next lun */
This is a somewhat imperfect means to try and bring FreeBSD forward in its ability to automatically scan and attach luns for modern storage which has luns in the 0..1000 range, not 0..7. The correct thing would be to do REPORT LUNS for devices whose LUN0 version shows a version >= SCSI3, but lacking that we should be able to search higher than LUN 7 if we're >= SCSI3 with no ill effects. This change keeps all of the QUIRK_HILUNS quirks, obeys the QUIRK_NOLUNS, and introduces a QUIRK_NOHILUNS which will keep searches above LUN 7 happening for devices that report >= SCSI3 compliance. I doubt the latter will be needed, but you never know. This allowed me to randomly scan and attach > 500 disks at a time in a situation where quirking for QUIRK_HILUNS wasn't practical (the vendor id and product id changes of the virtualization changes constantly). Reviewed by: ken@freebsd.org, scottl@freebsd.org, gibbs@freebsd.org MFC after: 2 weeks
2005-01-22 22:46:45 +00:00
if (lun_id < (CAM_SCSI2_MAXLUN-1)
Make the exploring of all luns supported by an HBA more of a tunable (until we get REPORT LUNS in place). If we're probing luns, and each probe succeeds, we keep going past lun 7 if we're a SCSI3 or better device (until we fail to probe). If we're probing luns, and a probe fails, we only keep going if we're quirked *for* it (CAM_QUIRK_HILUNS), and if we're not quirked *against* it (CAM_QUIRK_NOHILUNS), or we're a SCSI3 or better device and the tunable (kern.cam.cam_srch_hi) is set non-zero. Reviewed by: nate@rootlabs.org, gibbs@scsiguy.com, ken@kdm.com, scottl@samsco.org MFC after: 1 week
2005-09-16 01:26:17 +00:00
|| CAN_SRCH_HI_DENSE(device))
Put in a limit (8) for lun width- do not exceed probing this unless a device is quirked for allowing it. This will allow SCSI cards to specify their true correct lun width (32 for Qlogic SCSI, 16 or 65536 for Qlogic Fibre Channel) without crippling the system because of f/w breakage of a lot of drives. Reviewed by: gibbs@freebsd.org
1999-08-16 17:47:39 +00:00
lun_id++;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
}
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
/*
* Free the current request path- we're done with it.
*/
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_free_path(request_ccb->ccb_h.path);
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
/*
* Check to see if we scan any further luns.
*/
if (lun_id == request_ccb->ccb_h.target_lun
|| lun_id > scan_info->cpi->max_lun) {
int done;
hop_again:
done = 0;
if (scan_info->cpi->hba_misc & PIM_SEQSCAN) {
scan_info->counter++;
style(9)
2008-01-02 01:45:31 +00:00
if (scan_info->counter ==
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
scan_info->cpi->initiator_id) {
scan_info->counter++;
}
if (scan_info->counter >=
scan_info->cpi->max_target+1) {
done = 1;
}
} else {
scan_info->counter--;
if (scan_info->counter == 0) {
done = 1;
}
}
if (done) {
xpt_free_ccb(request_ccb);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_free_ccb((union ccb *)scan_info->cpi);
request_ccb = scan_info->request_ccb;
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(scan_info, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
request_ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(request_ccb);
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
break;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
if ((scan_info->cpi->hba_misc & PIM_SEQSCAN) == 0) {
break;
}
status = xpt_create_path(&path, xpt_periph,
scan_info->request_ccb->ccb_h.path_id,
scan_info->counter, 0);
if (status != CAM_REQ_CMP) {
printf("xpt_scan_bus: xpt_create_path failed"
" with status %#x, bus scan halted\n",
status);
xpt_free_ccb(request_ccb);
xpt_free_ccb((union ccb *)scan_info->cpi);
request_ccb = scan_info->request_ccb;
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(scan_info, M_CAMXPT);
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
request_ccb->ccb_h.status = status;
xpt_done(request_ccb);
break;
}
xpt_setup_ccb(&request_ccb->ccb_h, path,
request_ccb->ccb_h.pinfo.priority);
request_ccb->ccb_h.func_code = XPT_SCAN_LUN;
request_ccb->ccb_h.cbfcnp = xpt_scan_bus;
request_ccb->ccb_h.ppriv_ptr0 = scan_info;
request_ccb->crcn.flags =
scan_info->request_ccb->crcn.flags;
} else {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
status = xpt_create_path(&path, xpt_periph,
path_id, target_id, lun_id);
if (status != CAM_REQ_CMP) {
printf("xpt_scan_bus: xpt_create_path failed "
"with status %#x, halting LUN scan\n",
status);
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
goto hop_again;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
xpt_setup_ccb(&request_ccb->ccb_h, path,
request_ccb->ccb_h.pinfo.priority);
request_ccb->ccb_h.func_code = XPT_SCAN_LUN;
request_ccb->ccb_h.cbfcnp = xpt_scan_bus;
request_ccb->ccb_h.ppriv_ptr0 = scan_info;
request_ccb->crcn.flags =
scan_info->request_ccb->crcn.flags;
}
Add PIM_SEQSCAN for HBA misc flags and code that understands what to do with it. This forces us to scan targets sequentially, not in parallel. The reason we might want to do this is that SPI negotiation might not work right at the SIM level if we try to do it in parallel. We *could* fix this for each SIM where this is broken, but it's a lot harder to do that when we can simply ask CAM to probe sequentially. If PIM_SEQSCAN is not set (default), the original behaviour for probing is unchanged. LUN probing is still done in parallel for each target in either case. While we're at it, clean up some resource leakage for error cases. Reviewed by: ken, scott, scsi@ MFC after: 1 week
2006-06-05 22:22:14 +00:00
xpt_action(request_ccb);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
}
default:
break;
}
}
typedef enum {
PROBE_TUR,
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
PROBE_INQUIRY, /* this counts as DV0 for Basic Domain Validation */
Remove garbage collected tags from their usage in the ccb_getdev structure. Remove usage of the (now gone) pd_type tag of same. Add an extra probing state such that if we successfully run an initial inquiry (36 bytes), rerun another one with a longer data size as informed by the 'additional length' field in the first returned inquiry data (making sure not to get bigger than the actual scsi_inquiry_data structure- which has also been modified- see separate checkin of scsi_all.h). This allows devices such as SAF-TE devices (which have identifying marks in offsets 48-53 in inquiry data) to be successfully found without special case inquiry commands. There are also a lot of other things such as version codes that are coming in in the SPC2 specification that it would be useful to get our hands on. Reviewed by: gibbs@freebsd.org, ken@freebsd.org
2000-01-17 06:20:08 +00:00
PROBE_FULL_INQUIRY,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
PROBE_MODE_SENSE,
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
PROBE_SERIAL_NUM_0,
PROBE_SERIAL_NUM_1,
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
PROBE_TUR_FOR_NEGOTIATION,
PROBE_INQUIRY_BASIC_DV1,
PROBE_INQUIRY_BASIC_DV2,
PROBE_DV_EXIT
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} probe_action;
typedef enum {
PROBE_INQUIRY_CKSUM = 0x01,
PROBE_SERIAL_CKSUM = 0x02,
PROBE_NO_ANNOUNCE = 0x04
} probe_flags;
typedef struct {
Back out the previous change to the queue(3) interface. It was not discussed and should probably not happen. Requested by: msmith and others
2000-05-26 02:09:24 +00:00
TAILQ_HEAD(, ccb_hdr) request_ccbs;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
probe_action action;
union ccb saved_ccb;
probe_flags flags;
MD5_CTX context;
u_int8_t digest[16];
} probe_softc;
static void
xpt_scan_lun(struct cam_periph *periph, struct cam_path *path,
cam_flags flags, union ccb *request_ccb)
{
Expand the hba_misc fied in the Path Inquiry ccb to allow a controller driver to specify that it does not provide initiator services (PIM_NOINITIATOR) and that the initial bus reset for device probing should be avoided (PIM_NOBUSRESET). Modify the XPT layer to honor these flags.
1998-12-10 04:05:49 +00:00
struct ccb_pathinq cpi;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
cam_status status;
struct cam_path *new_path;
struct cam_periph *old_periph;
Retire the spl() markers. Add in some minor missed locking as a result.
2007-04-19 23:34:51 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
CAM_DEBUG(request_ccb->ccb_h.path, CAM_DEBUG_TRACE,
("xpt_scan_lun\n"));
style(9)
2008-01-02 01:45:31 +00:00
Expand the hba_misc fied in the Path Inquiry ccb to allow a controller driver to specify that it does not provide initiator services (PIM_NOINITIATOR) and that the initial bus reset for device probing should be avoided (PIM_NOBUSRESET). Modify the XPT layer to honor these flags.
1998-12-10 04:05:49 +00:00
xpt_setup_ccb(&cpi.ccb_h, path, /*priority*/1);
cpi.ccb_h.func_code = XPT_PATH_INQ;
xpt_action((union ccb *)&cpi);
if (cpi.ccb_h.status != CAM_REQ_CMP) {
if (request_ccb != NULL) {
request_ccb->ccb_h.status = cpi.ccb_h.status;
xpt_done(request_ccb);
}
return;
}
if ((cpi.hba_misc & PIM_NOINITIATOR) != 0) {
/*
* Can't scan the bus on an adapter that
* cannot perform the initiator role.
*/
if (request_ccb != NULL) {
request_ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(request_ccb);
}
return;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (request_ccb == NULL) {
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
request_ccb = malloc(sizeof(union ccb), M_CAMXPT, M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (request_ccb == NULL) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(path, "xpt_scan_lun: can't allocate CCB, "
"can't continue\n");
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return;
}
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
new_path = malloc(sizeof(*new_path), M_CAMXPT, M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (new_path == NULL) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(path, "xpt_scan_lun: can't allocate path, "
"can't continue\n");
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(request_ccb, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return;
}
Fix a bug in the scan lun code that showed up when we did the following sequence of things: - spin up a disk - send an async event to refresh the inquiry data - run through xpt_scan_lun() to re-probe the device - eventually finish the probe, but panic in xpt_done() because the periph pointer wasn't set. Reviewed by: gibbs Reported by: Philippe Regnauld <regnauld@deepo.prosa.dk>
1998-10-13 21:29:04 +00:00
status = xpt_compile_path(new_path, xpt_periph,
path->bus->path_id,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
path->target->target_id,
path->device->lun_id);
if (status != CAM_REQ_CMP) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(path, "xpt_scan_lun: can't compile path, "
"can't continue\n");
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(request_ccb, M_CAMXPT);
free(new_path, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return;
}
xpt_setup_ccb(&request_ccb->ccb_h, new_path, /*priority*/ 1);
request_ccb->ccb_h.cbfcnp = xptscandone;
request_ccb->ccb_h.func_code = XPT_SCAN_LUN;
request_ccb->crcn.flags = flags;
}
if ((old_periph = cam_periph_find(path, "probe")) != NULL) {
probe_softc *softc;
softc = (probe_softc *)old_periph->softc;
TAILQ_INSERT_TAIL(&softc->request_ccbs, &request_ccb->ccb_h,
periph_links.tqe);
} else {
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
status = cam_periph_alloc(proberegister, NULL, probecleanup,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
probestart, "probe",
CAM_PERIPH_BIO,
request_ccb->ccb_h.path, NULL, 0,
request_ccb);
if (status != CAM_REQ_CMP) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(path, "xpt_scan_lun: cam_alloc_periph "
"returned an error, can't continue probe\n");
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
request_ccb->ccb_h.status = status;
xpt_done(request_ccb);
}
}
}
static void
xptscandone(struct cam_periph *periph, union ccb *done_ccb)
{
xpt_release_path(done_ccb->ccb_h.path);
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(done_ccb->ccb_h.path, M_CAMXPT);
free(done_ccb, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
static cam_status
proberegister(struct cam_periph *periph, void *arg)
{
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
union ccb *request_ccb; /* CCB representing the probe request */
A silly buglet found by Coverity- check the return value from cam_periph_acquire.
2007-03-15 06:42:54 +00:00
cam_status status;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
probe_softc *softc;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
request_ccb = (union ccb *)arg;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (periph == NULL) {
printf("proberegister: periph was NULL!!\n");
return(CAM_REQ_CMP_ERR);
}
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
if (request_ccb == NULL) {
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
printf("proberegister: no probe CCB, "
"can't register device\n");
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(CAM_REQ_CMP_ERR);
}
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
softc = (probe_softc *)malloc(sizeof(*softc), M_CAMXPT, M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (softc == NULL) {
printf("proberegister: Unable to probe new device. "
style(9)
2008-01-02 01:45:31 +00:00
"Unable to allocate softc\n");
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(CAM_REQ_CMP_ERR);
}
TAILQ_INIT(&softc->request_ccbs);
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
TAILQ_INSERT_TAIL(&softc->request_ccbs, &request_ccb->ccb_h,
periph_links.tqe);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
softc->flags = 0;
periph->softc = softc;
A silly buglet found by Coverity- check the return value from cam_periph_acquire.
2007-03-15 06:42:54 +00:00
status = cam_periph_acquire(periph);
if (status != CAM_REQ_CMP) {
return (status);
}
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
/*
* Ensure we've waited at least a bus settle
* delay before attempting to probe the device.
minor comment update
1999-12-13 19:40:00 +00:00
* For HBAs that don't do bus resets, this won't make a difference.
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
*/
cam_periph_freeze_after_event(periph, &periph->path->bus->last_reset,
Make SCSI_DELAY setable at boot time and runtime via the kern.cam.scsi_delay tunable/sysctl. Reviewed by: mdodd, njl
2002-09-02 20:10:19 +00:00
scsi_delay);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
probeschedule(periph);
return(CAM_REQ_CMP);
}
static void
probeschedule(struct cam_periph *periph)
{
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
struct ccb_pathinq cpi;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
union ccb *ccb;
probe_softc *softc;
softc = (probe_softc *)periph->softc;
ccb = (union ccb *)TAILQ_FIRST(&softc->request_ccbs);
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
xpt_setup_ccb(&cpi.ccb_h, periph->path, /*priority*/1);
cpi.ccb_h.func_code = XPT_PATH_INQ;
xpt_action((union ccb *)&cpi);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* If a device has gone away and another device, or the same one,
* is back in the same place, it should have a unit attention
* condition pending. It will not report the unit attention in
* response to an inquiry, which may leave invalid transfer
* negotiations in effect. The TUR will reveal the unit attention
style(9)
2008-01-02 01:45:31 +00:00
* condition. Only send the TUR for lun 0, since some devices
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* will get confused by commands other than inquiry to non-existent
* luns. If you think a device has gone away start your scan from
* lun 0. This will insure that any bogus transfer settings are
* invalidated.
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
*
* If we haven't seen the device before and the controller supports
* some kind of transfer negotiation, negotiate with the first
* sent command if no bus reset was performed at startup. This
* ensures that the device is not confused by transfer negotiation
* settings left over by loader or BIOS action.
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
*/
Add support for wildcard device entries in the EDT. The target mode 'Black Hole' device uses this feature to schedule itself against any target or lun attached to a controller that receives an unwanted request from an initiator instead of having an instance per potential target/lun request. Use the wildcard entries to simplify wildcard async callback storage. Don't announce devices twice to peripheral drivers. The devices will be announced as soon as the AC_PATH_REGISTERED event is registered by the peripheral driver, so no manaul push of this event is required. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1999-01-14 06:03:59 +00:00
if (((ccb->ccb_h.path->device->flags & CAM_DEV_UNCONFIGURED) == 0)
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
&& (ccb->ccb_h.target_lun == 0)) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
softc->action = PROBE_TUR;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
} else if ((cpi.hba_inquiry & (PI_WIDE_32|PI_WIDE_16|PI_SDTR_ABLE)) != 0
&& (cpi.hba_misc & PIM_NOBUSRESET) != 0) {
proberequestdefaultnegotiation(periph);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
softc->action = PROBE_INQUIRY;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
} else {
softc->action = PROBE_INQUIRY;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (ccb->crcn.flags & CAM_EXPECT_INQ_CHANGE)
softc->flags |= PROBE_NO_ANNOUNCE;
else
softc->flags &= ~PROBE_NO_ANNOUNCE;
xpt_schedule(periph, ccb->ccb_h.pinfo.priority);
}
static void
probestart(struct cam_periph *periph, union ccb *start_ccb)
{
/* Probe the device that our peripheral driver points to */
struct ccb_scsiio *csio;
probe_softc *softc;
CAM_DEBUG(start_ccb->ccb_h.path, CAM_DEBUG_TRACE, ("probestart\n"));
softc = (probe_softc *)periph->softc;
csio = &start_ccb->csio;
switch (softc->action) {
case PROBE_TUR:
case PROBE_TUR_FOR_NEGOTIATION:
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
case PROBE_DV_EXIT:
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
scsi_test_unit_ready(csio,
cam: Retry TEST UNIT READY command if not successful. This fixes problems with discovering some USB devices that are very slow to respond during initialisation. When a USB device is inserted, CAM performs the sequence: 1) INQUIRY 2) INQUIRY (second time with other parameters) 3) TEST UNIT READY 4) READ CAPACITY Before this change CAM didn't check if TEST UNIT READY was successful and went on blindly to the next state and sent READ CAPACITY. If the device was still not ready by then, CAM ended with error message. This patch adds checking for the status of TEST UNIT READY command and retrying up to 10 times with 0.5 sec interval. Submitted by: Grzegorz Bernacki gjb ! semihalf dot com Reviewed by: scottl
2009-01-08 10:16:02 +00:00
/*retries*/10,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
probedone,
MSG_SIMPLE_Q_TAG,
SSD_FULL_SIZE,
cam_xpt.c: Add quirk entry for a Samsung drive that doesn't like experiencing the queue full condition. Bump the timeouts for all probe activities to 60s. We don't know what the seletion timeout (or equivelent on other mediums) is for controllers, which can make the transactions at the tail end of a parallel probe take a while to complete. The DPT seems to be a card that takes a long time to see a selection timeout. cam_periph.c: Don't call a device "gone" after a single selection timeout. We need to come up with a better policy. Until that time, you'll have to manually re-scan a bus via camcontrol for the system to decide that a device is really gone. This should give devices experiencing temporary insanity to escape death.
1998-09-20 07:14:36 +00:00
/*timeout*/60000);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
}
case PROBE_INQUIRY:
Remove garbage collected tags from their usage in the ccb_getdev structure. Remove usage of the (now gone) pd_type tag of same. Add an extra probing state such that if we successfully run an initial inquiry (36 bytes), rerun another one with a longer data size as informed by the 'additional length' field in the first returned inquiry data (making sure not to get bigger than the actual scsi_inquiry_data structure- which has also been modified- see separate checkin of scsi_all.h). This allows devices such as SAF-TE devices (which have identifying marks in offsets 48-53 in inquiry data) to be successfully found without special case inquiry commands. There are also a lot of other things such as version codes that are coming in in the SPC2 specification that it would be useful to get our hands on. Reviewed by: gibbs@freebsd.org, ken@freebsd.org
2000-01-17 06:20:08 +00:00
case PROBE_FULL_INQUIRY:
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
case PROBE_INQUIRY_BASIC_DV1:
case PROBE_INQUIRY_BASIC_DV2:
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
Redo the SHORT INQUIRY length stuff to be a bit cleaner. Obtained from:gibbs@freebsd.org
2000-01-25 18:25:22 +00:00
u_int inquiry_len;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct scsi_inquiry_data *inq_buf;
inq_buf = &periph->path->device->inq_data;
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* If the device is currently configured, we calculate an
* MD5 checksum of the inquiry data, and if the serial number
* length is greater than 0, add the serial number data
* into the checksum as well. Once the inquiry and the
* serial number check finish, we attempt to figure out
* whether we still have the same device.
*/
if ((periph->path->device->flags & CAM_DEV_UNCONFIGURED) == 0) {
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
MD5Init(&softc->context);
MD5Update(&softc->context, (unsigned char *)inq_buf,
sizeof(struct scsi_inquiry_data));
softc->flags |= PROBE_INQUIRY_CKSUM;
if (periph->path->device->serial_num_len > 0) {
MD5Update(&softc->context,
periph->path->device->serial_num,
periph->path->device->serial_num_len);
softc->flags |= PROBE_SERIAL_CKSUM;
}
MD5Final(softc->digest, &softc->context);
style(9)
2008-01-02 01:45:31 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Remove garbage collected tags from their usage in the ccb_getdev structure. Remove usage of the (now gone) pd_type tag of same. Add an extra probing state such that if we successfully run an initial inquiry (36 bytes), rerun another one with a longer data size as informed by the 'additional length' field in the first returned inquiry data (making sure not to get bigger than the actual scsi_inquiry_data structure- which has also been modified- see separate checkin of scsi_all.h). This allows devices such as SAF-TE devices (which have identifying marks in offsets 48-53 in inquiry data) to be successfully found without special case inquiry commands. There are also a lot of other things such as version codes that are coming in in the SPC2 specification that it would be useful to get our hands on. Reviewed by: gibbs@freebsd.org, ken@freebsd.org
2000-01-17 06:20:08 +00:00
if (softc->action == PROBE_INQUIRY)
Redo the SHORT INQUIRY length stuff to be a bit cleaner. Obtained from:gibbs@freebsd.org
2000-01-25 18:25:22 +00:00
inquiry_len = SHORT_INQUIRY_LENGTH;
else
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
inquiry_len = SID_ADDITIONAL_LENGTH(inq_buf);
Correct an off-by-one in the calculation of the inquiry buffer length to request from devices during the "long inquiry" portion of our probe. This same bug was fixed in the 4.x stream a few years ago, but the fix was never propogated to -current. This fix is slightly different than in -stable: o Use offsetof() instead of a hard coded constant so as the make the code more self-explainatory. o Round odd long inquiry lengths up so as to avoid tickling ignore wide residue bugs in broken parallel SCSI devices running with a wide transfer negotiation. MFC: 3 days
2004-11-18 04:14:32 +00:00
/*
* Some parallel SCSI devices fail to send an
* ignore wide residue message when dealing with
* odd length inquiry requests. Round up to be
* safe.
*/
inquiry_len = roundup2(inquiry_len, 2);
style(9)
2008-01-02 01:45:31 +00:00
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
if (softc->action == PROBE_INQUIRY_BASIC_DV1
|| softc->action == PROBE_INQUIRY_BASIC_DV2) {
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
inq_buf = malloc(inquiry_len, M_CAMXPT, M_NOWAIT);
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
}
if (inq_buf == NULL) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(periph->path, "malloc failure- skipping Basic"
"Domain Validation\n");
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
softc->action = PROBE_DV_EXIT;
scsi_test_unit_ready(csio,
/*retries*/4,
probedone,
MSG_SIMPLE_Q_TAG,
SSD_FULL_SIZE,
/*timeout*/60000);
break;
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
scsi_inquiry(csio,
/*retries*/4,
probedone,
MSG_SIMPLE_Q_TAG,
(u_int8_t *)inq_buf,
Redo the SHORT INQUIRY length stuff to be a bit cleaner. Obtained from:gibbs@freebsd.org
2000-01-25 18:25:22 +00:00
inquiry_len,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*evpd*/FALSE,
/*page_code*/0,
SSD_MIN_SIZE,
cam_xpt.c: Add quirk entry for a Samsung drive that doesn't like experiencing the queue full condition. Bump the timeouts for all probe activities to 60s. We don't know what the seletion timeout (or equivelent on other mediums) is for controllers, which can make the transactions at the tail end of a parallel probe take a while to complete. The DPT seems to be a card that takes a long time to see a selection timeout. cam_periph.c: Don't call a device "gone" after a single selection timeout. We need to come up with a better policy. Until that time, you'll have to manually re-scan a bus via camcontrol for the system to decide that a device is really gone. This should give devices experiencing temporary insanity to escape death.
1998-09-20 07:14:36 +00:00
/*timeout*/60 * 1000);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
}
case PROBE_MODE_SENSE:
{
void *mode_buf;
int mode_buf_len;
mode_buf_len = sizeof(struct scsi_mode_header_6)
+ sizeof(struct scsi_mode_blk_desc)
+ sizeof(struct scsi_control_page);
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
mode_buf = malloc(mode_buf_len, M_CAMXPT, M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (mode_buf != NULL) {
scsi_mode_sense(csio,
/*retries*/4,
probedone,
MSG_SIMPLE_Q_TAG,
/*dbd*/FALSE,
SMS_PAGE_CTRL_CURRENT,
SMS_CONTROL_MODE_PAGE,
mode_buf,
mode_buf_len,
SSD_FULL_SIZE,
cam_xpt.c: Add quirk entry for a Samsung drive that doesn't like experiencing the queue full condition. Bump the timeouts for all probe activities to 60s. We don't know what the seletion timeout (or equivelent on other mediums) is for controllers, which can make the transactions at the tail end of a parallel probe take a while to complete. The DPT seems to be a card that takes a long time to see a selection timeout. cam_periph.c: Don't call a device "gone" after a single selection timeout. We need to come up with a better policy. Until that time, you'll have to manually re-scan a bus via camcontrol for the system to decide that a device is really gone. This should give devices experiencing temporary insanity to escape death.
1998-09-20 07:14:36 +00:00
/*timeout*/60000);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
}
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(periph->path, "Unable to mode sense control page - "
"malloc failure\n");
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
softc->action = PROBE_SERIAL_NUM_0;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Add /* FALLTHROUGH */ Move /* FALLTHROUGH */ to correct location. Remove unused variable(s). Found by: FlexeLint
2003-05-31 20:46:21 +00:00
/* FALLTHROUGH */
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
case PROBE_SERIAL_NUM_0:
{
struct scsi_vpd_supported_page_list *vpd_list = NULL;
struct cam_ed *device;
device = periph->path->device;
if ((device->quirk->quirks & CAM_QUIRK_NOSERIAL) == 0) {
vpd_list = malloc(sizeof(*vpd_list), M_CAMXPT,
M_NOWAIT | M_ZERO);
}
if (vpd_list != NULL) {
scsi_inquiry(csio,
/*retries*/4,
probedone,
MSG_SIMPLE_Q_TAG,
(u_int8_t *)vpd_list,
sizeof(*vpd_list),
/*evpd*/TRUE,
SVPD_SUPPORTED_PAGE_LIST,
SSD_MIN_SIZE,
/*timeout*/60 * 1000);
break;
}
/*
* We'll have to do without, let our probedone
* routine finish up for us.
*/
start_ccb->csio.data_ptr = NULL;
probedone(periph, start_ccb);
return;
}
case PROBE_SERIAL_NUM_1:
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
struct scsi_vpd_unit_serial_number *serial_buf;
struct cam_ed* device;
serial_buf = NULL;
device = periph->path->device;
device->serial_num = NULL;
device->serial_num_len = 0;
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
serial_buf = (struct scsi_vpd_unit_serial_number *)
malloc(sizeof(*serial_buf), M_CAMXPT, M_NOWAIT|M_ZERO);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (serial_buf != NULL) {
scsi_inquiry(csio,
/*retries*/4,
probedone,
MSG_SIMPLE_Q_TAG,
(u_int8_t *)serial_buf,
sizeof(*serial_buf),
/*evpd*/TRUE,
SVPD_UNIT_SERIAL_NUMBER,
SSD_MIN_SIZE,
cam_xpt.c: Add quirk entry for a Samsung drive that doesn't like experiencing the queue full condition. Bump the timeouts for all probe activities to 60s. We don't know what the seletion timeout (or equivelent on other mediums) is for controllers, which can make the transactions at the tail end of a parallel probe take a while to complete. The DPT seems to be a card that takes a long time to see a selection timeout. cam_periph.c: Don't call a device "gone" after a single selection timeout. We need to come up with a better policy. Until that time, you'll have to manually re-scan a bus via camcontrol for the system to decide that a device is really gone. This should give devices experiencing temporary insanity to escape death.
1998-09-20 07:14:36 +00:00
/*timeout*/60 * 1000);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
break;
}
/*
* We'll have to do without, let our probedone
* routine finish up for us.
*/
start_ccb->csio.data_ptr = NULL;
probedone(periph, start_ccb);
return;
}
}
xpt_action(start_ccb);
}
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
static void
proberequestdefaultnegotiation(struct cam_periph *periph)
{
struct ccb_trans_settings cts;
xpt_setup_ccb(&cts.ccb_h, periph->path, /*priority*/1);
cts.ccb_h.func_code = XPT_GET_TRAN_SETTINGS;
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
cts.type = CTS_TYPE_USER_SETTINGS;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
xpt_action((union ccb *)&cts);
XPT_GET_TRAN_SETTINGS can fail- so check for an error and don't blindly assume it succeeded.
2006-11-07 23:06:00 +00:00
if ((cts.ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
return;
}
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
cts.ccb_h.func_code = XPT_SET_TRAN_SETTINGS;
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
cts.type = CTS_TYPE_CURRENT_SETTINGS;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
xpt_action((union ccb *)&cts);
}
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
/*
* Backoff Negotiation Code- only pertinent for SPI devices.
*/
static int
proberequestbackoff(struct cam_periph *periph, struct cam_ed *device)
{
struct ccb_trans_settings cts;
struct ccb_trans_settings_spi *spi;
memset(&cts, 0, sizeof (cts));
xpt_setup_ccb(&cts.ccb_h, periph->path, /*priority*/1);
cts.ccb_h.func_code = XPT_GET_TRAN_SETTINGS;
cts.type = CTS_TYPE_CURRENT_SETTINGS;
xpt_action((union ccb *)&cts);
if ((cts.ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
if (bootverbose) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(periph->path,
"failed to get current device settings\n");
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
}
return (0);
}
if (cts.transport != XPORT_SPI) {
if (bootverbose) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(periph->path, "not SPI transport\n");
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
}
return (0);
}
spi = &cts.xport_specific.spi;
/*
* We cannot renegotiate sync rate if we don't have one.
*/
if ((spi->valid & CTS_SPI_VALID_SYNC_RATE) == 0) {
if (bootverbose) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(periph->path, "no sync rate known\n");
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
}
return (0);
}
/*
* We'll assert that we don't have to touch PPR options- the
* SIM will see what we do with period and offset and adjust
* the PPR options as appropriate.
*/
/*
* A sync rate with unknown or zero offset is nonsensical.
* A sync period of zero means Async.
*/
if ((spi->valid & CTS_SPI_VALID_SYNC_OFFSET) == 0
|| spi->sync_offset == 0 || spi->sync_period == 0) {
if (bootverbose) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(periph->path, "no sync rate available\n");
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
}
return (0);
}
if (device->flags & CAM_DEV_DV_HIT_BOTTOM) {
CAM_DEBUG(periph->path, CAM_DEBUG_INFO,
("hit async: giving up on DV\n"));
return (0);
}
/*
* Jump sync_period up by one, but stop at 5MHz and fall back to Async.
* We don't try to remember 'last' settings to see if the SIM actually
* gets into the speed we want to set. We check on the SIM telling
* us that a requested speed is bad, but otherwise don't try and
* check the speed due to the asynchronous and handshake nature
* of speed setting.
*/
spi->valid = CTS_SPI_VALID_SYNC_RATE | CTS_SPI_VALID_SYNC_OFFSET;
for (;;) {
spi->sync_period++;
if (spi->sync_period >= 0xf) {
spi->sync_period = 0;
spi->sync_offset = 0;
CAM_DEBUG(periph->path, CAM_DEBUG_INFO,
("setting to async for DV\n"));
/*
* Once we hit async, we don't want to try
* any more settings.
*/
device->flags |= CAM_DEV_DV_HIT_BOTTOM;
} else if (bootverbose) {
CAM_DEBUG(periph->path, CAM_DEBUG_INFO,
("DV: period 0x%x\n", spi->sync_period));
printf("setting period to 0x%x\n", spi->sync_period);
}
cts.ccb_h.func_code = XPT_SET_TRAN_SETTINGS;
cts.type = CTS_TYPE_CURRENT_SETTINGS;
xpt_action((union ccb *)&cts);
if ((cts.ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP) {
break;
}
CAM_DEBUG(periph->path, CAM_DEBUG_INFO,
("DV: failed to set period 0x%x\n", spi->sync_period));
if (spi->sync_period == 0) {
return (0);
}
}
return (1);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static void
probedone(struct cam_periph *periph, union ccb *done_ccb)
{
probe_softc *softc;
struct cam_path *path;
u_int32_t priority;
CAM_DEBUG(done_ccb->ccb_h.path, CAM_DEBUG_TRACE, ("probedone\n"));
softc = (probe_softc *)periph->softc;
path = done_ccb->ccb_h.path;
priority = done_ccb->ccb_h.pinfo.priority;
switch (softc->action) {
case PROBE_TUR:
{
if ((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
if (cam_periph_error(done_ccb, 0,
SF_NO_PRINT, NULL) == ERESTART)
return;
else if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0)
/* Don't wedge the queue */
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(done_ccb->ccb_h.path,
/*count*/1,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*run_queue*/TRUE);
}
softc->action = PROBE_INQUIRY;
xpt_release_ccb(done_ccb);
xpt_schedule(periph, priority);
return;
}
case PROBE_INQUIRY:
Remove garbage collected tags from their usage in the ccb_getdev structure. Remove usage of the (now gone) pd_type tag of same. Add an extra probing state such that if we successfully run an initial inquiry (36 bytes), rerun another one with a longer data size as informed by the 'additional length' field in the first returned inquiry data (making sure not to get bigger than the actual scsi_inquiry_data structure- which has also been modified- see separate checkin of scsi_all.h). This allows devices such as SAF-TE devices (which have identifying marks in offsets 48-53 in inquiry data) to be successfully found without special case inquiry commands. There are also a lot of other things such as version codes that are coming in in the SPC2 specification that it would be useful to get our hands on. Reviewed by: gibbs@freebsd.org, ken@freebsd.org
2000-01-17 06:20:08 +00:00
case PROBE_FULL_INQUIRY:
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
if ((done_ccb->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP) {
struct scsi_inquiry_data *inq_buf;
u_int8_t periph_qual;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
path->device->flags |= CAM_DEV_INQUIRY_DATA_VALID;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
inq_buf = &path->device->inq_data;
periph_qual = SID_QUAL(inq_buf);
style(9)
2008-01-02 01:45:31 +00:00
Bring the probe inquiry code in line with the SCSI spec. It is legal to have a device with device type 0x1f, that just means that the device is of unknown type. Instead, only check the peripheral qualifier when deciding whether or not to reject a device based on its inquiry information. Tested by: julian MFC after: 3 weeks
2001-11-27 03:34:22 +00:00
switch(periph_qual) {
case SID_QUAL_LU_CONNECTED:
{
Correct another location where inquiry length as calculated from the inquiry additional length field is off by one. MFC: 3 days
2004-11-18 07:24:29 +00:00
u_int8_t len;
Redo the SHORT INQUIRY length stuff to be a bit cleaner. Obtained from:gibbs@freebsd.org
2000-01-25 18:25:22 +00:00
Bring the probe inquiry code in line with the SCSI spec. It is legal to have a device with device type 0x1f, that just means that the device is of unknown type. Instead, only check the peripheral qualifier when deciding whether or not to reject a device based on its inquiry information. Tested by: julian MFC after: 3 weeks
2001-11-27 03:34:22 +00:00
/*
* We conservatively request only
* SHORT_INQUIRY_LEN bytes of inquiry
* information during our first try
* at sending an INQUIRY. If the device
* has more information to give,
* perform a second request specifying
* the amount of information the device
* is willing to give.
*/
Correct another location where inquiry length as calculated from the inquiry additional length field is off by one. MFC: 3 days
2004-11-18 07:24:29 +00:00
len = inq_buf->additional_length
+ offsetof(struct scsi_inquiry_data,
additional_length) + 1;
Bring the probe inquiry code in line with the SCSI spec. It is legal to have a device with device type 0x1f, that just means that the device is of unknown type. Instead, only check the peripheral qualifier when deciding whether or not to reject a device based on its inquiry information. Tested by: julian MFC after: 3 weeks
2001-11-27 03:34:22 +00:00
if (softc->action == PROBE_INQUIRY
style(9)
2008-01-02 01:45:31 +00:00
&& len > SHORT_INQUIRY_LENGTH) {
Bring the probe inquiry code in line with the SCSI spec. It is legal to have a device with device type 0x1f, that just means that the device is of unknown type. Instead, only check the peripheral qualifier when deciding whether or not to reject a device based on its inquiry information. Tested by: julian MFC after: 3 weeks
2001-11-27 03:34:22 +00:00
softc->action = PROBE_FULL_INQUIRY;
xpt_release_ccb(done_ccb);
xpt_schedule(periph, priority);
return;
}
Redo the SHORT INQUIRY length stuff to be a bit cleaner. Obtained from:gibbs@freebsd.org
2000-01-25 18:25:22 +00:00
Bring the probe inquiry code in line with the SCSI spec. It is legal to have a device with device type 0x1f, that just means that the device is of unknown type. Instead, only check the peripheral qualifier when deciding whether or not to reject a device based on its inquiry information. Tested by: julian MFC after: 3 weeks
2001-11-27 03:34:22 +00:00
xpt_find_quirk(path->device);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Bring the probe inquiry code in line with the SCSI spec. It is legal to have a device with device type 0x1f, that just means that the device is of unknown type. Instead, only check the peripheral qualifier when deciding whether or not to reject a device based on its inquiry information. Tested by: julian MFC after: 3 weeks
2001-11-27 03:34:22 +00:00
xpt_devise_transport(path);
Handle some of the inquiry flags that have come into usage as of SPC2r20. Specifically, handle the BQueue flag which will indicate that a device supports the Basic Queueing model (no Head of Queue or Ordered tags). When this flag is set, SID_CmdQueue is clear. This has causes FreeBSD to assume that the device did not support tagged operations. MFC after: 1 month
2006-05-30 22:44:00 +00:00
if (INQ_DATA_TQ_ENABLED(inq_buf))
Bring the probe inquiry code in line with the SCSI spec. It is legal to have a device with device type 0x1f, that just means that the device is of unknown type. Instead, only check the peripheral qualifier when deciding whether or not to reject a device based on its inquiry information. Tested by: julian MFC after: 3 weeks
2001-11-27 03:34:22 +00:00
softc->action = PROBE_MODE_SENSE;
else
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
softc->action = PROBE_SERIAL_NUM_0;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Bring the probe inquiry code in line with the SCSI spec. It is legal to have a device with device type 0x1f, that just means that the device is of unknown type. Instead, only check the peripheral qualifier when deciding whether or not to reject a device based on its inquiry information. Tested by: julian MFC after: 3 weeks
2001-11-27 03:34:22 +00:00
path->device->flags &= ~CAM_DEV_UNCONFIGURED;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Bring the probe inquiry code in line with the SCSI spec. It is legal to have a device with device type 0x1f, that just means that the device is of unknown type. Instead, only check the peripheral qualifier when deciding whether or not to reject a device based on its inquiry information. Tested by: julian MFC after: 3 weeks
2001-11-27 03:34:22 +00:00
xpt_release_ccb(done_ccb);
xpt_schedule(periph, priority);
return;
}
default:
break;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
} else if (cam_periph_error(done_ccb, 0,
done_ccb->ccb_h.target_lun > 0
? SF_RETRY_UA|SF_QUIET_IR
: SF_RETRY_UA,
&softc->saved_ccb) == ERESTART) {
return;
} else if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) {
/* Don't wedge the queue */
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(done_ccb->ccb_h.path, /*count*/1,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*run_queue*/TRUE);
}
/*
* If we get to this point, we got an error status back
* from the inquiry and the error status doesn't require
* automatically retrying the command. Therefore, the
* inquiry failed. If we had inquiry information before
* for this device, but this latest inquiry command failed,
* the device has probably gone away. If this device isn't
* already marked unconfigured, notify the peripheral
* drivers that this device is no more.
*/
if ((path->device->flags & CAM_DEV_UNCONFIGURED) == 0)
/* Send the async notification. */
xpt_async(AC_LOST_DEVICE, path, NULL);
xpt_release_ccb(done_ccb);
break;
}
case PROBE_MODE_SENSE:
{
struct ccb_scsiio *csio;
struct scsi_mode_header_6 *mode_hdr;
csio = &done_ccb->csio;
mode_hdr = (struct scsi_mode_header_6 *)csio->data_ptr;
if ((csio->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP) {
struct scsi_control_page *page;
u_int8_t *offset;
offset = ((u_int8_t *)&mode_hdr[1])
+ mode_hdr->blk_desc_len;
page = (struct scsi_control_page *)offset;
path->device->queue_flags = page->queue_flags;
} else if (cam_periph_error(done_ccb, 0,
SF_RETRY_UA|SF_NO_PRINT,
&softc->saved_ccb) == ERESTART) {
return;
} else if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) {
/* Don't wedge the queue */
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(done_ccb->ccb_h.path,
/*count*/1, /*run_queue*/TRUE);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
xpt_release_ccb(done_ccb);
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(mode_hdr, M_CAMXPT);
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
softc->action = PROBE_SERIAL_NUM_0;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_schedule(periph, priority);
return;
}
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
case PROBE_SERIAL_NUM_0:
{
struct ccb_scsiio *csio;
struct scsi_vpd_supported_page_list *page_list;
int length, serialnum_supported, i;
serialnum_supported = 0;
csio = &done_ccb->csio;
page_list =
(struct scsi_vpd_supported_page_list *)csio->data_ptr;
if (page_list == NULL) {
/*
* Don't process the command as it was never sent
*/
} else if ((csio->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP
&& (page_list->length > 0)) {
length = min(page_list->length,
SVPD_SUPPORTED_PAGES_SIZE);
for (i = 0; i < length; i++) {
if (page_list->list[i] ==
SVPD_UNIT_SERIAL_NUMBER) {
serialnum_supported = 1;
break;
}
}
} else if (cam_periph_error(done_ccb, 0,
SF_RETRY_UA|SF_NO_PRINT,
&softc->saved_ccb) == ERESTART) {
return;
} else if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) {
/* Don't wedge the queue */
xpt_release_devq(done_ccb->ccb_h.path, /*count*/1,
/*run_queue*/TRUE);
}
if (page_list != NULL)
free(page_list, M_DEVBUF);
if (serialnum_supported) {
xpt_release_ccb(done_ccb);
softc->action = PROBE_SERIAL_NUM_1;
xpt_schedule(periph, priority);
return;
}
xpt_release_ccb(done_ccb);
softc->action = PROBE_TUR_FOR_NEGOTIATION;
Fix use after free. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation Found with: Coverity Prevent(tm) CID: 3712
2009-01-14 21:29:20 +00:00
xpt_schedule(periph, priority);
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
return;
}
Fix whitespace.
2008-08-01 15:13:56 +00:00
When probing a newly found device, don't automatically assume that the device supports retrieving a serial number. Instead, first query the list of VPD pages it does support, and only query the serial number if it's supported, else silently move on. This eliminates a lot of noise during verbose booting, and will likely eliminate the need for most NOSERIAL quirks.
2008-02-27 08:47:13 +00:00
case PROBE_SERIAL_NUM_1:
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
struct ccb_scsiio *csio;
struct scsi_vpd_unit_serial_number *serial_buf;
u_int32_t priority;
int changed;
int have_serialnum;
changed = 1;
have_serialnum = 0;
csio = &done_ccb->csio;
priority = done_ccb->ccb_h.pinfo.priority;
serial_buf =
(struct scsi_vpd_unit_serial_number *)csio->data_ptr;
/* Clean up from previous instance of this device */
if (path->device->serial_num != NULL) {
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
free(path->device->serial_num, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
path->device->serial_num = NULL;
path->device->serial_num_len = 0;
}
if (serial_buf == NULL) {
/*
* Don't process the command as it was never sent
*/
} else if ((csio->ccb_h.status & CAM_STATUS_MASK) == CAM_REQ_CMP
Release bus entries in the EDT when a path is released. Null out path entries when the path is released to catch bugs. Free our configuration hook when we're done with it.
1999-04-23 23:25:48 +00:00
&& (serial_buf->length > 0)) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
have_serialnum = 1;
path->device->serial_num =
(u_int8_t *)malloc((serial_buf->length + 1),
- Providing fine-grained malloc statistic by replacing M_DEVBUF with module-specific malloc types. These should help us to pinpoint the possible memory leakage in the future. - Implementing xpt_alloc_ccb_nowait() and replacing all malloc/free based CCB memory management with xpt_alloc_ccb[_nowait]/xpt_free_ccb. Hopefully this would be helpful if someday we move the CCB allocator to use UMA instead of malloc(). Encouraged by: jeffr, rwatson Reviewed by: gibbs, scottl Approved by: re (scottl)
2005-07-01 15:21:30 +00:00
M_CAMXPT, M_NOWAIT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (path->device->serial_num != NULL) {
bcopy(serial_buf->serial_num,
path->device->serial_num,
serial_buf->length);
path->device->serial_num_len =
serial_buf->length;
path->device->serial_num[serial_buf->length]
= '\0';
}
} else if (cam_periph_error(done_ccb, 0,
SF_RETRY_UA|SF_NO_PRINT,
&softc->saved_ccb) == ERESTART) {
return;
} else if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) {
/* Don't wedge the queue */
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(done_ccb->ccb_h.path, /*count*/1,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*run_queue*/TRUE);
}
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*
* Let's see if we have seen this device before.
*/
if ((softc->flags & PROBE_INQUIRY_CKSUM) != 0) {
MD5_CTX context;
u_int8_t digest[16];
MD5Init(&context);
style(9)
2008-01-02 01:45:31 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
MD5Update(&context,
(unsigned char *)&path->device->inq_data,
sizeof(struct scsi_inquiry_data));
if (have_serialnum)
MD5Update(&context, serial_buf->serial_num,
serial_buf->length);
MD5Final(digest, &context);
if (bcmp(softc->digest, digest, 16) == 0)
changed = 0;
/*
* XXX Do we need to do a TUR in order to ensure
* that the device really hasn't changed???
*/
if ((changed != 0)
&& ((softc->flags & PROBE_NO_ANNOUNCE) == 0))
xpt_async(AC_LOST_DEVICE, path, NULL);
}
if (serial_buf != NULL)
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(serial_buf, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (changed != 0) {
/*
* Now that we have all the necessary
* information to safely perform transfer
* negotiations... Controllers don't perform
* any negotiation or tagged queuing until
* after the first XPT_SET_TRAN_SETTINGS ccb is
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
* received. So, on a new device, just retrieve
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* the user settings, and set them as the current
* settings to set the device up.
*/
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
proberequestdefaultnegotiation(periph);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_release_ccb(done_ccb);
/*
* Perform a TUR to allow the controller to
* perform any necessary transfer negotiation.
*/
softc->action = PROBE_TUR_FOR_NEGOTIATION;
xpt_schedule(periph, priority);
return;
}
xpt_release_ccb(done_ccb);
break;
}
case PROBE_TUR_FOR_NEGOTIATION:
cam: Retry TEST UNIT READY command if not successful. This fixes problems with discovering some USB devices that are very slow to respond during initialisation. When a USB device is inserted, CAM performs the sequence: 1) INQUIRY 2) INQUIRY (second time with other parameters) 3) TEST UNIT READY 4) READ CAPACITY Before this change CAM didn't check if TEST UNIT READY was successful and went on blindly to the next state and sent READ CAPACITY. If the device was still not ready by then, CAM ended with error message. This patch adds checking for the status of TEST UNIT READY command and retrying up to 10 times with 0.5 sec interval. Submitted by: Grzegorz Bernacki gjb ! semihalf dot com Reviewed by: scottl
2009-01-08 10:16:02 +00:00
if ((done_ccb->ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
DELAY(500000);
if (cam_periph_error(done_ccb, 0, SF_RETRY_UA,
NULL) == ERESTART)
return;
}
/* FALLTHROUGH */
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
case PROBE_DV_EXIT:
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) {
/* Don't wedge the queue */
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(done_ccb->ccb_h.path, /*count*/1,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*run_queue*/TRUE);
}
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
/*
* Do Domain Validation for lun 0 on devices that claim
* to support Synchronous Transfer modes.
*/
if (softc->action == PROBE_TUR_FOR_NEGOTIATION
&& done_ccb->ccb_h.target_lun == 0
&& (path->device->inq_data.flags & SID_Sync) != 0
&& (path->device->flags & CAM_DEV_IN_DV) == 0) {
CAM_DEBUG(periph->path, CAM_DEBUG_INFO,
("Begin Domain Validation\n"));
path->device->flags |= CAM_DEV_IN_DV;
xpt_release_ccb(done_ccb);
softc->action = PROBE_INQUIRY_BASIC_DV1;
xpt_schedule(periph, priority);
return;
}
if (softc->action == PROBE_DV_EXIT) {
CAM_DEBUG(periph->path, CAM_DEBUG_INFO,
("Leave Domain Validation\n"));
}
path->device->flags &=
~(CAM_DEV_UNCONFIGURED|CAM_DEV_IN_DV|CAM_DEV_DV_HIT_BOTTOM);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if ((softc->flags & PROBE_NO_ANNOUNCE) == 0) {
/* Inform the XPT that a new device has been found */
done_ccb->ccb_h.func_code = XPT_GDEV_TYPE;
xpt_action(done_ccb);
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
xpt_async(AC_FOUND_DEVICE, done_ccb->ccb_h.path,
done_ccb);
}
xpt_release_ccb(done_ccb);
break;
case PROBE_INQUIRY_BASIC_DV1:
case PROBE_INQUIRY_BASIC_DV2:
{
struct scsi_inquiry_data *nbuf;
struct ccb_scsiio *csio;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
if ((done_ccb->ccb_h.status & CAM_DEV_QFRZN) != 0) {
/* Don't wedge the queue */
xpt_release_devq(done_ccb->ccb_h.path, /*count*/1,
/*run_queue*/TRUE);
}
csio = &done_ccb->csio;
nbuf = (struct scsi_inquiry_data *)csio->data_ptr;
if (bcmp(nbuf, &path->device->inq_data, SHORT_INQUIRY_LENGTH)) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(path,
"inquiry data fails comparison at DV%d step\n",
style(9)
2008-01-02 01:45:31 +00:00
softc->action == PROBE_INQUIRY_BASIC_DV1 ? 1 : 2);
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
if (proberequestbackoff(periph, path->device)) {
path->device->flags &= ~CAM_DEV_IN_DV;
softc->action = PROBE_TUR_FOR_NEGOTIATION;
} else {
/* give up */
softc->action = PROBE_DV_EXIT;
}
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(nbuf, M_CAMXPT);
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
xpt_release_ccb(done_ccb);
xpt_schedule(periph, priority);
return;
}
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(nbuf, M_CAMXPT);
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
if (softc->action == PROBE_INQUIRY_BASIC_DV1) {
softc->action = PROBE_INQUIRY_BASIC_DV2;
xpt_release_ccb(done_ccb);
xpt_schedule(periph, priority);
return;
}
if (softc->action == PROBE_DV_EXIT) {
CAM_DEBUG(periph->path, CAM_DEBUG_INFO,
("Leave Domain Validation Successfully\n"));
}
path->device->flags &=
~(CAM_DEV_UNCONFIGURED|CAM_DEV_IN_DV|CAM_DEV_DV_HIT_BOTTOM);
if ((softc->flags & PROBE_NO_ANNOUNCE) == 0) {
/* Inform the XPT that a new device has been found */
done_ccb->ccb_h.func_code = XPT_GDEV_TYPE;
xpt_action(done_ccb);
Restore queue depth settings across tag disable events. The system often disables tag queuing temporarily in order to allow controllers a window to safely perform transfer negotiation with non-compliant devices. Before this change, CAM would restore the queue depth to the controller specified maximum or device quirk level rather than any depth determined by reactions to QUEUE FULL/BUSY events or an explicit user setting. During device probe, initialize the flags field for XPT_SCAN_BUS. The uninitialized value often confused CAM into not bothering to issue an AC_FOUND_DEVICE async event for new devices. The reason this bug wasn't reported earlier is that CAM manually announces devices after the initial system bus scans. MFC: 3 days
2005-03-23 16:43:29 +00:00
xpt_async(AC_FOUND_DEVICE, done_ccb->ccb_h.path,
done_ccb);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
xpt_release_ccb(done_ccb);
break;
}
Initial cut at Basic Domain Validation- just add some states to traverse through during probing. Reviewed by: scsi (scottl)
2006-12-04 23:04:13 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
done_ccb = (union ccb *)TAILQ_FIRST(&softc->request_ccbs);
TAILQ_REMOVE(&softc->request_ccbs, &done_ccb->ccb_h, periph_links.tqe);
done_ccb->ccb_h.status = CAM_REQ_CMP;
xpt_done(done_ccb);
if (TAILQ_FIRST(&softc->request_ccbs) == NULL) {
cam_periph_invalidate(periph);
Periph driver fixes, second try. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation
2008-12-19 14:31:40 +00:00
cam_periph_release_locked(periph);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
} else {
probeschedule(periph);
}
}
static void
probecleanup(struct cam_periph *periph)
{
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(periph->softc, M_CAMXPT);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
static void
xpt_find_quirk(struct cam_ed *device)
{
caddr_t match;
match = cam_quirkmatch((caddr_t)&device->inq_data,
(caddr_t)xpt_quirk_table,
sizeof(xpt_quirk_table)/sizeof(*xpt_quirk_table),
sizeof(*xpt_quirk_table), scsi_inquiry_match);
if (match == NULL)
panic("xpt_find_quirk: device didn't match wildcard entry!!");
device->quirk = (struct xpt_quirk_entry *)match;
}
Make the exploring of all luns supported by an HBA more of a tunable (until we get REPORT LUNS in place). If we're probing luns, and each probe succeeds, we keep going past lun 7 if we're a SCSI3 or better device (until we fail to probe). If we're probing luns, and a probe fails, we only keep going if we're quirked *for* it (CAM_QUIRK_HILUNS), and if we're not quirked *against* it (CAM_QUIRK_NOHILUNS), or we're a SCSI3 or better device and the tunable (kern.cam.cam_srch_hi) is set non-zero. Reviewed by: nate@rootlabs.org, gibbs@scsiguy.com, ken@kdm.com, scottl@samsco.org MFC after: 1 week
2005-09-16 01:26:17 +00:00
static int
sysctl_cam_search_luns(SYSCTL_HANDLER_ARGS)
{
int error, bool;
bool = cam_srch_hi;
Despite several examples in the kernel, the third argument of sysctl_handle_int is not sizeof the int type you want to export. The type must always be an int or an unsigned int. Remove the instances where a sizeof(variable) is passed to stop people accidently cut and pasting these examples. In a few places this was sysctl_handle_int was being used on 64 bit types, which would truncate the value to be exported. In these cases use sysctl_handle_quad to export them and change the format to Q so that sysctl(1) can still print them.
2007-06-04 18:25:08 +00:00
error = sysctl_handle_int(oidp, &bool, 0, req);
Make the exploring of all luns supported by an HBA more of a tunable (until we get REPORT LUNS in place). If we're probing luns, and each probe succeeds, we keep going past lun 7 if we're a SCSI3 or better device (until we fail to probe). If we're probing luns, and a probe fails, we only keep going if we're quirked *for* it (CAM_QUIRK_HILUNS), and if we're not quirked *against* it (CAM_QUIRK_NOHILUNS), or we're a SCSI3 or better device and the tunable (kern.cam.cam_srch_hi) is set non-zero. Reviewed by: nate@rootlabs.org, gibbs@scsiguy.com, ken@kdm.com, scottl@samsco.org MFC after: 1 week
2005-09-16 01:26:17 +00:00
if (error != 0 || req->newptr == NULL)
return (error);
if (bool == 0 || bool == 1) {
cam_srch_hi = bool;
return (0);
} else {
return (EINVAL);
}
}
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
static void
xpt_devise_transport(struct cam_path *path)
{
struct ccb_pathinq cpi;
struct ccb_trans_settings cts;
struct scsi_inquiry_data *inq_buf;
/* Get transport information from the SIM */
xpt_setup_ccb(&cpi.ccb_h, path, /*priority*/1);
cpi.ccb_h.func_code = XPT_PATH_INQ;
xpt_action((union ccb *)&cpi);
inq_buf = NULL;
if ((path->device->flags & CAM_DEV_INQUIRY_DATA_VALID) != 0)
inq_buf = &path->device->inq_data;
path->device->protocol = PROTO_SCSI;
path->device->protocol_version =
inq_buf != NULL ? SID_ANSI_REV(inq_buf) : cpi.protocol_version;
path->device->transport = cpi.transport;
path->device->transport_version = cpi.transport_version;
/*
* Any device not using SPI3 features should
* be considered SPI2 or lower.
*/
if (inq_buf != NULL) {
if (path->device->transport == XPORT_SPI
&& (inq_buf->spi3data & SID_SPI_MASK) == 0
&& path->device->transport_version > 2)
path->device->transport_version = 2;
} else {
struct cam_ed* otherdev;
for (otherdev = TAILQ_FIRST(&path->target->ed_entries);
otherdev != NULL;
otherdev = TAILQ_NEXT(otherdev, links)) {
if (otherdev != path->device)
break;
}
style(9)
2008-01-02 01:45:31 +00:00
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 (otherdev != NULL) {
/*
* Initially assume the same versioning as
* prior luns for this target.
*/
path->device->protocol_version =
otherdev->protocol_version;
path->device->transport_version =
otherdev->transport_version;
} else {
/* Until we know better, opt for safty */
path->device->protocol_version = 2;
if (path->device->transport == XPORT_SPI)
path->device->transport_version = 2;
else
path->device->transport_version = 0;
}
}
/*
* XXX
* For a device compliant with SPC-2 we should be able
* to determine the transport version supported by
* scrutinizing the version descriptors in the
* inquiry buffer.
*/
/* Tell the controller what we think */
xpt_setup_ccb(&cts.ccb_h, path, /*priority*/1);
cts.ccb_h.func_code = XPT_SET_TRAN_SETTINGS;
cts.type = CTS_TYPE_CURRENT_SETTINGS;
cts.transport = path->device->transport;
cts.transport_version = path->device->transport_version;
cts.protocol = path->device->protocol;
cts.protocol_version = path->device->protocol_version;
cts.proto_specific.valid = 0;
cts.xport_specific.valid = 0;
xpt_action((union ccb *)&cts);
}
static void
xpt_set_transfer_settings(struct ccb_trans_settings *cts, struct cam_ed *device,
int async_update)
{
struct ccb_pathinq cpi;
struct ccb_trans_settings cur_cts;
struct ccb_trans_settings_scsi *scsi;
struct ccb_trans_settings_scsi *cur_scsi;
struct cam_sim *sim;
struct scsi_inquiry_data *inq_data;
if (device == NULL) {
cts->ccb_h.status = CAM_PATH_INVALID;
xpt_done((union ccb *)cts);
return;
}
if (cts->protocol == PROTO_UNKNOWN
|| cts->protocol == PROTO_UNSPECIFIED) {
cts->protocol = device->protocol;
cts->protocol_version = device->protocol_version;
}
if (cts->protocol_version == PROTO_VERSION_UNKNOWN
|| cts->protocol_version == PROTO_VERSION_UNSPECIFIED)
cts->protocol_version = device->protocol_version;
if (cts->protocol != device->protocol) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(cts->ccb_h.path, "Uninitialized Protocol %x:%x?\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
cts->protocol, device->protocol);
cts->protocol = device->protocol;
}
if (cts->protocol_version > device->protocol_version) {
if (bootverbose) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(cts->ccb_h.path, "Down reving Protocol "
"Version from %d to %d?\n", cts->protocol_version,
device->protocol_version);
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
}
cts->protocol_version = device->protocol_version;
}
if (cts->transport == XPORT_UNKNOWN
|| cts->transport == XPORT_UNSPECIFIED) {
cts->transport = device->transport;
cts->transport_version = device->transport_version;
}
if (cts->transport_version == XPORT_VERSION_UNKNOWN
|| cts->transport_version == XPORT_VERSION_UNSPECIFIED)
cts->transport_version = device->transport_version;
if (cts->transport != device->transport) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(cts->ccb_h.path, "Uninitialized Transport %x:%x?\n",
cts->transport, device->transport);
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
cts->transport = device->transport;
}
if (cts->transport_version > device->transport_version) {
if (bootverbose) {
Add an xpt_print function to reduce most of the xpt_print_path/printf pairs. Convert the core code to use it. Reviewed by: scsi (various folks weighing in with good advice) MFC after: 2 weeks
2006-12-05 07:45:28 +00:00
xpt_print(cts->ccb_h.path, "Down reving Transport "
"Version from %d to %d?\n", cts->transport_version,
device->transport_version);
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
}
cts->transport_version = device->transport_version;
}
sim = cts->ccb_h.path->bus->sim;
/*
* Nothing more of interest to do unless
* this is a device connected via the
* SCSI protocol.
*/
if (cts->protocol != PROTO_SCSI) {
style(9)
2008-01-02 01:45:31 +00:00
if (async_update == FALSE)
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
(*(sim->sim_action))(sim, (union ccb *)cts);
return;
}
inq_data = &device->inq_data;
scsi = &cts->proto_specific.scsi;
xpt_setup_ccb(&cpi.ccb_h, cts->ccb_h.path, /*priority*/1);
cpi.ccb_h.func_code = XPT_PATH_INQ;
xpt_action((union ccb *)&cpi);
/* SCSI specific sanity checking */
if ((cpi.hba_inquiry & PI_TAG_ABLE) == 0
Handle some of the inquiry flags that have come into usage as of SPC2r20. Specifically, handle the BQueue flag which will indicate that a device supports the Basic Queueing model (no Head of Queue or Ordered tags). When this flag is set, SID_CmdQueue is clear. This has causes FreeBSD to assume that the device did not support tagged operations. MFC after: 1 month
2006-05-30 22:44:00 +00:00
|| (INQ_DATA_TQ_ENABLED(inq_data)) == 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
|| (device->queue_flags & SCP_QUEUE_DQUE) != 0
|| (device->quirk->mintags == 0)) {
/*
* Can't tag on hardware that doesn't support tags,
* doesn't have it enabled, or has broken tag support.
*/
scsi->flags &= ~CTS_SCSI_FLAGS_TAG_ENB;
}
if (async_update == FALSE) {
/*
* Perform sanity checking against what the
* controller and device can do.
*/
xpt_setup_ccb(&cur_cts.ccb_h, cts->ccb_h.path, /*priority*/1);
cur_cts.ccb_h.func_code = XPT_GET_TRAN_SETTINGS;
cur_cts.type = cts->type;
xpt_action((union ccb *)&cur_cts);
XPT_GET_TRAN_SETTINGS can fail- so check for an error and don't blindly assume it succeeded.
2006-11-07 23:06:00 +00:00
if ((cur_cts.ccb_h.status & CAM_STATUS_MASK) != CAM_REQ_CMP) {
return;
}
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
cur_scsi = &cur_cts.proto_specific.scsi;
if ((scsi->valid & CTS_SCSI_VALID_TQ) == 0) {
scsi->flags &= ~CTS_SCSI_FLAGS_TAG_ENB;
scsi->flags |= cur_scsi->flags & CTS_SCSI_FLAGS_TAG_ENB;
}
if ((cur_scsi->valid & CTS_SCSI_VALID_TQ) == 0)
scsi->flags &= ~CTS_SCSI_FLAGS_TAG_ENB;
}
/* SPI specific sanity checking */
cleanup some of the XPORT_FC code
2001-06-24 18:19:24 +00:00
if (cts->transport == XPORT_SPI && async_update == FALSE) {
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 spi3caps;
struct ccb_trans_settings_spi *spi;
struct ccb_trans_settings_spi *cur_spi;
spi = &cts->xport_specific.spi;
cur_spi = &cur_cts.xport_specific.spi;
/* Fill in any gaps in what the user gave us */
if ((spi->valid & CTS_SPI_VALID_SYNC_RATE) == 0)
spi->sync_period = cur_spi->sync_period;
if ((cur_spi->valid & CTS_SPI_VALID_SYNC_RATE) == 0)
spi->sync_period = 0;
if ((spi->valid & CTS_SPI_VALID_SYNC_OFFSET) == 0)
spi->sync_offset = cur_spi->sync_offset;
if ((cur_spi->valid & CTS_SPI_VALID_SYNC_OFFSET) == 0)
spi->sync_offset = 0;
if ((spi->valid & CTS_SPI_VALID_PPR_OPTIONS) == 0)
spi->ppr_options = cur_spi->ppr_options;
if ((cur_spi->valid & CTS_SPI_VALID_PPR_OPTIONS) == 0)
spi->ppr_options = 0;
if ((spi->valid & CTS_SPI_VALID_BUS_WIDTH) == 0)
spi->bus_width = cur_spi->bus_width;
if ((cur_spi->valid & CTS_SPI_VALID_BUS_WIDTH) == 0)
spi->bus_width = 0;
if ((spi->valid & CTS_SPI_VALID_DISC) == 0) {
spi->flags &= ~CTS_SPI_FLAGS_DISC_ENB;
spi->flags |= cur_spi->flags & CTS_SPI_FLAGS_DISC_ENB;
}
if ((cur_spi->valid & CTS_SPI_VALID_DISC) == 0)
spi->flags &= ~CTS_SPI_FLAGS_DISC_ENB;
if (((device->flags & CAM_DEV_INQUIRY_DATA_VALID) != 0
&& (inq_data->flags & SID_Sync) == 0
&& cts->type == CTS_TYPE_CURRENT_SETTINGS)
|| ((cpi.hba_inquiry & PI_SDTR_ABLE) == 0)
Fix the hanging chad if the NEW_TRAN_CODE change and change the qualifier to force async from cur_spi to spi.
2006-11-07 05:51:40 +00:00
|| (spi->sync_offset == 0)
|| (spi->sync_period == 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
/* Force async */
spi->sync_period = 0;
spi->sync_offset = 0;
}
switch (spi->bus_width) {
case MSG_EXT_WDTR_BUS_32_BIT:
if (((device->flags & CAM_DEV_INQUIRY_DATA_VALID) == 0
|| (inq_data->flags & SID_WBus32) != 0
|| cts->type == CTS_TYPE_USER_SETTINGS)
&& (cpi.hba_inquiry & PI_WIDE_32) != 0)
break;
/* Fall Through to 16-bit */
case MSG_EXT_WDTR_BUS_16_BIT:
if (((device->flags & CAM_DEV_INQUIRY_DATA_VALID) == 0
|| (inq_data->flags & SID_WBus16) != 0
|| cts->type == CTS_TYPE_USER_SETTINGS)
&& (cpi.hba_inquiry & PI_WIDE_16) != 0) {
spi->bus_width = MSG_EXT_WDTR_BUS_16_BIT;
break;
}
/* Fall Through to 8-bit */
default: /* New bus width?? */
case MSG_EXT_WDTR_BUS_8_BIT:
/* All targets can do this */
spi->bus_width = MSG_EXT_WDTR_BUS_8_BIT;
break;
}
spi3caps = cpi.xport_specific.spi.ppr_options;
if ((device->flags & CAM_DEV_INQUIRY_DATA_VALID) != 0
&& cts->type == CTS_TYPE_CURRENT_SETTINGS)
spi3caps &= inq_data->spi3data;
if ((spi3caps & SID_SPI_CLOCK_DT) == 0)
spi->ppr_options &= ~MSG_EXT_PPR_DT_REQ;
if ((spi3caps & SID_SPI_IUS) == 0)
spi->ppr_options &= ~MSG_EXT_PPR_IU_REQ;
if ((spi3caps & SID_SPI_QAS) == 0)
spi->ppr_options &= ~MSG_EXT_PPR_QAS_REQ;
/* No SPI Transfer settings are allowed unless we are wide */
if (spi->bus_width == 0)
spi->ppr_options = 0;
if ((spi->flags & CTS_SPI_FLAGS_DISC_ENB) == 0) {
/*
* Can't tag queue without disconnection.
*/
scsi->flags &= ~CTS_SCSI_FLAGS_TAG_ENB;
scsi->valid |= CTS_SCSI_VALID_TQ;
}
/*
* If we are currently performing tagged transactions to
* this device and want to change its negotiation parameters,
* go non-tagged for a bit to give the controller a chance to
* negotiate unhampered by tag messages.
*/
if (cts->type == CTS_TYPE_CURRENT_SETTINGS
&& (device->inq_flags & SID_CmdQue) != 0
&& (scsi->flags & CTS_SCSI_FLAGS_TAG_ENB) != 0
&& (spi->flags & (CTS_SPI_VALID_SYNC_RATE|
CTS_SPI_VALID_SYNC_OFFSET|
CTS_SPI_VALID_BUS_WIDTH)) != 0)
xpt_toggle_tags(cts->ccb_h.path);
}
if (cts->type == CTS_TYPE_CURRENT_SETTINGS
&& (scsi->valid & CTS_SCSI_VALID_TQ) != 0) {
int device_tagenb;
/*
* If we are transitioning from tags to no-tags or
* vice-versa, we need to carefully freeze and restart
* the queue so that we don't overlap tagged and non-tagged
* commands. We also temporarily stop tags if there is
* a change in transfer negotiation settings to allow
* "tag-less" negotiation.
*/
if ((device->flags & CAM_DEV_TAG_AFTER_COUNT) != 0
|| (device->inq_flags & SID_CmdQue) != 0)
device_tagenb = TRUE;
else
device_tagenb = FALSE;
if (((scsi->flags & CTS_SCSI_FLAGS_TAG_ENB) != 0
&& device_tagenb == FALSE)
|| ((scsi->flags & CTS_SCSI_FLAGS_TAG_ENB) == 0
&& device_tagenb == TRUE)) {
if ((scsi->flags & CTS_SCSI_FLAGS_TAG_ENB) != 0) {
/*
* Delay change to use tags until after a
* few commands have gone to this device so
* the controller has time to perform transfer
* negotiations without tagged messages getting
* in the way.
*/
device->tag_delay_count = CAM_TAG_DELAY_COUNT;
device->flags |= CAM_DEV_TAG_AFTER_COUNT;
} else {
struct ccb_relsim crs;
xpt_freeze_devq(cts->ccb_h.path, /*count*/1);
device->inq_flags &= ~SID_CmdQue;
xpt_dev_ccbq_resize(cts->ccb_h.path,
sim->max_dev_openings);
device->flags &= ~CAM_DEV_TAG_AFTER_COUNT;
device->tag_delay_count = 0;
xpt_setup_ccb(&crs.ccb_h, cts->ccb_h.path,
/*priority*/1);
crs.ccb_h.func_code = XPT_REL_SIMQ;
crs.release_flags = RELSIM_RELEASE_AFTER_QEMPTY;
crs.openings
style(9)
2008-01-02 01:45:31 +00:00
= crs.release_timeout
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
= crs.qfrozen_cnt
= 0;
xpt_action((union ccb *)&crs);
}
}
}
style(9)
2008-01-02 01:45:31 +00:00
if (async_update == FALSE)
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
(*(sim->sim_action))(sim, (union ccb *)cts);
}
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
static void
Fix a few problems with the tag delay code: - Tagged devices were limited to one transaction (oops) - We revert to untagged with a tag delay if the user changed the transfer negotiation values (via perhaps camcontrol some day). - xpt_async did not use the expanded path in some cases which could cause a panic.
1998-09-24 22:43:54 +00:00
xpt_toggle_tags(struct cam_path *path)
{
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
struct cam_ed *dev;
Fix a few problems with the tag delay code: - Tagged devices were limited to one transaction (oops) - We revert to untagged with a tag delay if the user changed the transfer negotiation values (via perhaps camcontrol some day). - xpt_async did not use the expanded path in some cases which could cause a panic.
1998-09-24 22:43:54 +00:00
/*
* Give controllers a chance to renegotiate
* before starting tag operations. We
* "toggle" tagged queuing off then on
* which causes the tag enable command delay
* counter to come into effect.
*/
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
dev = path->device;
if ((dev->flags & CAM_DEV_TAG_AFTER_COUNT) != 0
|| ((dev->inq_flags & SID_CmdQue) != 0
&& (dev->inq_flags & (SID_Sync|SID_WBus16|SID_WBus32)) != 0)) {
Fix a few problems with the tag delay code: - Tagged devices were limited to one transaction (oops) - We revert to untagged with a tag delay if the user changed the transfer negotiation values (via perhaps camcontrol some day). - xpt_async did not use the expanded path in some cases which could cause a panic.
1998-09-24 22:43:54 +00:00
struct ccb_trans_settings cts;
xpt_setup_ccb(&cts.ccb_h, path, 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
cts.protocol = PROTO_SCSI;
cts.protocol_version = PROTO_VERSION_UNSPECIFIED;
cts.transport = XPORT_UNSPECIFIED;
cts.transport_version = XPORT_VERSION_UNSPECIFIED;
cts.proto_specific.scsi.flags = 0;
cts.proto_specific.scsi.valid = CTS_SCSI_VALID_TQ;
Correct problems with xpt_set_transfer_settings and async transfer negotiation changes with wildcarded paths.
1998-09-25 22:35:56 +00:00
xpt_set_transfer_settings(&cts, path->device,
/*async_update*/TRUE);
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
cts.proto_specific.scsi.flags = CTS_SCSI_FLAGS_TAG_ENB;
Correct problems with xpt_set_transfer_settings and async transfer negotiation changes with wildcarded paths.
1998-09-25 22:35:56 +00:00
xpt_set_transfer_settings(&cts, path->device,
/*async_update*/TRUE);
Fix a few problems with the tag delay code: - Tagged devices were limited to one transaction (oops) - We revert to untagged with a tag delay if the user changed the transfer negotiation values (via perhaps camcontrol some day). - xpt_async did not use the expanded path in some cases which could cause a panic.
1998-09-24 22:43:54 +00:00
}
}
static void
xpt_start_tags(struct cam_path *path)
{
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
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;
Restore queue depth settings across tag disable events. The system often disables tag queuing temporarily in order to allow controllers a window to safely perform transfer negotiation with non-compliant devices. Before this change, CAM would restore the queue depth to the controller specified maximum or device quirk level rather than any depth determined by reactions to QUEUE FULL/BUSY events or an explicit user setting. During device probe, initialize the flags field for XPT_SCAN_BUS. The uninitialized value often confused CAM into not bothering to issue an AC_FOUND_DEVICE async event for new devices. The reason this bug wasn't reported earlier is that CAM manually announces devices after the initial system bus scans. MFC: 3 days
2005-03-23 16:43:29 +00:00
if (device->tag_saved_openings != 0)
newopenings = device->tag_saved_openings;
else
newopenings = min(device->quirk->maxtags,
sim->max_tagged_dev_openings);
Fix a few problems with the tag delay code: - Tagged devices were limited to one transaction (oops) - We revert to untagged with a tag delay if the user changed the transfer negotiation values (via perhaps camcontrol some day). - xpt_async did not use the expanded path in some cases which could cause a panic.
1998-09-24 22:43:54 +00:00
xpt_dev_ccbq_resize(path, newopenings);
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
xpt_setup_ccb(&crs.ccb_h, path, /*priority*/1);
crs.ccb_h.func_code = XPT_REL_SIMQ;
crs.release_flags = RELSIM_RELEASE_AFTER_QEMPTY;
crs.openings
style(9)
2008-01-02 01:45:31 +00:00
= crs.release_timeout
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
= crs.qfrozen_cnt
= 0;
xpt_action((union ccb *)&crs);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static int busses_to_config;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
static int busses_to_reset;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static int
xptconfigbuscountfunc(struct cam_eb *bus, void *arg)
{
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(bus->sim->mtx, MA_OWNED);
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
if (bus->path_id != CAM_XPT_PATH_ID) {
struct cam_path path;
struct ccb_pathinq cpi;
int can_negotiate;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
busses_to_config++;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
xpt_compile_path(&path, NULL, bus->path_id,
CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD);
xpt_setup_ccb(&cpi.ccb_h, &path, /*priority*/1);
cpi.ccb_h.func_code = XPT_PATH_INQ;
xpt_action((union ccb *)&cpi);
can_negotiate = cpi.hba_inquiry;
can_negotiate &= (PI_WIDE_32|PI_WIDE_16|PI_SDTR_ABLE);
if ((cpi.hba_misc & PIM_NOBUSRESET) == 0
&& can_negotiate)
busses_to_reset++;
xpt_release_path(&path);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(1);
}
static int
xptconfigfunc(struct cam_eb *bus, void *arg)
{
struct cam_path *path;
union ccb *work_ccb;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_assert(bus->sim->mtx, MA_OWNED);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (bus->path_id != CAM_XPT_PATH_ID) {
cam_status status;
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
int can_negotiate;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Revert a driver API change to xpt_alloc_ccb that isn't necessary. Fix a couple of associated error checks.
2007-04-18 04:58:53 +00:00
work_ccb = xpt_alloc_ccb_nowait();
if (work_ccb == NULL) {
busses_to_config--;
xpt_finishconfig(xpt_periph, NULL);
return(0);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if ((status = xpt_create_path(&path, xpt_periph, bus->path_id,
CAM_TARGET_WILDCARD,
CAM_LUN_WILDCARD)) !=CAM_REQ_CMP){
printf("xptconfigfunc: xpt_create_path failed with "
"status %#x for bus %d\n", status, bus->path_id);
printf("xptconfigfunc: halting bus configuration\n");
xpt_free_ccb(work_ccb);
Expand the hba_misc fied in the Path Inquiry ccb to allow a controller driver to specify that it does not provide initiator services (PIM_NOINITIATOR) and that the initial bus reset for device probing should be avoided (PIM_NOBUSRESET). Modify the XPT layer to honor these flags.
1998-12-10 04:05:49 +00:00
busses_to_config--;
xpt_finishconfig(xpt_periph, NULL);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
return(0);
}
xpt_setup_ccb(&work_ccb->ccb_h, path, /*priority*/1);
Expand the hba_misc fied in the Path Inquiry ccb to allow a controller driver to specify that it does not provide initiator services (PIM_NOINITIATOR) and that the initial bus reset for device probing should be avoided (PIM_NOBUSRESET). Modify the XPT layer to honor these flags.
1998-12-10 04:05:49 +00:00
work_ccb->ccb_h.func_code = XPT_PATH_INQ;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_action(work_ccb);
Expand the hba_misc fied in the Path Inquiry ccb to allow a controller driver to specify that it does not provide initiator services (PIM_NOINITIATOR) and that the initial bus reset for device probing should be avoided (PIM_NOBUSRESET). Modify the XPT layer to honor these flags.
1998-12-10 04:05:49 +00:00
if (work_ccb->ccb_h.status != CAM_REQ_CMP) {
printf("xptconfigfunc: CPI failed on bus %d "
"with status %d\n", bus->path_id,
work_ccb->ccb_h.status);
xpt_finishconfig(xpt_periph, work_ccb);
return(1);
}
Add the XPT_PATH_STATS and XPT_GDEV_STATS function codes. These ccb types allow the reporting of error counts and other statistics. Currently we provide information on the last BDR or bus reset as well as active transaction inforamtion, but this will be expanded as more information is added to aid in error recovery. Use the 'last reset' information to better handle bus settle delays. Peripheral drivers now control whether a bus settle delay occurs and for how long. This allows target mode peripheral drivers to avoid having their device queue frozen by the XPT for what shoudl only be initiator type behavior. Don't perform a bus reset if the target device is incapable of performing transfer negotiation (e.g. Fiber Channel). If we don't perform a bus reset but the controller is capable of transfer negotiations, force negotiations on the first transaction to go to the device. This ensures that we aren't tripped up by a left over negotiation from the prom, BIOS, loader, etc. Add a default async handler funstion to cam_periph.c to remove duplicated code in all initiator type peripheral drivers. Allow mapping of XPT_CONT_TARGET_IO ccbs from userland. They are itentical to XPT_SCSI_IO ccbs as far as data mapping is concerned.
1999-05-22 21:58:47 +00:00
can_negotiate = work_ccb->cpi.hba_inquiry;
can_negotiate &= (PI_WIDE_32|PI_WIDE_16|PI_SDTR_ABLE);
if ((work_ccb->cpi.hba_misc & PIM_NOBUSRESET) == 0
&& (can_negotiate != 0)) {
Expand the hba_misc fied in the Path Inquiry ccb to allow a controller driver to specify that it does not provide initiator services (PIM_NOINITIATOR) and that the initial bus reset for device probing should be avoided (PIM_NOBUSRESET). Modify the XPT layer to honor these flags.
1998-12-10 04:05:49 +00:00
xpt_setup_ccb(&work_ccb->ccb_h, path, /*priority*/1);
work_ccb->ccb_h.func_code = XPT_RESET_BUS;
work_ccb->ccb_h.cbfcnp = NULL;
CAM_DEBUG(path, CAM_DEBUG_SUBTRACE,
("Resetting Bus\n"));
xpt_action(work_ccb);
xpt_finishconfig(xpt_periph, work_ccb);
} else {
/* Act as though we performed a successful BUS RESET */
work_ccb->ccb_h.func_code = XPT_RESET_BUS;
xpt_finishconfig(xpt_periph, work_ccb);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
return(1);
}
static void
xpt_config(void *arg)
{
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
/*
* Now that interrupts are enabled, go find our devices
*/
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
#ifdef CAMDEBUG
/* Setup debugging flags and path */
#ifdef CAM_DEBUG_FLAGS
cam_dflags = CAM_DEBUG_FLAGS;
#else /* !CAM_DEBUG_FLAGS */
cam_dflags = CAM_DEBUG_NONE;
#endif /* CAM_DEBUG_FLAGS */
#ifdef CAM_DEBUG_BUS
if (cam_dflags != CAM_DEBUG_NONE) {
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
/*
* Locking is specifically omitted here. No SIMs have
* registered yet, so xpt_create_path will only be searching
* empty lists of targets and devices.
*/
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (xpt_create_path(&cam_dpath, xpt_periph,
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;
#else /* !CAM_DEBUG_BUS */
cam_dpath = NULL;
#endif /* CAM_DEBUG_BUS */
#endif /* CAMDEBUG */
Protect access to the EDT with splcam() to protect against corruption caused by temporary EDT allocations performed by controller drivers in their interrupt routiens. Reference count bus entries in the EDT in preparation for support for dynamic controller arrival and departure. Have children of the EDT hold references to their parents. Correct routing of the XPT_IMMED_NOTIFY ccb type for use in target mode applications. Fix a few cases where the generation count for EDT data members was not being updated when a modification occurred. splcam() problem Noticed by: Tor Egge <tegge@FreeBSD.org>
1999-03-05 23:18:16 +00:00
/*
* Scan all installed busses.
*/
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_for_all_busses(xptconfigbuscountfunc, NULL);
Only deregister out configuration hook manually if there are no SCSI busses to configure.
1998-09-20 05:03:34 +00:00
if (busses_to_config == 0) {
/* Call manually because we don't have any busses */
xpt_finishconfig(xpt_periph, NULL);
If the bus delay is >= 2 seconds, notify the user that we are waiting for devices to settle. This will hopefully allay any 'first installation' fears that the machine has hung.
1998-10-10 21:10:36 +00:00
} else {
Make SCSI_DELAY setable at boot time and runtime via the kern.cam.scsi_delay tunable/sysctl. Reviewed by: mdodd, njl
2002-09-02 20:10:19 +00:00
if (busses_to_reset > 0 && scsi_delay >= 2000) {
If the bus delay is >= 2 seconds, notify the user that we are waiting for devices to settle. This will hopefully allay any 'first installation' fears that the machine has hung.
1998-10-10 21:10:36 +00:00
printf("Waiting %d seconds for SCSI "
Make SCSI_DELAY setable at boot time and runtime via the kern.cam.scsi_delay tunable/sysctl. Reviewed by: mdodd, njl
2002-09-02 20:10:19 +00:00
"devices to settle\n", scsi_delay/1000);
If the bus delay is >= 2 seconds, notify the user that we are waiting for devices to settle. This will hopefully allay any 'first installation' fears that the machine has hung.
1998-10-10 21:10:36 +00:00
}
Only deregister out configuration hook manually if there are no SCSI busses to configure.
1998-09-20 05:03:34 +00:00
xpt_for_all_busses(xptconfigfunc, NULL);
If the bus delay is >= 2 seconds, notify the user that we are waiting for devices to settle. This will hopefully allay any 'first installation' fears that the machine has hung.
1998-10-10 21:10:36 +00:00
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
/*
* 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
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xpt_finishconfig_task(void *context, int pending)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
struct periph_driver **p_drv;
int i;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
if (busses_to_config == 0) {
/* Register all the peripheral drivers */
/* XXX This will have to change when we have loadable modules */
p_drv = periph_drivers;
for (i = 0; p_drv[i] != NULL; i++) {
(*p_drv[i]->init)();
}
/*
* Check for devices with no "standard" peripheral driver
* attached. For any devices like that, announce the
* passthrough driver so the user will see something.
*/
xpt_for_all_devices(xptpassannouncefunc, NULL);
/* Release our hook so that the boot can continue. */
config_intrhook_disestablish(xsoftc.xpt_config_hook);
Eliminate the use of M_TEMP.
2007-05-14 21:48:53 +00:00
free(xsoftc.xpt_config_hook, M_CAMXPT);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xsoftc.xpt_config_hook = NULL;
}
free(context, M_CAMXPT);
}
static void
xpt_finishconfig(struct cam_periph *periph, union ccb *done_ccb)
{
struct xpt_task *task;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (done_ccb != NULL) {
CAM_DEBUG(done_ccb->ccb_h.path, CAM_DEBUG_TRACE,
("xpt_finishconfig\n"));
switch(done_ccb->ccb_h.func_code) {
case XPT_RESET_BUS:
if (done_ccb->ccb_h.status == CAM_REQ_CMP) {
done_ccb->ccb_h.func_code = XPT_SCAN_BUS;
done_ccb->ccb_h.cbfcnp = xpt_finishconfig;
Restore queue depth settings across tag disable events. The system often disables tag queuing temporarily in order to allow controllers a window to safely perform transfer negotiation with non-compliant devices. Before this change, CAM would restore the queue depth to the controller specified maximum or device quirk level rather than any depth determined by reactions to QUEUE FULL/BUSY events or an explicit user setting. During device probe, initialize the flags field for XPT_SCAN_BUS. The uninitialized value often confused CAM into not bothering to issue an AC_FOUND_DEVICE async event for new devices. The reason this bug wasn't reported earlier is that CAM manually announces devices after the initial system bus scans. MFC: 3 days
2005-03-23 16:43:29 +00:00
done_ccb->crcn.flags = 0;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_action(done_ccb);
return;
}
/* FALLTHROUGH */
case XPT_SCAN_BUS:
Expand the hba_misc fied in the Path Inquiry ccb to allow a controller driver to specify that it does not provide initiator services (PIM_NOINITIATOR) and that the initial bus reset for device probing should be avoided (PIM_NOBUSRESET). Modify the XPT layer to honor these flags.
1998-12-10 04:05:49 +00:00
default:
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
xpt_free_path(done_ccb->ccb_h.path);
busses_to_config--;
break;
}
}
Only schedule the xpt_finishconfig_task once. This fixes some potential panics on boot.
2007-04-27 14:23:05 +00:00
if (busses_to_config == 0) {
task = malloc(sizeof(struct xpt_task), M_CAMXPT, M_NOWAIT);
if (task != NULL) {
TASK_INIT(&task->task, 0, xpt_finishconfig_task, task);
taskqueue_enqueue(taskqueue_thread, &task->task);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if (done_ccb != NULL)
xpt_free_ccb(done_ccb);
}
Add a helper function for registering async callbacks. Besides eliminating a lot of duplicated code, this also fixes a locking edge case.
2007-05-16 16:54:23 +00:00
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) {
mtx_lock(&xsoftc.xpt_lock);
status = xpt_create_path(&path, /*periph*/NULL, CAM_XPT_PATH_ID,
CAM_TARGET_WILDCARD, CAM_LUN_WILDCARD);
if (status != CAM_REQ_CMP) {
mtx_unlock(&xsoftc.xpt_lock);
return (status);
}
xptpath = 1;
}
xpt_setup_ccb(&csa.ccb_h, path, /*priority*/5);
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;
if (xptpath) {
xpt_free_path(path);
mtx_unlock(&xsoftc.xpt_lock);
}
return (status);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
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;
strncpy(cpi->sim_vid, "FreeBSD", SIM_IDLEN);
strncpy(cpi->hba_vid, "", HBA_IDLEN);
strncpy(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;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
cpi->ccb_h.status = CAM_REQ_CMP;
xpt_done(work_ccb);
break;
}
default:
work_ccb->ccb_h.status = CAM_REQ_INVALID;
xpt_done(work_ccb);
break;
}
}
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
/*
* The xpt as a "controller" has no interrupt sources, so polling
* is a no-op.
*/
static void
xptpoll(struct cam_sim *sim)
{
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
void
xpt_lock_buses(void)
{
mtx_lock(&xsoftc.xpt_topo_lock);
}
void
xpt_unlock_buses(void)
{
mtx_unlock(&xsoftc.xpt_topo_lock);
}
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
static void
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
camisr(void *dummy)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
{
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
cam_simq_t queue;
struct cam_sim *sim;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
mtx_lock(&cam_simq_lock);
Provide locking for the ccb_bioq. This allows xpt_done() to be called without Giant held. In camisr(), move the ccb_bioq elements to a temporary local list and then process the elements off of that list. This enables the list to be processed by only taking the ccb_bioq_lock once and only for a very short time. ccb_bioq_lock is a leaf mutex, so it's fine to call xpt_done() with other locks held. This is just a very minor step in the work to lock CAM, but it allows us to avoid some messy locking/unlock dances in certain drivers.
2005-02-09 11:44:15 +00:00
TAILQ_INIT(&queue);
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
TAILQ_CONCAT(&queue, &cam_simq, links);
mtx_unlock(&cam_simq_lock);
while ((sim = TAILQ_FIRST(&queue)) != NULL) {
TAILQ_REMOVE(&queue, sim, links);
Inline cam_periph_lock|unlock to make debugging easier. Use more CAM_SIM_LOCK() more uniformly.
2007-04-19 22:46:26 +00:00
CAM_SIM_LOCK(sim);
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
sim->flags &= ~CAM_SIM_ON_DONEQ;
camisr_runqueue(&sim->sim_doneq);
Inline cam_periph_lock|unlock to make debugging easier. Use more CAM_SIM_LOCK() more uniformly.
2007-04-19 22:46:26 +00:00
CAM_SIM_UNLOCK(sim);
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
}
}
Provide locking for the ccb_bioq. This allows xpt_done() to be called without Giant held. In camisr(), move the ccb_bioq elements to a temporary local list and then process the elements off of that list. This enables the list to be processed by only taking the ccb_bioq_lock once and only for a very short time. ccb_bioq_lock is a leaf mutex, so it's fine to call xpt_done() with other locks held. This is just a very minor step in the work to lock CAM, but it allows us to avoid some messy locking/unlock dances in certain drivers.
2005-02-09 11:44:15 +00:00
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
static void
camisr_runqueue(void *V_queue)
{
cam_isrq_t *queue = V_queue;
struct ccb_hdr *ccb_h;
while ((ccb_h = TAILQ_FIRST(queue)) != NULL) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
int runq;
Split the camisr into per-SIM done queues. This optimizes the locking a little bit and allows for direct dispatch of the doneq from certain contexts that would otherwise face recursive locking problems.
2007-04-19 14:28:43 +00:00
TAILQ_REMOVE(queue, ccb_h, sim_links.tqe);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
ccb_h->pinfo.index = CAM_UNQUEUED_INDEX;
CAM_DEBUG(ccb_h->path, CAM_DEBUG_TRACE,
Add \n to various printfs. grep shows these should be the last ones.
2002-08-27 18:43:14 +00:00
("camisr\n"));
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
runq = FALSE;
if (ccb_h->flags & CAM_HIGH_POWER) {
struct highpowerlist *hphead;
union ccb *send_ccb;
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_lock(&xsoftc.xpt_lock);
hphead = &xsoftc.highpowerq;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
send_ccb = (union ccb *)STAILQ_FIRST(hphead);
/*
* Increment the count since this command is done.
*/
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
xsoftc.num_highpower++;
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
style(9)
2008-01-02 01:45:31 +00:00
/*
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
* Any high powered commands queued up?
*/
if (send_ccb != NULL) {
STAILQ_REMOVE_HEAD(hphead, xpt_links.stqe);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
mtx_unlock(&xsoftc.xpt_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(send_ccb->ccb_h.path,
/*count*/1, /*runqueue*/TRUE);
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
} else
mtx_unlock(&xsoftc.xpt_lock);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
Remove Giant from CAM. Drivers (SIMs) now register a mutex that CAM will use to synchornize and protect all data objects that are used for that SIM. Drivers that are not yet MPSAFE register Giant and operate as usual. RIght now, no drivers are MPSAFE, though a few will be changed in the coming week as this work settles down. The driver API has changed, so all CAM drivers will need to be recompiled. The userland API has not changed, so tools like camcontrol do not need to be recompiled.
2007-04-15 08:49:19 +00:00
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
if ((ccb_h->func_code & XPT_FC_USER_CCB) == 0) {
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
struct cam_ed *dev;
dev = ccb_h->path->device;
cam_ccbq_ccb_done(&dev->ccbq, (union ccb *)ccb_h);
Get rid of dead_sim. There is no way to make it work - any attempt to actually use it would panic on mtx operation, as dead_sim doesn't have a proper mutex. Even if it had a properly initialized mutex, it wouldn't have properly locked and owned one. Reviewed by: scottl Approved by: rwatson (mentor) Sponsored by: FreeBSD Foundation
2008-12-16 16:54:51 +00:00
ccb_h->path->bus->sim->devq->send_active--;
ccb_h->path->bus->sim->devq->send_openings++;
style(9)
2008-01-02 01:45:31 +00:00
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 (((dev->flags & CAM_DEV_REL_ON_COMPLETE) != 0
&& (ccb_h->status&CAM_STATUS_MASK) != CAM_REQUEUE_REQ)
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
|| ((dev->flags & CAM_DEV_REL_ON_QUEUE_EMPTY) != 0
&& (dev->ccbq.dev_active == 0))) {
style(9)
2008-01-02 01:45:31 +00:00
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(ccb_h->path, /*count*/1,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*run_queue*/TRUE);
}
Allow 5 untagged commands to go to a device before enabling tags after enabling transfer negotiations, a BDR, or a bus reset to allow the controller driver to negotiate without tagged messages getting in the way. Some devices are confused by attempts to negotiate and tag at the same time. Some controllers (e.g. BT MultiMaster with certain firmware revs) will never negotiate if you don't give them an untagged "window" to perform negotiation in. Bump the maximum tag count to 255. The system reclaims unused tag space as the tag count is dropped anyway, so we might as well try the max. We should probably use a larger type than u_int8_t to hold our tag value as SCSI over certain mediums allows for higher values. Reviewed by: Kenneth Merry <ken@FreeBSD.org>
1998-09-23 03:03:19 +00:00
if ((dev->flags & CAM_DEV_TAG_AFTER_COUNT) != 0
&& (--dev->tag_delay_count == 0))
xpt_start_tags(ccb_h->path);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
if ((dev->ccbq.queue.entries > 0)
&& (dev->qfrozen_cnt == 0)
&& (device_is_send_queued(dev) == 0)) {
runq = xpt_schedule_dev_sendq(ccb_h->path->bus,
dev);
}
}
if (ccb_h->status & CAM_RELEASE_SIMQ) {
xpt_release_simq(ccb_h->path->bus->sim,
/*run_queue*/TRUE);
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
ccb_h->status &= ~CAM_RELEASE_SIMQ;
runq = FALSE;
style(9)
2008-01-02 01:45:31 +00:00
}
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
if ((ccb_h->flags & CAM_DEV_QFRZDIS)
&& (ccb_h->status & CAM_DEV_QFRZN)) {
Implement the guts of the XPT_ABORT function code. Clear residuals for XPT_TARGET_IO/XPT_CONT_TARGET_IO calls just as we do for XPT_SCSI_IO calls. Make xpt_release_devq a proper reciprocal of xpt_freeze_devq by allowing the caller to decrement the count more than once per call. Change the semantics of xpt_path_cmp to return 1 for a match with a wildcard from path1 and 2 for a match with wildcards from path2. This change doesn't affect most callers as they only check for >=0 (exact match or one containing a wildcard). When cleaning up a device node, ensure that there are no timeouts outstanding that reference it.
1999-08-16 22:22:41 +00:00
xpt_release_devq(ccb_h->path, /*count*/1,
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
/*run_queue*/TRUE);
ccb_h->status &= ~CAM_DEV_QFRZN;
} else if (runq) {
xpt_run_dev_sendq(ccb_h->path->bus);
}
/* Call the peripheral driver's callback */
Clean up the xpt_sim creation to mirror how HBA drivers perform this task. Clean up node creation in the EDT so that initialization follows member declaration. Sort registered paths by pathid so that we probe busses in order of ascending pathid. This makes hardwiring of busses without wiring individual targets do what the user expects. (submitted by tegge@FreeBSD.org) Fix an EDT node leak. Target nodes would never go away. Implement xpt_bus_deregister(). (prodded by some patches from T. Ichinoseki, but implemented differently.)
2000-01-14 23:08:46 +00:00
(*ccb_h->cbfcnp)(ccb_h->path->periph, (union ccb *)ccb_h);
CAM Transport Layer (XPT). Submitted by: The CAM Team
1998-09-15 06:33:23 +00:00
}
}
When deregistering a bus, attempt to flush out all outstanding operations before returning. Point the bus at a dummy cam_sim structure so that any CCBs will complete immediately with a CAM_DEV_NOT_THERE status, and ensure that any xpt_schedule() calls on the bus's devices will immediately call the peripheral's periph_start() routine. Also repeat the async messages because devices that were part of the way through being probed may appear after the original AC_LOST_DEVICE was sent, and would otherwise never go away. These changes make it possible to deregister a bus and free the SIM at most stages during bus probing without the usual crashes in camisr(). In particular, plugging in a umass device and then unplugging it as soon as the first probe messages appeared would almost always result in a crash. Now the device just goes away with a few CAM errors and all references to the CAM bus, target and device are dropped correctly.
2006-01-11 02:06:08 +00:00
Reference in New Issue Copy Permalink