d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
d7a3cd4807
freebsd-skq/sys/dev/aac/aac.c

3049 lines
76 KiB
C
Raw Normal View History

A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
/*-
* Copyright (c) 2000 Michael Smith
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
* Copyright (c) 2001 Scott Long
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Copyright (c) 2000 BSDi
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
* Copyright (c) 2001 Adaptec, Inc.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +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.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* 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(). Also some minor style cleanups.
2003-08-24 17:55:58 +00:00
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
/*
* Driver for the Adaptec 'FSA' family of PCI/SCSI RAID adapters.
*/
Bugfixes. Close a race and logic bug in the timeout handling, don't call the interrupt handler from the upper half, etc. This fixes some serious stability problems that we were seeing on our production server. These patches have been tested for almost 6 months and are a highly recommended MFC candidate. Reviewed by: gibbs, merry, msmith MFC after: 4 days
2001-08-05 06:20:13 +00:00
#include "opt_aac.h"
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
/* #include <stddef.h> */
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/malloc.h>
#include <sys/kernel.h>
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
#include <sys/kthread.h>
Once again throw out the calculation of si_iosize_max and set it to 64K. Although it can go higher, it is not safe to so do on arrays with many members. Compromise by adding a tunable, "hw.aac.iosize_max" that can be set at boottime. Also document in the aac(4) manpage. MFC after: 4 weeks
2001-08-05 07:26:31 +00:00
#include <sys/sysctl.h>
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
#include <sys/poll.h>
Use bioq_flush() to drain a bio queue with a specific error code. Retain the mistake of not updating the devstat API for now. Spell bioq_disksort() consistently with the remaining bioq_*(). #include <geom/geom_disk.h> where this is more appropriate.
2003-04-01 15:06:26 +00:00
#include <sys/ioccom.h>
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
#include <sys/bus.h>
#include <sys/conf.h>
#include <sys/signalvar.h>
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
#include <sys/time.h>
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
#include <sys/eventhandler.h>
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
#include <machine/bus_memio.h>
#include <machine/bus.h>
#include <machine/resource.h>
#include <dev/aac/aacreg.h>
Complete the repo-copy of aac_ioctl.h from sys/dev/aac/to sys/sys.
2004-12-09 22:20:25 +00:00
#include <sys/aac_ioctl.h>
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
#include <dev/aac/aacvar.h>
#include <dev/aac/aac_tables.h>
static void aac_startup(void *arg);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
static void aac_add_container(struct aac_softc *sc,
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
struct aac_mntinforesp *mir, int f);
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
static void aac_get_bus_info(struct aac_softc *sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
/* Command Processing */
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
static void aac_timeout(struct aac_softc *sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
static void aac_complete(void *context, int pending);
static int aac_bio_command(struct aac_softc *sc, struct aac_command **cmp);
static void aac_bio_complete(struct aac_command *cm);
Remove the 'timeout' argument from aac_wait_command() as it isn't used and never will be. Update the XXX comment for this function to accurately reflect why things are the way they are.
2004-04-14 19:11:29 +00:00
static int aac_wait_command(struct aac_command *cm);
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
static void aac_command_thread(struct aac_softc *sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
/* Command Buffer Management */
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
static void aac_map_command_sg(void *arg, bus_dma_segment_t *segs,
int nseg, int error);
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
static void aac_map_command_helper(void *arg, bus_dma_segment_t *segs,
int nseg, int error);
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
static int aac_alloc_commands(struct aac_softc *sc);
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
static void aac_free_commands(struct aac_softc *sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
static void aac_unmap_command(struct aac_command *cm);
/* Hardware Interface */
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
static void aac_common_map(void *arg, bus_dma_segment_t *segs, int nseg,
int error);
Don't attach to Dell PERC2/QC cards that have a firmware rev of 1.x. This check is complicated by the fact that the Adaptec 5400S cards claim to use 1.x firmware also. PERC2/QC 1.x firmware is not compatible with this driver and will cause a system hang. MFC after: 3 days
2002-02-06 01:34:09 +00:00
static int aac_check_firmware(struct aac_softc *sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
static int aac_init(struct aac_softc *sc);
static int aac_sync_command(struct aac_softc *sc, u_int32_t command,
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
u_int32_t arg0, u_int32_t arg1, u_int32_t arg2,
u_int32_t arg3, u_int32_t *sp);
static int aac_enqueue_fib(struct aac_softc *sc, int queue,
Bugfixes. Close a race and logic bug in the timeout handling, don't call the interrupt handler from the upper half, etc. This fixes some serious stability problems that we were seeing on our production server. These patches have been tested for almost 6 months and are a highly recommended MFC candidate. Reviewed by: gibbs, merry, msmith MFC after: 4 days
2001-08-05 06:20:13 +00:00
struct aac_command *cm);
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
static int aac_dequeue_fib(struct aac_softc *sc, int queue,
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
u_int32_t *fib_size, struct aac_fib **fib_addr);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
static int aac_enqueue_response(struct aac_softc *sc, int queue,
struct aac_fib *fib);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
/* Falcon/PPC interface */
static int aac_fa_get_fwstatus(struct aac_softc *sc);
static void aac_fa_qnotify(struct aac_softc *sc, int qbit);
static int aac_fa_get_istatus(struct aac_softc *sc);
static void aac_fa_clear_istatus(struct aac_softc *sc, int mask);
static void aac_fa_set_mailbox(struct aac_softc *sc, u_int32_t command,
u_int32_t arg0, u_int32_t arg1,
u_int32_t arg2, u_int32_t arg3);
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
static int aac_fa_get_mailbox(struct aac_softc *sc, int mb);
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static void aac_fa_set_interrupts(struct aac_softc *sc, int enable);
struct aac_interface aac_fa_interface = {
aac_fa_get_fwstatus,
aac_fa_qnotify,
aac_fa_get_istatus,
aac_fa_clear_istatus,
aac_fa_set_mailbox,
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
aac_fa_get_mailbox,
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
aac_fa_set_interrupts
};
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
/* StrongARM interface */
static int aac_sa_get_fwstatus(struct aac_softc *sc);
static void aac_sa_qnotify(struct aac_softc *sc, int qbit);
static int aac_sa_get_istatus(struct aac_softc *sc);
static void aac_sa_clear_istatus(struct aac_softc *sc, int mask);
static void aac_sa_set_mailbox(struct aac_softc *sc, u_int32_t command,
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
u_int32_t arg0, u_int32_t arg1,
u_int32_t arg2, u_int32_t arg3);
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
static int aac_sa_get_mailbox(struct aac_softc *sc, int mb);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
static void aac_sa_set_interrupts(struct aac_softc *sc, int enable);
struct aac_interface aac_sa_interface = {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
aac_sa_get_fwstatus,
aac_sa_qnotify,
aac_sa_get_istatus,
aac_sa_clear_istatus,
aac_sa_set_mailbox,
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
aac_sa_get_mailbox,
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
aac_sa_set_interrupts
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
};
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* i960Rx interface */
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
static int aac_rx_get_fwstatus(struct aac_softc *sc);
static void aac_rx_qnotify(struct aac_softc *sc, int qbit);
static int aac_rx_get_istatus(struct aac_softc *sc);
static void aac_rx_clear_istatus(struct aac_softc *sc, int mask);
static void aac_rx_set_mailbox(struct aac_softc *sc, u_int32_t command,
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
u_int32_t arg0, u_int32_t arg1,
u_int32_t arg2, u_int32_t arg3);
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
static int aac_rx_get_mailbox(struct aac_softc *sc, int mb);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
static void aac_rx_set_interrupts(struct aac_softc *sc, int enable);
struct aac_interface aac_rx_interface = {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
aac_rx_get_fwstatus,
aac_rx_qnotify,
aac_rx_get_istatus,
aac_rx_clear_istatus,
aac_rx_set_mailbox,
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
aac_rx_get_mailbox,
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
aac_rx_set_interrupts
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
};
Add support for the Adaptec RAID-On-Chip architecture. This in turn provides support for the Adaptec 2130S adapter. Thanks to Adaptec for providing hardware for this.
2004-08-13 01:44:09 +00:00
/* Rocket/MIPS interface */
static int aac_rkt_get_fwstatus(struct aac_softc *sc);
static void aac_rkt_qnotify(struct aac_softc *sc, int qbit);
static int aac_rkt_get_istatus(struct aac_softc *sc);
static void aac_rkt_clear_istatus(struct aac_softc *sc, int mask);
static void aac_rkt_set_mailbox(struct aac_softc *sc, u_int32_t command,
u_int32_t arg0, u_int32_t arg1,
u_int32_t arg2, u_int32_t arg3);
static int aac_rkt_get_mailbox(struct aac_softc *sc, int mb);
static void aac_rkt_set_interrupts(struct aac_softc *sc, int enable);
struct aac_interface aac_rkt_interface = {
aac_rkt_get_fwstatus,
aac_rkt_qnotify,
aac_rkt_get_istatus,
aac_rkt_clear_istatus,
aac_rkt_set_mailbox,
aac_rkt_get_mailbox,
aac_rkt_set_interrupts
};
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
/* Debugging and Diagnostics */
static void aac_describe_controller(struct aac_softc *sc);
Doh! Fix a comma that disappeared along the way. Pointy-hat by: peter
2001-08-30 13:09:09 +00:00
static char *aac_describe_code(struct aac_code_lookup *table,
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
u_int32_t code);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
/* Management Interface */
static d_open_t aac_open;
static d_close_t aac_close;
static d_ioctl_t aac_ioctl;
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static d_poll_t aac_poll;
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
static int aac_ioctl_sendfib(struct aac_softc *sc, caddr_t ufib);
static void aac_handle_aif(struct aac_softc *sc,
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_fib *fib);
Put the linux compat ioctl interface on death row. Add FreeBSD definitions of all of the ioctls, plus round out all ioctl definitions to match what exists for linux. Allow ioctls to be called through either the native or linux interface, though no apps exist (yet) that can take advantage of native calling.
2001-08-06 04:34:47 +00:00
static int aac_rev_check(struct aac_softc *sc, caddr_t udata);
static int aac_getnext_aif(struct aac_softc *sc, caddr_t arg);
static int aac_return_aif(struct aac_softc *sc, caddr_t uptr);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
static int aac_query_disk(struct aac_softc *sc, caddr_t uptr);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
static struct cdevsw aac_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,
.d_flags = D_NEEDGIANT,
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 = aac_open,
.d_close = aac_close,
.d_ioctl = aac_ioctl,
.d_poll = aac_poll,
.d_name = "aac",
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
};
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
MALLOC_DEFINE(M_AACBUF, "aacbuf", "Buffers for the AAC driver");
Once again throw out the calculation of si_iosize_max and set it to 64K. Although it can go higher, it is not safe to so do on arrays with many members. Compromise by adding a tunable, "hw.aac.iosize_max" that can be set at boottime. Also document in the aac(4) manpage. MFC after: 4 weeks
2001-08-05 07:26:31 +00:00
/* sysctl node */
SYSCTL_NODE(_hw, OID_AUTO, aac, CTLFLAG_RD, 0, "AAC driver parameters");
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Device Interface
*/
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Initialise the controller and softc
*/
int
aac_attach(struct aac_softc *sc)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
int error, unit;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Initialise per-controller queues.
*/
aac_initq_free(sc);
aac_initq_ready(sc);
aac_initq_busy(sc);
aac_initq_bio(sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Initialise command-completion task.
*/
TASK_INIT(&sc->aac_task_complete, 0, aac_complete, sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* disable interrupts before we enable anything */
AAC_MASK_INTERRUPTS(sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* mark controller as suspended until we get ourselves organised */
sc->aac_state |= AAC_STATE_SUSPEND;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Don't attach to Dell PERC2/QC cards that have a firmware rev of 1.x. This check is complicated by the fact that the Adaptec 5400S cards claim to use 1.x firmware also. PERC2/QC 1.x firmware is not compatible with this driver and will cause a system hang. MFC after: 3 days
2002-02-06 01:34:09 +00:00
/*
* Check that the firmware on the card is supported.
*/
if ((error = aac_check_firmware(sc)) != 0)
return(error);
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
/*
* Initialize locks
*/
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_init(&sc->aac_aifq_lock, "AAC AIF lock", NULL, MTX_DEF);
mtx_init(&sc->aac_io_lock, "AAC I/O lock", NULL, MTX_DEF);
mtx_init(&sc->aac_container_lock, "AAC container lock", NULL, MTX_DEF);
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
TAILQ_INIT(&sc->aac_container_tqh);
Fix a couple of bugs with AIF handling: - Correct the logic for the AIF array index pointers so that correct slot is always looked at. - Copy the full FIB payload size when copying AIF's, not just the first 64 bytes. Thanks to Mirapoint, Inc, for pointing these problems out and offering a solution.
2003-10-17 21:44:06 +00:00
/* Initialize the local AIF queue pointers */
sc->aac_aifq_head = sc->aac_aifq_tail = AAC_AIFQ_LENGTH;
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Initialise the adapter.
*/
if ((error = aac_init(sc)) != 0)
return(error);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Print a little information about the controller.
*/
aac_describe_controller(sc);
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
/*
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
* Register to probe our containers later.
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
*/
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_ich.ich_func = aac_startup;
sc->aac_ich.ich_arg = sc;
if (config_intrhook_establish(&sc->aac_ich) != 0) {
device_printf(sc->aac_dev,
"can't establish configuration hook\n");
return(ENXIO);
}
/*
* Make the control device.
*/
unit = device_get_unit(sc->aac_dev);
Use OPERATOR instead of WHEEL for make_dev(), use 0640 instead of 0644 for disk devices to be consistent with other storage devices. Submitted by: kris Reviewed by: scottl
2002-12-27 17:52:16 +00:00
sc->aac_dev_t = make_dev(&aac_cdevsw, unit, UID_ROOT, GID_OPERATOR,
0640, "aac%d", unit);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
(void)make_dev_alias(sc->aac_dev_t, "afa%d", unit);
(void)make_dev_alias(sc->aac_dev_t, "hpn%d", unit);
sc->aac_dev_t->si_drv1 = sc;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* Create the AIF thread */
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
if (kthread_create((void(*)(void *))aac_command_thread, sc,
Some kernel threads try to do significant work, and the default KSTACK_PAGES doesn't give them enough stack to do much before blowing away the pcb. This adds MI and MD code to allow the allocation of an alternate kstack who's size can be speficied when calling kthread_create. Passing the value 0 prevents the alternate kstack from being created. Note that the ia64 MD code is missing for now, and PowerPC was only partially written due to the pmap.c being incomplete there. Though this patch does not modify anything to make use of the alternate kstack, acpi and usb are good candidates. Reviewed by: jake, peter, jhb
2002-10-02 07:44:29 +00:00
&sc->aifthread, 0, 0, "aac%daif", unit))
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
panic("Could not create AIF thread\n");
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* Register the shutdown method to only be called post-dump */
Don't include aac_cam.h, since it was nuked in the last commit. Deregister the shutdown eventhandler on unload.
2003-02-06 00:07:12 +00:00
if ((sc->eh = EVENTHANDLER_REGISTER(shutdown_final, aac_shutdown,
sc->aac_dev, SHUTDOWN_PRI_DEFAULT)) == NULL)
device_printf(sc->aac_dev,
"shutdown event registration failed\n");
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
/* Register with CAM for the non-DASD devices */
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
if ((sc->flags & AAC_FLAGS_ENABLE_CAM) != 0) {
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
TAILQ_INIT(&sc->aac_sim_tqh);
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
aac_get_bus_info(sc);
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
}
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(0);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Probe for containers, create disks.
*/
static void
aac_startup(void *arg)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
struct aac_fib *fib;
struct aac_mntinfo *mi;
struct aac_mntinforesp *mir = NULL;
When scanning for changed containers, don't assume that the controller will respond in a sane manner. Thanks to Petri Helenius <pete@he.iki.fi> for spotting this and pestering me to fix it.
2003-06-03 02:10:55 +00:00
int count = 0, i = 0;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc = (struct aac_softc *)arg;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* disconnect ourselves from the intrhook chain */
config_intrhook_disestablish(&sc->aac_ich);
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
Collapse sync fib locking into normal i/o locking. The former didn't protect the registers so it was trivially possible for a sync command and i/o command to fight each other and confuse the controller. Make the sync fib alloc/release functions inline and remove the somewhat worthless AAC_SYNC_LOCK_FORCE flag. Thanks to Adil Katchi for helping me to track this down in RELENG_4.
2004-06-02 18:15:48 +00:00
aac_alloc_sync_fib(sc, &fib);
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
mi = (struct aac_mntinfo *)&fib->data[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* loop over possible containers */
do {
/* request information on this container */
bzero out the sync command buffer when sending commands. This was causing problems when enumerating multiple arrays. This is an MFC candidate.
2002-05-03 00:07:50 +00:00
bzero(mi, sizeof(struct aac_mntinfo));
mi->Command = VM_NameServe;
mi->MntType = FT_FILESYS;
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
mi->MntCount = i;
if (aac_sync_fib(sc, ContainerCommand, 0, fib,
sizeof(struct aac_mntinfo))) {
When scanning for changed containers, don't assume that the controller will respond in a sane manner. Thanks to Petri Helenius <pete@he.iki.fi> for spotting this and pestering me to fix it.
2003-06-03 02:10:55 +00:00
printf("error probing container %d", i);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
continue;
}
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
mir = (struct aac_mntinforesp *)&fib->data[0];
When scanning for changed containers, don't assume that the controller will respond in a sane manner. Thanks to Petri Helenius <pete@he.iki.fi> for spotting this and pestering me to fix it.
2003-06-03 02:10:55 +00:00
/* XXX Need to check if count changed */
count = mir->MntRespCount;
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
aac_add_container(sc, mir, 0);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
i++;
When scanning for changed containers, don't assume that the controller will respond in a sane manner. Thanks to Petri Helenius <pete@he.iki.fi> for spotting this and pestering me to fix it.
2003-06-03 02:10:55 +00:00
} while ((i < count) && (i < AAC_MAX_CONTAINERS));
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
aac_release_sync_fib(sc);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* poke the bus to actually attach the child devices */
if (bus_generic_attach(sc->aac_dev))
device_printf(sc->aac_dev, "bus_generic_attach failed\n");
/* mark the controller up */
sc->aac_state &= ~AAC_STATE_SUSPEND;
/* enable interrupts now */
AAC_UNMASK_INTERRUPTS(sc);
}
/*
* Create a device to respresent a new container
*/
static void
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
aac_add_container(struct aac_softc *sc, struct aac_mntinforesp *mir, int f)
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
{
struct aac_container *co;
device_t child;
/*
* Check container volume type for validity. Note that many of
* the possible types may never show up.
*/
if ((mir->Status == ST_OK) && (mir->MntTable[0].VolType != CT_NONE)) {
Free the container and sim objects on detach. Convert several uses of MALLOC to malloc.
2003-02-06 00:37:37 +00:00
co = (struct aac_container *)malloc(sizeof *co, M_AACBUF,
M_NOWAIT | M_ZERO);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (co == NULL)
panic("Out of memory?!\n");
debug(1, "id %x name '%.16s' size %u type %d",
mir->MntTable[0].ObjectId,
mir->MntTable[0].FileSystemName,
mir->MntTable[0].Capacity, mir->MntTable[0].VolType);
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
if ((child = device_add_child(sc->aac_dev, "aacd", -1)) == NULL)
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
device_printf(sc->aac_dev, "device_add_child failed\n");
else
device_set_ivars(child, co);
device_set_desc(child, aac_describe_code(aac_container_types,
mir->MntTable[0].VolType));
co->co_disk = child;
co->co_found = f;
bcopy(&mir->MntTable[0], &co->co_mntobj,
sizeof(struct aac_mntobj));
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_container_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
TAILQ_INSERT_TAIL(&sc->aac_container_tqh, co, co_link);
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_container_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Free all of the resources associated with (sc)
*
* Should not be called if the controller is active.
*/
void
aac_free(struct aac_softc *sc)
{
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
/* remove the control device */
if (sc->aac_dev_t != NULL)
destroy_dev(sc->aac_dev_t);
/* throw away any FIB buffers, discard the FIB DMA tag */
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
aac_free_commands(sc);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (sc->aac_fib_dmat)
bus_dma_tag_destroy(sc->aac_fib_dmat);
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
free(sc->aac_commands, M_AACBUF);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* destroy the common area */
if (sc->aac_common) {
bus_dmamap_unload(sc->aac_common_dmat, sc->aac_common_dmamap);
bus_dmamem_free(sc->aac_common_dmat, sc->aac_common,
sc->aac_common_dmamap);
}
if (sc->aac_common_dmat)
bus_dma_tag_destroy(sc->aac_common_dmat);
/* disconnect the interrupt handler */
if (sc->aac_intr)
bus_teardown_intr(sc->aac_dev, sc->aac_irq, sc->aac_intr);
if (sc->aac_irq != NULL)
bus_release_resource(sc->aac_dev, SYS_RES_IRQ, sc->aac_irq_rid,
sc->aac_irq);
/* destroy data-transfer DMA tag */
if (sc->aac_buffer_dmat)
bus_dma_tag_destroy(sc->aac_buffer_dmat);
/* destroy the parent DMA tag */
if (sc->aac_parent_dmat)
bus_dma_tag_destroy(sc->aac_parent_dmat);
/* release the register window mapping */
if (sc->aac_regs_resource != NULL)
bus_release_resource(sc->aac_dev, SYS_RES_MEMORY,
sc->aac_regs_rid, sc->aac_regs_resource);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Disconnect from the controller completely, in preparation for unload.
*/
int
aac_detach(device_t dev)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
struct aac_container *co;
struct aac_sim *sim;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
int error;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
sc = device_get_softc(dev);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (sc->aac_state & AAC_STATE_OPEN)
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
return(EBUSY);
/* Remove the child containers */
Free the container and sim objects on detach. Convert several uses of MALLOC to malloc.
2003-02-06 00:37:37 +00:00
while ((co = TAILQ_FIRST(&sc->aac_container_tqh)) != NULL) {
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
error = device_delete_child(dev, co->co_disk);
if (error)
return (error);
On detach, don't remove the child from our list of children unless it has successfully detached.
2003-02-20 23:13:48 +00:00
TAILQ_REMOVE(&sc->aac_container_tqh, co, co_link);
Free the container and sim objects on detach. Convert several uses of MALLOC to malloc.
2003-02-06 00:37:37 +00:00
free(co, M_AACBUF);
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
}
/* Remove the CAM SIMs */
Free the container and sim objects on detach. Convert several uses of MALLOC to malloc.
2003-02-06 00:37:37 +00:00
while ((sim = TAILQ_FIRST(&sc->aac_sim_tqh)) != NULL) {
TAILQ_REMOVE(&sc->aac_sim_tqh, sim, sim_link);
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
error = device_delete_child(dev, sim->sim_dev);
if (error)
return (error);
Free the container and sim objects on detach. Convert several uses of MALLOC to malloc.
2003-02-06 00:37:37 +00:00
free(sim, M_AACBUF);
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (sc->aifflags & AAC_AIFFLAGS_RUNNING) {
sc->aifflags |= AAC_AIFFLAGS_EXIT;
wakeup(sc->aifthread);
tsleep(sc->aac_dev, PUSER | PCATCH, "aacdch", 30 * hz);
}
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (sc->aifflags & AAC_AIFFLAGS_RUNNING)
panic("Cannot shutdown AIF thread\n");
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if ((error = aac_shutdown(dev)))
return(error);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Don't include aac_cam.h, since it was nuked in the last commit. Deregister the shutdown eventhandler on unload.
2003-02-06 00:07:12 +00:00
EVENTHANDLER_DEREGISTER(shutdown_final, sc->eh);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
aac_free(sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Destroy mutexes on detach.
2004-08-12 05:31:17 +00:00
mtx_destroy(&sc->aac_aifq_lock);
mtx_destroy(&sc->aac_io_lock);
mtx_destroy(&sc->aac_container_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(0);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Bring the controller down to a dormant state and detach all child devices.
*
* This function is called before detach or system shutdown.
*
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
* Note that we can assume that the bioq on the controller is empty, as we won't
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* allow shutdown if any device is open.
*/
int
aac_shutdown(device_t dev)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
struct aac_fib *fib;
struct aac_close_command *cc;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc = device_get_softc(dev);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_state |= AAC_STATE_SUSPEND;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Send a Container shutdown followed by a HostShutdown FIB to the
* controller to convince it that we don't want to talk to it anymore.
* We've been closed and all I/O completed already
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
*/
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
device_printf(sc->aac_dev, "shutting down controller...");
Collapse sync fib locking into normal i/o locking. The former didn't protect the registers so it was trivially possible for a sync command and i/o command to fight each other and confuse the controller. Make the sync fib alloc/release functions inline and remove the somewhat worthless AAC_SYNC_LOCK_FORCE flag. Thanks to Adil Katchi for helping me to track this down in RELENG_4.
2004-06-02 18:15:48 +00:00
aac_alloc_sync_fib(sc, &fib);
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
cc = (struct aac_close_command *)&fib->data[0];
bzero out the sync command buffer when sending commands. This was causing problems when enumerating multiple arrays. This is an MFC candidate.
2002-05-03 00:07:50 +00:00
bzero(cc, sizeof(struct aac_close_command));
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
cc->Command = VM_CloseAll;
cc->ContainerId = 0xffffffff;
if (aac_sync_fib(sc, ContainerCommand, 0, fib,
sizeof(struct aac_close_command)))
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
printf("FAILED.\n");
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
else
printf("done\n");
#if 0
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
else {
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
fib->data[0] = 0;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* XXX Issuing this command to the controller makes it shut down
* but also keeps it from coming back up without a reset of the
* PCI bus. This is not desirable if you are just unloading the
* driver module with the intent to reload it later.
*/
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
if (aac_sync_fib(sc, FsaHostShutdown, AAC_FIBSTATE_SHUTDOWN,
fib, 1)) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
printf("FAILED.\n");
} else {
printf("done.\n");
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
#endif
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
AAC_MASK_INTERRUPTS(sc);
Release the sync fib after the controller has been shut down. This also releases the I/O lock instead of just leaking it.
2004-08-12 04:47:32 +00:00
aac_release_sync_fib(sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(0);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Bring the controller to a quiescent state, ready for system suspend.
*/
int
aac_suspend(device_t dev)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc = device_get_softc(dev);
sc->aac_state |= AAC_STATE_SUSPEND;
AAC_MASK_INTERRUPTS(sc);
return(0);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Bring the controller back to a state ready for operation.
*/
int
aac_resume(device_t dev)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
sc = device_get_softc(dev);
sc->aac_state &= ~AAC_STATE_SUSPEND;
AAC_UNMASK_INTERRUPTS(sc);
return(0);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Take an interrupt.
*/
void
aac_intr(void *arg)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
u_int16_t reason;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc = (struct aac_softc *)arg;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
/*
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
* Read the status register directly. This is faster than taking the
* driver lock and reading the queues directly. It also saves having
* to turn parts of the driver lock into a spin mutex, which would be
* ugly.
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
*/
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
reason = AAC_GET_ISTATUS(sc);
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
AAC_CLEAR_ISTATUS(sc, reason);
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
/* handle completion processing */
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
if (reason & AAC_DB_RESPONSE_READY)
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
taskqueue_enqueue_fast(taskqueue_fast, &sc->aac_task_complete);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
/* controller wants to talk to us */
if (reason & (AAC_DB_PRINTF | AAC_DB_COMMAND_READY)) {
/*
* XXX Make sure that we don't get fooled by strange messages
* that start with a NULL.
*/
if ((reason & AAC_DB_PRINTF) &&
(sc->aac_common->ac_printf[0] == 0))
sc->aac_common->ac_printf[0] = 32;
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
/*
* This might miss doing the actual wakeup. However, the
- Broaden the scope of locking in aac_command_thread() again to catch some edge cases in the loop. - Try to grab a command before dequeueing the bio from the bioq. The old behaviour of requeuing deferred bios to the end of the bioq is arguably wrong. This should be fixed in the future to check the bioq head without automatically dequeueing the bio.
2004-02-07 03:26:38 +00:00
* msleep that this is waking up has a timeout, so it will
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
* wake up eventually. AIFs and printfs are low enough
* priority that they can handle hanging out for a few seconds
* if needed.
*/
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
wakeup(sc->aifthread);
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
}
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Command Processing
*/
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Start as much queued I/O as possible on the controller
*/
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
void
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
aac_startio(struct aac_softc *sc)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_command *cm;
Collapse aac_map_command() into aac_startio(). Check the AAC_QUEUE_FRZN in every iteration of aac_startio(). This ensures that a command that is deferred for lack of resources doesn't immediately get retried in the aac_startio() loop. This avoids an almost certain livelock.
2004-06-01 05:32:26 +00:00
int error;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
for (;;) {
Collapse aac_map_command() into aac_startio(). Check the AAC_QUEUE_FRZN in every iteration of aac_startio(). This ensures that a command that is deferred for lack of resources doesn't immediately get retried in the aac_startio() loop. This avoids an almost certain livelock.
2004-06-01 05:32:26 +00:00
/*
* This flag might be set if the card is out of resources.
* Checking it here prevents an infinite loop of deferrals.
*/
if (sc->flags & AAC_QUEUE_FRZN)
break;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Try to get a command that's been put off for lack of
* resources
*/
cm = aac_dequeue_ready(sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Try to build a command off the bio queue (ignore error
* return)
*/
if (cm == NULL)
aac_bio_command(sc, &cm);
/* nothing to do? */
if (cm == NULL)
break;
Collapse aac_map_command() into aac_startio(). Check the AAC_QUEUE_FRZN in every iteration of aac_startio(). This ensures that a command that is deferred for lack of resources doesn't immediately get retried in the aac_startio() loop. This avoids an almost certain livelock.
2004-06-01 05:32:26 +00:00
/* don't map more than once */
if (cm->cm_flags & AAC_CMD_MAPPED)
panic("aac: command %p already mapped", cm);
If a command has to be deferred because there are no more resources for it on the card, unmap it first. This allows it to be picked up properly when the queue gets kicked again. This was the root problem for the lost command (i.e. stuck in getblk/vinvalb) problem. While here, panic if commands don't map correctly instead of just silently ignoring the problem and dropping command. Also slow down the dynamic allocation of new commands. It should be safe to go back into the aac waters. Thanks to everyone who suffered through this and provided good feedback.
2004-02-07 10:30:22 +00:00
/*
Collapse aac_map_command() into aac_startio(). Check the AAC_QUEUE_FRZN in every iteration of aac_startio(). This ensures that a command that is deferred for lack of resources doesn't immediately get retried in the aac_startio() loop. This avoids an almost certain livelock.
2004-06-01 05:32:26 +00:00
* Set up the command to go to the controller. If there are no
* data buffers associated with the command then it can bypass
* busdma.
If a command has to be deferred because there are no more resources for it on the card, unmap it first. This allows it to be picked up properly when the queue gets kicked again. This was the root problem for the lost command (i.e. stuck in getblk/vinvalb) problem. While here, panic if commands don't map correctly instead of just silently ignoring the problem and dropping command. Also slow down the dynamic allocation of new commands. It should be safe to go back into the aac waters. Thanks to everyone who suffered through this and provided good feedback.
2004-02-07 10:30:22 +00:00
*/
Collapse aac_map_command() into aac_startio(). Check the AAC_QUEUE_FRZN in every iteration of aac_startio(). This ensures that a command that is deferred for lack of resources doesn't immediately get retried in the aac_startio() loop. This avoids an almost certain livelock.
2004-06-01 05:32:26 +00:00
if (cm->cm_datalen != 0) {
error = bus_dmamap_load(sc->aac_buffer_dmat,
cm->cm_datamap, cm->cm_data,
cm->cm_datalen,
aac_map_command_sg, cm, 0);
if (error == EINPROGRESS) {
debug(1, "freezing queue\n");
sc->flags |= AAC_QUEUE_FRZN;
error = 0;
Commit the correct version of the patch from last night. This fixes an immediate panic when doing any i/o, and it closes a completion race.
2004-06-01 15:50:11 +00:00
} else if (error != 0)
Collapse aac_map_command() into aac_startio(). Check the AAC_QUEUE_FRZN in every iteration of aac_startio(). This ensures that a command that is deferred for lack of resources doesn't immediately get retried in the aac_startio() loop. This avoids an almost certain livelock.
2004-06-01 05:32:26 +00:00
panic("aac_startio: unexpected error %d from "
"busdma\n", error);
} else
aac_map_command_sg(cm, NULL, 0, 0);
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Handle notification of one or more FIBs coming from the controller.
*/
static void
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
aac_command_thread(struct aac_softc *sc)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_fib *fib;
u_int32_t fib_size;
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
int size, retval;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_io_lock);
- Broaden the scope of locking in aac_command_thread() again to catch some edge cases in the loop. - Try to grab a command before dequeueing the bio from the bioq. The old behaviour of requeuing deferred bios to the end of the bioq is arguably wrong. This should be fixed in the future to check the bioq head without automatically dequeueing the bio.
2004-02-07 03:26:38 +00:00
sc->aifflags = AAC_AIFFLAGS_RUNNING;
while ((sc->aifflags & AAC_AIFFLAGS_EXIT) == 0) {
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
- Broaden the scope of locking in aac_command_thread() again to catch some edge cases in the loop. - Try to grab a command before dequeueing the bio from the bioq. The old behaviour of requeuing deferred bios to the end of the bioq is arguably wrong. This should be fixed in the future to check the bioq head without automatically dequeueing the bio.
2004-02-07 03:26:38 +00:00
retval = 0;
if ((sc->aifflags & AAC_AIFFLAGS_PENDING) == 0)
retval = msleep(sc->aifthread, &sc->aac_io_lock, PRIBIO,
"aifthd", AAC_PERIODIC_INTERVAL * hz);
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
/*
* First see if any FIBs need to be allocated. This needs
* to be called without the driver lock because contigmalloc
* will grab Giant, and would result in an LOR.
*/
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
if ((sc->aifflags & AAC_AIFFLAGS_ALLOCFIBS) != 0) {
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_io_lock);
- Broaden the scope of locking in aac_command_thread() again to catch some edge cases in the loop. - Try to grab a command before dequeueing the bio from the bioq. The old behaviour of requeuing deferred bios to the end of the bioq is arguably wrong. This should be fixed in the future to check the bioq head without automatically dequeueing the bio.
2004-02-07 03:26:38 +00:00
aac_alloc_commands(sc);
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_io_lock);
If a command has to be deferred because there are no more resources for it on the card, unmap it first. This allows it to be picked up properly when the queue gets kicked again. This was the root problem for the lost command (i.e. stuck in getblk/vinvalb) problem. While here, panic if commands don't map correctly instead of just silently ignoring the problem and dropping command. Also slow down the dynamic allocation of new commands. It should be safe to go back into the aac waters. Thanks to everyone who suffered through this and provided good feedback.
2004-02-07 10:30:22 +00:00
sc->aifflags &= ~AAC_AIFFLAGS_ALLOCFIBS;
- Broaden the scope of locking in aac_command_thread() again to catch some edge cases in the loop. - Try to grab a command before dequeueing the bio from the bioq. The old behaviour of requeuing deferred bios to the end of the bioq is arguably wrong. This should be fixed in the future to check the bioq head without automatically dequeueing the bio.
2004-02-07 03:26:38 +00:00
aac_startio(sc);
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
}
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Take the plunge and make this driver be INTR_FAST. This re-arranges the interrupt handler so that no locks are needed, and schedules the command completion routine with a taskqueue_fast. This also corrects the locking in the command thread and removes the need for operation flags. Simple load tests show that this is now considerably faster than FreeBSD 4.x in the SMP case when multiple i/o tasks are running.
2004-01-30 07:04:39 +00:00
/*
* While we're here, check to see if any commands are stuck.
* This is pretty low-priority, so it's ok if it doesn't
* always fire.
*/
if (retval == EWOULDBLOCK)
aac_timeout(sc);
/* Check the hardware printf message buffer */
if (sc->aac_common->ac_printf[0] != 0)
aac_print_printf(sc);
/* Also check to see if the adapter has a command for us. */
while (aac_dequeue_fib(sc, AAC_HOST_NORM_CMD_QUEUE,
&fib_size, &fib) == 0) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
AAC_PRINT_FIB(sc, fib);
switch (fib->Header.Command) {
case AifRequest:
aac_handle_aif(sc, fib);
break;
default:
device_printf(sc->aac_dev, "unknown command "
"from controller\n");
break;
}
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if ((fib->Header.XferState == 0) ||
(fib->Header.StructType != AAC_FIBTYPE_TFIB))
break;
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
/* Return the AIF to the controller. */
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (fib->Header.XferState & AAC_FIBSTATE_FROMADAP) {
fib->Header.XferState |= AAC_FIBSTATE_DONEHOST;
*(AAC_FSAStatus*)fib->data = ST_OK;
/* XXX Compute the Size field? */
size = fib->Header.Size;
if (size > sizeof(struct aac_fib)) {
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
size = sizeof(struct aac_fib);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
fib->Header.Size = size;
}
/*
* Since we did not generate this command, it
* cannot go through the normal
* enqueue->startio chain.
*/
aac_enqueue_response(sc,
AAC_ADAP_NORM_RESP_QUEUE,
fib);
}
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
sc->aifflags &= ~AAC_AIFFLAGS_RUNNING;
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_io_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
wakeup(sc->aac_dev);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
kthread_exit(0);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Make the aac driver be INTR_MPSAFE. Once the interrupt handler determines that a command completion happened, all further processing is deferred to a taskqueue. The taskqueue itself runs implicetely under Giant, but we already used a taskqueue for the biodone() processing, so this at least saves the contesting of Giant in the interrupt handler.
2003-02-19 21:58:34 +00:00
* Process completed commands.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*/
static void
Make the aac driver be INTR_MPSAFE. Once the interrupt handler determines that a command completion happened, all further processing is deferred to a taskqueue. The taskqueue itself runs implicetely under Giant, but we already used a taskqueue for the biodone() processing, so this at least saves the contesting of Giant in the interrupt handler.
2003-02-19 21:58:34 +00:00
aac_complete(void *context, int pending)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Make the aac driver be INTR_MPSAFE. Once the interrupt handler determines that a command completion happened, all further processing is deferred to a taskqueue. The taskqueue itself runs implicetely under Giant, but we already used a taskqueue for the biodone() processing, so this at least saves the contesting of Giant in the interrupt handler.
2003-02-19 21:58:34 +00:00
struct aac_softc *sc;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_command *cm;
struct aac_fib *fib;
u_int32_t fib_size;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Make the aac driver be INTR_MPSAFE. Once the interrupt handler determines that a command completion happened, all further processing is deferred to a taskqueue. The taskqueue itself runs implicetely under Giant, but we already used a taskqueue for the biodone() processing, so this at least saves the contesting of Giant in the interrupt handler.
2003-02-19 21:58:34 +00:00
sc = (struct aac_softc *)context;
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_io_lock);
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
Make the aac driver be INTR_MPSAFE. Once the interrupt handler determines that a command completion happened, all further processing is deferred to a taskqueue. The taskqueue itself runs implicetely under Giant, but we already used a taskqueue for the biodone() processing, so this at least saves the contesting of Giant in the interrupt handler.
2003-02-19 21:58:34 +00:00
/* pull completed commands off the queue */
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
for (;;) {
/* look for completed FIBs on our queue */
if (aac_dequeue_fib(sc, AAC_HOST_NORM_RESP_QUEUE, &fib_size,
&fib))
break; /* nothing to do */
Make the aac driver be INTR_MPSAFE. Once the interrupt handler determines that a command completion happened, all further processing is deferred to a taskqueue. The taskqueue itself runs implicetely under Giant, but we already used a taskqueue for the biodone() processing, so this at least saves the contesting of Giant in the interrupt handler.
2003-02-19 21:58:34 +00:00
Do some small cleanups to comments and remove AACQ_COMPLETE definitions since the completion queue is long-gone.
2004-02-07 17:40:38 +00:00
/* get the command, unmap and hand off for processing */
Fix a 64-bit bogon. The hardware command structure only has one 32 bit field for holding driver-dependant data. Instead of putting the pointer to the driver command struct in there, take advantage of these structs being a (virtually) contiguous array and just put the array index in the field.
2003-02-19 23:33:56 +00:00
cm = sc->aac_commands + fib->Header.SenderData;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (cm == NULL) {
AAC_PRINT_FIB(sc, fib);
Make the aac driver be INTR_MPSAFE. Once the interrupt handler determines that a command completion happened, all further processing is deferred to a taskqueue. The taskqueue itself runs implicetely under Giant, but we already used a taskqueue for the biodone() processing, so this at least saves the contesting of Giant in the interrupt handler.
2003-02-19 21:58:34 +00:00
break;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Make the aac driver be INTR_MPSAFE. Once the interrupt handler determines that a command completion happened, all further processing is deferred to a taskqueue. The taskqueue itself runs implicetely under Giant, but we already used a taskqueue for the biodone() processing, so this at least saves the contesting of Giant in the interrupt handler.
2003-02-19 21:58:34 +00:00
aac_remove_busy(cm);
Do some small cleanups to comments and remove AACQ_COMPLETE definitions since the completion queue is long-gone.
2004-02-07 17:40:38 +00:00
aac_unmap_command(cm);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
cm->cm_flags |= AAC_CMD_COMPLETED;
/* is there a completion handler? */
if (cm->cm_complete != NULL) {
cm->cm_complete(cm);
} else {
/* assume that someone is sleeping on this command */
wakeup(cm);
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* see if we can start some more I/O */
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
sc->flags &= ~AAC_QUEUE_FRZN;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
aac_startio(sc);
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_io_lock);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Handle a bio submitted from a disk device.
*/
void
aac_submit_bio(struct bio *bp)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_disk *ad;
struct aac_softc *sc;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Move to 'struct disk*' API Submitted by: phk
2003-02-26 02:50:27 +00:00
ad = (struct aac_disk *)bp->bio_disk->d_drv1;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc = ad->ad_controller;
/* queue the BIO and try to get some work done */
aac_enqueue_bio(sc, bp);
aac_startio(sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Get a bio and build a command to go with it.
*/
static int
aac_bio_command(struct aac_softc *sc, struct aac_command **cmp)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_command *cm;
struct aac_fib *fib;
struct aac_disk *ad;
struct bio *bp;
debug_called(2);
/* get the resources we will need */
cm = NULL;
- Broaden the scope of locking in aac_command_thread() again to catch some edge cases in the loop. - Try to grab a command before dequeueing the bio from the bioq. The old behaviour of requeuing deferred bios to the end of the bioq is arguably wrong. This should be fixed in the future to check the bioq head without automatically dequeueing the bio.
2004-02-07 03:26:38 +00:00
bp = NULL;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (aac_alloc_command(sc, &cm)) /* get a command */
goto fail;
- Broaden the scope of locking in aac_command_thread() again to catch some edge cases in the loop. - Try to grab a command before dequeueing the bio from the bioq. The old behaviour of requeuing deferred bios to the end of the bioq is arguably wrong. This should be fixed in the future to check the bioq head without automatically dequeueing the bio.
2004-02-07 03:26:38 +00:00
if ((bp = aac_dequeue_bio(sc)) == NULL)
goto fail;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* fill out the command */
cm->cm_data = (void *)bp->bio_data;
cm->cm_datalen = bp->bio_bcount;
cm->cm_complete = aac_bio_complete;
cm->cm_private = bp;
cm->cm_timestamp = time_second;
cm->cm_queue = AAC_ADAP_NORM_CMD_QUEUE;
/* build the FIB */
fib = cm->cm_fib;
Add the ability to send 64-bit scatter/gather elements to aac cards. This is enabled when both the size of bus_addr_t > 4 and the card claims support. Don't wake up the kthread to allocate more commands if we know that we've already allocated the max number of commands.
2003-03-30 21:47:16 +00:00
fib->Header.Size = sizeof(struct aac_fib_header);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
fib->Header.XferState =
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
AAC_FIBSTATE_HOSTOWNED |
AAC_FIBSTATE_INITIALISED |
AAC_FIBSTATE_EMPTY |
AAC_FIBSTATE_FROMHOST |
AAC_FIBSTATE_REXPECTED |
AAC_FIBSTATE_NORM |
AAC_FIBSTATE_ASYNC |
AAC_FIBSTATE_FAST_RESPONSE;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* build the read/write request */
Move to 'struct disk*' API Submitted by: phk
2003-02-26 02:50:27 +00:00
ad = (struct aac_disk *)bp->bio_disk->d_drv1;
Add the ability to send 64-bit scatter/gather elements to aac cards. This is enabled when both the size of bus_addr_t > 4 and the card claims support. Don't wake up the kthread to allocate more commands if we know that we've already allocated the max number of commands.
2003-03-30 21:47:16 +00:00
if ((sc->flags & AAC_FLAGS_SG_64BIT) == 0) {
fib->Header.Command = ContainerCommand;
if (bp->bio_cmd == BIO_READ) {
struct aac_blockread *br;
br = (struct aac_blockread *)&fib->data[0];
br->Command = VM_CtBlockRead;
br->ContainerId = ad->ad_container->co_mntobj.ObjectId;
br->BlockNumber = bp->bio_pblkno;
br->ByteCount = bp->bio_bcount;
fib->Header.Size += sizeof(struct aac_blockread);
cm->cm_sgtable = &br->SgMap;
cm->cm_flags |= AAC_CMD_DATAIN;
} else {
struct aac_blockwrite *bw;
bw = (struct aac_blockwrite *)&fib->data[0];
bw->Command = VM_CtBlockWrite;
bw->ContainerId = ad->ad_container->co_mntobj.ObjectId;
bw->BlockNumber = bp->bio_pblkno;
bw->ByteCount = bp->bio_bcount;
bw->Stable = CUNSTABLE;
fib->Header.Size += sizeof(struct aac_blockwrite);
cm->cm_flags |= AAC_CMD_DATAOUT;
cm->cm_sgtable = &bw->SgMap;
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
} else {
Add the ability to send 64-bit scatter/gather elements to aac cards. This is enabled when both the size of bus_addr_t > 4 and the card claims support. Don't wake up the kthread to allocate more commands if we know that we've already allocated the max number of commands.
2003-03-30 21:47:16 +00:00
fib->Header.Command = ContainerCommand64;
if (bp->bio_cmd == BIO_READ) {
struct aac_blockread64 *br;
br = (struct aac_blockread64 *)&fib->data[0];
br->Command = VM_CtHostRead64;
br->ContainerId = ad->ad_container->co_mntobj.ObjectId;
br->SectorCount = bp->bio_bcount / AAC_BLOCK_SIZE;
br->BlockNumber = bp->bio_pblkno;
br->Pad = 0;
br->Flags = 0;
fib->Header.Size += sizeof(struct aac_blockread64);
cm->cm_flags |= AAC_CMD_DATAOUT;
Avoid casts as lvalues.
2004-07-28 06:21:53 +00:00
cm->cm_sgtable = (struct aac_sg_table *)&br->SgMap64;
Add the ability to send 64-bit scatter/gather elements to aac cards. This is enabled when both the size of bus_addr_t > 4 and the card claims support. Don't wake up the kthread to allocate more commands if we know that we've already allocated the max number of commands.
2003-03-30 21:47:16 +00:00
} else {
struct aac_blockwrite64 *bw;
bw = (struct aac_blockwrite64 *)&fib->data[0];
bw->Command = VM_CtHostWrite64;
bw->ContainerId = ad->ad_container->co_mntobj.ObjectId;
bw->SectorCount = bp->bio_bcount / AAC_BLOCK_SIZE;
bw->BlockNumber = bp->bio_pblkno;
bw->Pad = 0;
bw->Flags = 0;
fib->Header.Size += sizeof(struct aac_blockwrite64);
cm->cm_flags |= AAC_CMD_DATAIN;
Avoid casts as lvalues.
2004-07-28 06:21:53 +00:00
cm->cm_sgtable = (struct aac_sg_table *)&bw->SgMap64;
Add the ability to send 64-bit scatter/gather elements to aac cards. This is enabled when both the size of bus_addr_t > 4 and the card claims support. Don't wake up the kthread to allocate more commands if we know that we've already allocated the max number of commands.
2003-03-30 21:47:16 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
*cmp = cm;
return(0);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
fail:
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (bp != NULL)
aac_enqueue_bio(sc, bp);
if (cm != NULL)
aac_release_command(cm);
return(ENOMEM);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Handle a bio-instigated command that has been completed.
*/
static void
aac_bio_complete(struct aac_command *cm)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_blockread_response *brr;
struct aac_blockwrite_response *bwr;
struct bio *bp;
AAC_FSAStatus status;
/* fetch relevant status and then release the command */
bp = (struct bio *)cm->cm_private;
The aac driver has evolved enough over the last few months that it no longer resembles the 4.x version very much. Garbage collect the legacy bits.
2003-03-01 05:07:19 +00:00
if (bp->bio_cmd == BIO_READ) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
brr = (struct aac_blockread_response *)&cm->cm_fib->data[0];
status = brr->Status;
} else {
bwr = (struct aac_blockwrite_response *)&cm->cm_fib->data[0];
status = bwr->Status;
}
aac_release_command(cm);
/* fix up the bio based on status */
if (status == ST_OK) {
bp->bio_resid = 0;
} else {
bp->bio_error = EIO;
bp->bio_flags |= BIO_ERROR;
/* pass an error string out to the disk layer */
bp->bio_driver1 = aac_describe_code(aac_command_status_table,
status);
}
aac_biodone(bp);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Submit a command to the controller, return when it completes.
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
* XXX This is very dangerous! If the card has gone out to lunch, we could
* be stuck here forever. At the same time, signals are not caught
Remove the 'timeout' argument from aac_wait_command() as it isn't used and never will be. Update the XXX comment for this function to accurately reflect why things are the way they are.
2004-04-14 19:11:29 +00:00
* because there is a risk that a signal could wakeup the sleep before
* the card has a chance to complete the command. Since there is no way
* to cancel a command that is in progress, we can't protect against the
* card completing a command late and spamming the command and data
* memory. So, we are held hostage until the command completes.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*/
static int
Remove the 'timeout' argument from aac_wait_command() as it isn't used and never will be. Update the XXX comment for this function to accurately reflect why things are the way they are.
2004-04-14 19:11:29 +00:00
aac_wait_command(struct aac_command *cm)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
struct aac_softc *sc;
Remove the 'timeout' argument from aac_wait_command() as it isn't used and never will be. Update the XXX comment for this function to accurately reflect why things are the way they are.
2004-04-14 19:11:29 +00:00
int error;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
sc = cm->cm_sc;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* Put the command on the ready queue and get things going */
cm->cm_queue = AAC_ADAP_NORM_CMD_QUEUE;
aac_enqueue_ready(cm);
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
aac_startio(sc);
Remove the 'timeout' argument from aac_wait_command() as it isn't used and never will be. Update the XXX comment for this function to accurately reflect why things are the way they are.
2004-04-14 19:11:29 +00:00
error = msleep(cm, &sc->aac_io_lock, PRIBIO, "aacwait", 0);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(error);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
*Command Buffer Management
*/
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Allocate a command.
*/
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
int
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
aac_alloc_command(struct aac_softc *sc, struct aac_command **cmp)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_command *cm;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
if ((cm = aac_dequeue_free(sc)) == NULL) {
Add the ability to send 64-bit scatter/gather elements to aac cards. This is enabled when both the size of bus_addr_t > 4 and the card claims support. Don't wake up the kthread to allocate more commands if we know that we've already allocated the max number of commands.
2003-03-30 21:47:16 +00:00
if (sc->total_fibs < sc->aac_max_fibs) {
sc->aifflags |= AAC_AIFFLAGS_ALLOCFIBS;
wakeup(sc->aifthread);
}
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
return (EBUSY);
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
*cmp = cm;
return(0);
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
* Release a command back to the freelist.
*/
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
void
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
aac_release_command(struct aac_command *cm)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
/* (re)initialise the command/FIB */
cm->cm_sgtable = NULL;
cm->cm_flags = 0;
cm->cm_complete = NULL;
cm->cm_private = NULL;
cm->cm_fib->Header.XferState = AAC_FIBSTATE_EMPTY;
cm->cm_fib->Header.StructType = AAC_FIBTYPE_TFIB;
cm->cm_fib->Header.Flags = 0;
cm->cm_fib->Header.SenderSize = sizeof(struct aac_fib);
/*
* These are duplicated in aac_start to cover the case where an
* intermediate stage may have destroyed them. They're left
* initialised here for debugging purposes only.
*/
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
cm->cm_fib->Header.ReceiverFibAddress = (u_int32_t)cm->cm_fibphys;
cm->cm_fib->Header.SenderData = 0;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
aac_enqueue_free(cm);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
* Map helper for command/FIB allocation.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*/
static void
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
aac_map_command_helper(void *arg, bus_dma_segment_t *segs, int nseg, int error)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
uint32_t *fibphys;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
fibphys = (uint32_t *)arg;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
*fibphys = segs[0].ds_addr;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
* Allocate and initialise commands/FIBs for this adapter.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*/
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
static int
aac_alloc_commands(struct aac_softc *sc)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_command *cm;
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
struct aac_fibmap *fm;
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
uint32_t fibphys;
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
int i, error;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
if (sc->total_fibs + AAC_FIB_COUNT > sc->aac_max_fibs)
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
return (ENOMEM);
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
fm = malloc(sizeof(struct aac_fibmap), M_AACBUF, M_NOWAIT|M_ZERO);
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
if (fm == NULL)
return (ENOMEM);
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* allocate the FIBs in DMAable memory and load them */
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
if (bus_dmamem_alloc(sc->aac_fib_dmat, (void **)&fm->aac_fibs,
BUS_DMA_NOWAIT, &fm->aac_fibmap)) {
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
device_printf(sc->aac_dev,
"Not enough contiguous memory available.\n");
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
free(fm, M_AACBUF);
The aac driver needs 256K of contiguous physical memory in order to attach. If it can't get that (highly likely if loaded as a module on a system that's been up for a while), give a more descriptive error message. Also clean up some nearby style nits. MFC after: 2 days
2002-08-30 05:02:07 +00:00
return (ENOMEM);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
Hack around a bug in the 2200 and 2120 controllers that connot DMA commands from below the first 8K of physical memory. A better fix is to modify the busdma api to allow either inclusion ranges or multiple exclusion ranges, but that debate is for another day. MFC After: 2 days
2003-01-23 01:01:44 +00:00
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
/* Ignore errors since this doesn't bounce */
(void)bus_dmamap_load(sc->aac_fib_dmat, fm->aac_fibmap, fm->aac_fibs,
AAC_FIB_COUNT * sizeof(struct aac_fib),
aac_map_command_helper, &fibphys, 0);
Hack around a bug in the 2200 and 2120 controllers that connot DMA commands from below the first 8K of physical memory. A better fix is to modify the busdma api to allow either inclusion ranges or multiple exclusion ranges, but that debate is for another day. MFC After: 2 days
2003-01-23 01:01:44 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* initialise constant fields in the command structure */
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_io_lock);
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
bzero(fm->aac_fibs, AAC_FIB_COUNT * sizeof(struct aac_fib));
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
for (i = 0; i < AAC_FIB_COUNT; i++) {
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
cm = sc->aac_commands + sc->total_fibs;
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
fm->aac_commands = cm;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
cm->cm_sc = sc;
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
cm->cm_fib = fm->aac_fibs + i;
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
cm->cm_fibphys = fibphys + (i * sizeof(struct aac_fib));
Fix a 64-bit bogon. The hardware command structure only has one 32 bit field for holding driver-dependant data. Instead of putting the pointer to the driver command struct in there, take advantage of these structs being a (virtually) contiguous array and just put the array index in the field.
2003-02-19 23:33:56 +00:00
cm->cm_index = sc->total_fibs;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
if ((error = bus_dmamap_create(sc->aac_buffer_dmat, 0,
&cm->cm_datamap)) == 0)
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
aac_release_command(cm);
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
else
break;
sc->total_fibs++;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
if (i > 0) {
TAILQ_INSERT_TAIL(&sc->aac_fibmap_tqh, fm, fm_link);
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
debug(1, "total_fibs= %d\n", sc->total_fibs);
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_io_lock);
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
return (0);
}
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_io_lock);
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
bus_dmamap_unload(sc->aac_fib_dmat, fm->aac_fibmap);
bus_dmamem_free(sc->aac_fib_dmat, fm->aac_fibs, fm->aac_fibmap);
free(fm, M_AACBUF);
return (ENOMEM);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
* Free FIBs owned by this adapter.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*/
static void
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
aac_free_commands(struct aac_softc *sc)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
struct aac_fibmap *fm;
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
struct aac_command *cm;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
int i;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
while ((fm = TAILQ_FIRST(&sc->aac_fibmap_tqh)) != NULL) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
TAILQ_REMOVE(&sc->aac_fibmap_tqh, fm, fm_link);
/*
* We check against total_fibs to handle partially
* allocated blocks.
*/
for (i = 0; i < AAC_FIB_COUNT && sc->total_fibs--; i++) {
cm = fm->aac_commands + i;
bus_dmamap_destroy(sc->aac_buffer_dmat, cm->cm_datamap);
}
bus_dmamap_unload(sc->aac_fib_dmat, fm->aac_fibmap);
bus_dmamem_free(sc->aac_fib_dmat, fm->aac_fibs, fm->aac_fibmap);
free(fm, M_AACBUF);
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Command-mapping helper function - populate this command's s/g table.
*/
static void
aac_map_command_sg(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
struct aac_softc *sc;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_command *cm;
struct aac_fib *fib;
int i;
debug_called(3);
cm = (struct aac_command *)arg;
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
sc = cm->cm_sc;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
fib = cm->cm_fib;
/* copy into the FIB */
Add the ability to send 64-bit scatter/gather elements to aac cards. This is enabled when both the size of bus_addr_t > 4 and the card claims support. Don't wake up the kthread to allocate more commands if we know that we've already allocated the max number of commands.
2003-03-30 21:47:16 +00:00
if (cm->cm_sgtable != NULL) {
if ((cm->cm_sc->flags & AAC_FLAGS_SG_64BIT) == 0) {
struct aac_sg_table *sg;
sg = cm->cm_sgtable;
sg->SgCount = nseg;
for (i = 0; i < nseg; i++) {
sg->SgEntry[i].SgAddress = segs[i].ds_addr;
sg->SgEntry[i].SgByteCount = segs[i].ds_len;
}
/* update the FIB size for the s/g count */
fib->Header.Size += nseg * sizeof(struct aac_sg_entry);
} else {
struct aac_sg_table64 *sg;
sg = (struct aac_sg_table64 *)cm->cm_sgtable;
sg->SgCount = nseg;
for (i = 0; i < nseg; i++) {
sg->SgEntry64[i].SgAddress = segs[i].ds_addr;
sg->SgEntry64[i].SgByteCount = segs[i].ds_len;
}
/* update the FIB size for the s/g count */
fib->Header.Size += nseg*sizeof(struct aac_sg_entry64);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
/* Fix up the address values in the FIB. Use the command array index
* instead of a pointer since these fields are only 32 bits. Shift
* the SenderFibAddress over to make room for the fast response bit.
*/
cm->cm_fib->Header.SenderFibAddress = (cm->cm_index << 1);
cm->cm_fib->Header.ReceiverFibAddress = cm->cm_fibphys;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
/* save a pointer to the command for speedy reverse-lookup */
cm->cm_fib->Header.SenderData = cm->cm_index;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
if (cm->cm_flags & AAC_CMD_DATAIN)
bus_dmamap_sync(sc->aac_buffer_dmat, cm->cm_datamap,
BUS_DMASYNC_PREREAD);
if (cm->cm_flags & AAC_CMD_DATAOUT)
bus_dmamap_sync(sc->aac_buffer_dmat, cm->cm_datamap,
BUS_DMASYNC_PREWRITE);
cm->cm_flags |= AAC_CMD_MAPPED;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Collapse aac_map_command() into aac_startio(). Check the AAC_QUEUE_FRZN in every iteration of aac_startio(). This ensures that a command that is deferred for lack of resources doesn't immediately get retried in the aac_startio() loop. This avoids an almost certain livelock.
2004-06-01 05:32:26 +00:00
/* Put the FIB on the outbound queue */
If a command has to be deferred because there are no more resources for it on the card, unmap it first. This allows it to be picked up properly when the queue gets kicked again. This was the root problem for the lost command (i.e. stuck in getblk/vinvalb) problem. While here, panic if commands don't map correctly instead of just silently ignoring the problem and dropping command. Also slow down the dynamic allocation of new commands. It should be safe to go back into the aac waters. Thanks to everyone who suffered through this and provided good feedback.
2004-02-07 10:30:22 +00:00
if (aac_enqueue_fib(sc, cm->cm_queue, cm) == EBUSY) {
aac_unmap_command(cm);
Collapse aac_map_command() into aac_startio(). Check the AAC_QUEUE_FRZN in every iteration of aac_startio(). This ensures that a command that is deferred for lack of resources doesn't immediately get retried in the aac_startio() loop. This avoids an almost certain livelock.
2004-06-01 05:32:26 +00:00
sc->flags |= AAC_QUEUE_FRZN;
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
aac_requeue_ready(cm);
If a command has to be deferred because there are no more resources for it on the card, unmap it first. This allows it to be picked up properly when the queue gets kicked again. This was the root problem for the lost command (i.e. stuck in getblk/vinvalb) problem. While here, panic if commands don't map correctly instead of just silently ignoring the problem and dropping command. Also slow down the dynamic allocation of new commands. It should be safe to go back into the aac waters. Thanks to everyone who suffered through this and provided good feedback.
2004-02-07 10:30:22 +00:00
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
return;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Unmap a command from controller-visible space.
*/
static void
aac_unmap_command(struct aac_command *cm)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc = cm->cm_sc;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (!(cm->cm_flags & AAC_CMD_MAPPED))
return;
if (cm->cm_datalen != 0) {
if (cm->cm_flags & AAC_CMD_DATAIN)
bus_dmamap_sync(sc->aac_buffer_dmat, cm->cm_datamap,
BUS_DMASYNC_POSTREAD);
if (cm->cm_flags & AAC_CMD_DATAOUT)
bus_dmamap_sync(sc->aac_buffer_dmat, cm->cm_datamap,
BUS_DMASYNC_POSTWRITE);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
bus_dmamap_unload(sc->aac_buffer_dmat, cm->cm_datamap);
}
cm->cm_flags &= ~AAC_CMD_MAPPED;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Hardware Interface
*/
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Initialise the adapter.
*/
static void
aac_common_map(void *arg, bus_dma_segment_t *segs, int nseg, int error)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
debug_called(1);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc = (struct aac_softc *)arg;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_common_busaddr = segs[0].ds_addr;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Don't attach to Dell PERC2/QC cards that have a firmware rev of 1.x. This check is complicated by the fact that the Adaptec 5400S cards claim to use 1.x firmware also. PERC2/QC 1.x firmware is not compatible with this driver and will cause a system hang. MFC after: 3 days
2002-02-06 01:34:09 +00:00
static int
aac_check_firmware(struct aac_softc *sc)
{
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
u_int32_t major, minor, options;
Don't attach to Dell PERC2/QC cards that have a firmware rev of 1.x. This check is complicated by the fact that the Adaptec 5400S cards claim to use 1.x firmware also. PERC2/QC 1.x firmware is not compatible with this driver and will cause a system hang. MFC after: 3 days
2002-02-06 01:34:09 +00:00
debug_called(1);
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
/*
* Retrieve the firmware version numbers. Dell PERC2/QC cards with
* firmware version 1.x are not compatible with this driver.
*/
if (sc->flags & AAC_FLAGS_PERC2QC) {
Don't attach to Dell PERC2/QC cards that have a firmware rev of 1.x. This check is complicated by the fact that the Adaptec 5400S cards claim to use 1.x firmware also. PERC2/QC 1.x firmware is not compatible with this driver and will cause a system hang. MFC after: 3 days
2002-02-06 01:34:09 +00:00
if (aac_sync_command(sc, AAC_MONKER_GETKERNVER, 0, 0, 0, 0,
NULL)) {
device_printf(sc->aac_dev,
"Error reading firmware version\n");
return (EIO);
}
/* These numbers are stored as ASCII! */
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
major = (AAC_GET_MAILBOX(sc, 1) & 0xff) - 0x30;
minor = (AAC_GET_MAILBOX(sc, 2) & 0xff) - 0x30;
Don't attach to Dell PERC2/QC cards that have a firmware rev of 1.x. This check is complicated by the fact that the Adaptec 5400S cards claim to use 1.x firmware also. PERC2/QC 1.x firmware is not compatible with this driver and will cause a system hang. MFC after: 3 days
2002-02-06 01:34:09 +00:00
if (major == 1) {
device_printf(sc->aac_dev,
"Firmware version %d.%d is not supported.\n",
major, minor);
return (EINVAL);
}
}
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
/*
* Retrieve the capabilities/supported options word so we know what
* work-arounds to enable.
*/
if (aac_sync_command(sc, AAC_MONKER_GETINFO, 0, 0, 0, 0, NULL)) {
device_printf(sc->aac_dev, "RequestAdapterInfo failed\n");
return (EIO);
}
options = AAC_GET_MAILBOX(sc, 1);
sc->supported_options = options;
if ((options & AAC_SUPPORTED_4GB_WINDOW) != 0 &&
(sc->flags & AAC_FLAGS_NO4GB) == 0)
sc->flags |= AAC_FLAGS_4GB_WINDOW;
if (options & AAC_SUPPORTED_NONDASD)
sc->flags |= AAC_FLAGS_ENABLE_CAM;
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
if ((options & AAC_SUPPORTED_SGMAP_HOST64) != 0
&& (sizeof(bus_addr_t) > 4)) {
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
device_printf(sc->aac_dev, "Enabling 64-bit address support\n");
sc->flags |= AAC_FLAGS_SG_64BIT;
}
/* Check for broken hardware that does a lower number of commands */
if ((sc->flags & AAC_FLAGS_256FIBS) == 0)
sc->aac_max_fibs = AAC_MAX_FIBS;
else
sc->aac_max_fibs = 256;
Don't attach to Dell PERC2/QC cards that have a firmware rev of 1.x. This check is complicated by the fact that the Adaptec 5400S cards claim to use 1.x firmware also. PERC2/QC 1.x firmware is not compatible with this driver and will cause a system hang. MFC after: 3 days
2002-02-06 01:34:09 +00:00
return (0);
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
static int
aac_init(struct aac_softc *sc)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_adapter_init *ip;
time_t then;
Make aac(4) compile cleanly on 64-bit machines. The code was already 64-bit safe, but some (unneeded and/or harmless) downcasts were generating warnings. The driver still is not endian-clean.
2003-08-19 21:31:45 +00:00
u_int32_t code, qoffset;
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
int error;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
/*
* First wait for the adapter to come ready.
*/
then = time_second;
do {
code = AAC_GET_FWSTATUS(sc);
if (code & AAC_SELF_TEST_FAILED) {
device_printf(sc->aac_dev, "FATAL: selftest failed\n");
return(ENXIO);
}
if (code & AAC_KERNEL_PANIC) {
device_printf(sc->aac_dev,
"FATAL: controller kernel panic\n");
return(ENXIO);
}
if (time_second > (then + AAC_BOOT_TIMEOUT)) {
device_printf(sc->aac_dev,
"FATAL: controller not coming ready, "
"status %x\n", code);
return(ENXIO);
}
} while (!(code & AAC_UP_AND_RUNNING));
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
error = ENOMEM;
/*
* Create DMA tag for mapping buffers into controller-addressable space.
*/
if (bus_dma_tag_create(sc->aac_parent_dmat, /* parent */
1, 0, /* algnmnt, boundary */
(sc->flags & AAC_FLAGS_SG_64BIT) ?
BUS_SPACE_MAXADDR :
BUS_SPACE_MAXADDR_32BIT, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
MAXBSIZE, /* maxsize */
AAC_MAXSGENTRIES, /* nsegments */
MAXBSIZE, /* maxsegsize */
BUS_DMA_ALLOCNOW, /* flags */
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
busdma_lock_mutex, /* lockfunc */
&sc->aac_io_lock, /* lockfuncarg */
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
&sc->aac_buffer_dmat)) {
device_printf(sc->aac_dev, "can't allocate buffer DMA tag\n");
goto out;
}
/*
* Create DMA tag for mapping FIBs into controller-addressable space..
*/
if (bus_dma_tag_create(sc->aac_parent_dmat, /* parent */
1, 0, /* algnmnt, boundary */
(sc->flags & AAC_FLAGS_4GB_WINDOW) ?
BUS_SPACE_MAXADDR_32BIT :
0x7fffffff, /* lowaddr */
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
AAC_FIB_COUNT *
sizeof(struct aac_fib), /* maxsize */
1, /* nsegments */
AAC_FIB_COUNT *
sizeof(struct aac_fib), /* maxsegsize */
Don't force busdma to pre-allocate bounce pages for static allocations.
2004-11-21 04:03:29 +00:00
0, /* flags */
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
NULL, NULL, /* No locking needed */
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
&sc->aac_fib_dmat)) {
device_printf(sc->aac_dev, "can't allocate FIB DMA tag\n");;
goto out;
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Create DMA tag for the common structure and allocate it.
*/
if (bus_dma_tag_create(sc->aac_parent_dmat, /* parent */
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
1, 0, /* algnmnt, boundary */
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
(sc->flags & AAC_FLAGS_4GB_WINDOW) ?
BUS_SPACE_MAXADDR_32BIT :
0x7fffffff, /* lowaddr */
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
BUS_SPACE_MAXADDR, /* highaddr */
NULL, NULL, /* filter, filterarg */
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
8192 + sizeof(struct aac_common), /* maxsize */
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
1, /* nsegments */
BUS_SPACE_MAXSIZE_32BIT, /* maxsegsize */
Don't force busdma to pre-allocate bounce pages for static allocations.
2004-11-21 04:03:29 +00:00
0, /* flags */
Mega busdma API commit. Add two new arguments to bus_dma_tag_create(): lockfunc and lockfuncarg. Lockfunc allows a driver to provide a function for managing its locking semantics while using busdma. At the moment, this is used for the asynchronous busdma_swi and callback mechanism. Two lockfunc implementations are provided: busdma_lock_mutex() performs standard mutex operations on the mutex that is specified from lockfuncarg. dftl_lock() is a panic implementation and is defaulted to when NULL, NULL are passed to bus_dma_tag_create(). The only time that NULL, NULL should ever be used is when the driver ensures that bus_dmamap_load() will not be deferred. Drivers that do not provide their own locking can pass busdma_lock_mutex,&Giant args in order to preserve the former behaviour. sparc64 and powerpc do not provide real busdma_swi functions, so this is largely a noop on those platforms. The busdma_swi on is64 is not properly locked yet, so warnings will be emitted on this platform when busdma callback deferrals happen. If anyone gets panics or warnings from dflt_lock() being called, please let me know right away. Reviewed by: tmm, gibbs
2003-07-01 15:52:06 +00:00
NULL, NULL, /* No locking needed */
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
&sc->aac_common_dmat)) {
device_printf(sc->aac_dev,
"can't allocate common structure DMA tag\n");
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
goto out;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (bus_dmamem_alloc(sc->aac_common_dmat, (void **)&sc->aac_common,
BUS_DMA_NOWAIT, &sc->aac_common_dmamap)) {
device_printf(sc->aac_dev, "can't allocate common structure\n");
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
goto out;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
/*
* Work around a bug in the 2120 and 2200 that cannot DMA commands
* below address 8192 in physical memory.
* XXX If the padding is not needed, can it be put to use instead
* of ignored?
*/
Handle the EINPROGRESS case of bus_dmamap_load() for data buffers.
2003-07-09 21:16:23 +00:00
(void)bus_dmamap_load(sc->aac_common_dmat, sc->aac_common_dmamap,
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
sc->aac_common, 8192 + sizeof(*sc->aac_common),
aac_common_map, sc, 0);
if (sc->aac_common_busaddr < 8192) {
Avoid casts as lvalues.
2004-07-28 06:21:53 +00:00
sc->aac_common = (struct aac_common *)
((uint8_t *)sc->aac_common + 8192);
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
sc->aac_common_busaddr += 8192;
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
bzero(sc->aac_common, sizeof(*sc->aac_common));
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
/* Allocate some FIBs and associated command structs */
TAILQ_INIT(&sc->aac_fibmap_tqh);
sc->aac_commands = malloc(AAC_MAX_FIBS * sizeof(struct aac_command),
o Move the cleanup of the fib maps into aac_free_commands() so as to retain symetry with aac_alloc_commans(). Since aac_alloc_commands() allocates fib maps and places them onto the fib lists, aac_free_commands() should reverse those operations. o Combine two ifs with the same body with an ||. o Switch from uintptr_t to uint32_t for fib map load operations. The target is a uint32_t so using this type for the map load call avoids an extra cast. uintptr_t should only be used when you need an "int sized the same as the machine's poitner size" which is not the case here. o Removed the commented out M_WAITOK flag in the allocation in aac_alloc_commands(). The kernel will only block in the allocator if it can grow the size of the kernel. This usually results in a page-out which could involve this aac device. Thus, sleeping here could deadlock the machine. Assuming this operation is occurring outside of attach time, we have enough fibs to operate anyway, so waiting for fibs to free up is okay if not optimal. o In aac_alloc_commands(), if we cannot dmamem_alloc additional fib space, free the fib map. o In aac_alloc_commands(), if we cannot create per-command dmamaps, don't lose track of the fib map that is mapping all of the commands that we have already released into the free pool. Instead, just cut out of the loop and modify aac_free_commands to not attempt to free maps that have not been allocated. o Don't use a magic number when pre-allocating fibs. o Use PAGE_SIZE to allocate in page sized chunks instead of an architecture specific constant. Submitted by: gibbs
2003-02-19 21:38:29 +00:00
M_AACBUF, M_WAITOK|M_ZERO);
while (sc->total_fibs < AAC_PREALLOCATE_FIBS) {
Implement a new dynamic command allocator. FIBs are allocated in 1 page blocks now, which should eliminate problems with the driver failing to attach due to insufficient contiguous RAM. Allow the FIB pool to grow from the default of 128 to the max of 512 as demand grows. Also pad the adapter init struct to work around the 2120/2200 DMA bug now that there is no longer a FIB slab.
2003-02-10 00:34:24 +00:00
if (aac_alloc_commands(sc) != 0)
break;
}
if (sc->total_fibs == 0)
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
goto out;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Fill in the init structure. This tells the adapter about the
* physical location of various important shared data structures.
*/
ip = &sc->aac_common->ac_init;
ip->InitStructRevision = AAC_INIT_STRUCT_REVISION;
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
ip->MiniPortRevision = AAC_INIT_STRUCT_MINIPORT_REVISION;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
ip->AdapterFibsPhysicalAddress = sc->aac_common_busaddr +
offsetof(struct aac_common, ac_fibs);
Don't pass pointers to kernel data structures through 32-bit fields as 64-bit architectures won't like this. Use virtual array indexes instead. This *should* allow the driver to work on 64-bit platforms, though it's still not endian clean.
2003-04-28 06:16:20 +00:00
ip->AdapterFibsVirtualAddress = 0;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
ip->AdapterFibsSize = AAC_ADAPTER_FIBS * sizeof(struct aac_fib);
ip->AdapterFibAlign = sizeof(struct aac_fib);
ip->PrintfBufferAddress = sc->aac_common_busaddr +
offsetof(struct aac_common, ac_printf);
ip->PrintfBufferSize = AAC_PRINTF_BUFSIZE;
Add a new quirk for cards that incorrectly interpret the amount of memory in the system. This might also have a small performance gain.
2003-07-09 19:19:16 +00:00
/*
* The adapter assumes that pages are 4K in size, except on some
* broken firmware versions that do the page->byte conversion twice,
* therefore 'assuming' that this value is in 16MB units (2^24).
* Round up since the granularity is so high.
*/
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
ip->HostPhysMemPages = ctob(physmem) / AAC_PAGE_SIZE;
Add a new quirk for cards that incorrectly interpret the amount of memory in the system. This might also have a small performance gain.
2003-07-09 19:19:16 +00:00
if (sc->flags & AAC_FLAGS_BROKEN_MEMMAP) {
ip->HostPhysMemPages =
(ip->HostPhysMemPages + AAC_PAGE_SIZE) / AAC_PAGE_SIZE;
Fix a missing } that got dropped from the last commit.
2003-07-09 19:30:30 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
ip->HostElapsedSeconds = time_second; /* reset later if invalid */
/*
* Initialise FIB queues. Note that it appears that the layout of the
* indexes and the segmentation of the entries may be mandated by the
* adapter, which is only told about the base of the queue index fields.
*
* The initial values of the indices are assumed to inform the adapter
* of the sizes of the respective queues, and theoretically it could
* work out the entire layout of the queue structures from this. We
* take the easy route and just lay this area out like everyone else
* does.
*
* The Linux driver uses a much more complex scheme whereby several
* header records are kept for each queue. We use a couple of generic
* list manipulation functions which 'know' the size of each list by
* virtue of a table.
*/
Make aac(4) compile cleanly on 64-bit machines. The code was already 64-bit safe, but some (unneeded and/or harmless) downcasts were generating warnings. The driver still is not endian-clean.
2003-08-19 21:31:45 +00:00
qoffset = offsetof(struct aac_common, ac_qbuf) + AAC_QUEUE_ALIGN;
If ~ chars were pennies, I'd be pennyless. This should fix all of the 'command not in queue' panics. Also fix a nearby style problem.
2003-09-01 00:20:29 +00:00
qoffset &= ~(AAC_QUEUE_ALIGN - 1);
sc->aac_queues =
(struct aac_queue_table *)((uintptr_t)sc->aac_common + qoffset);
Make aac(4) compile cleanly on 64-bit machines. The code was already 64-bit safe, but some (unneeded and/or harmless) downcasts were generating warnings. The driver still is not endian-clean.
2003-08-19 21:31:45 +00:00
ip->CommHeaderAddress = sc->aac_common_busaddr + qoffset;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_HOST_NORM_CMD_QUEUE][AAC_PRODUCER_INDEX] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_HOST_NORM_CMD_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_HOST_NORM_CMD_QUEUE][AAC_CONSUMER_INDEX] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_HOST_NORM_CMD_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_HOST_HIGH_CMD_QUEUE][AAC_PRODUCER_INDEX] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_HOST_HIGH_CMD_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_HOST_HIGH_CMD_QUEUE][AAC_CONSUMER_INDEX] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_HOST_HIGH_CMD_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_ADAP_NORM_CMD_QUEUE][AAC_PRODUCER_INDEX] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_ADAP_NORM_CMD_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_ADAP_NORM_CMD_QUEUE][AAC_CONSUMER_INDEX] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_ADAP_NORM_CMD_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_ADAP_HIGH_CMD_QUEUE][AAC_PRODUCER_INDEX] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_ADAP_HIGH_CMD_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_ADAP_HIGH_CMD_QUEUE][AAC_CONSUMER_INDEX] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_ADAP_HIGH_CMD_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_HOST_NORM_RESP_QUEUE][AAC_PRODUCER_INDEX]=
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_HOST_NORM_RESP_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_HOST_NORM_RESP_QUEUE][AAC_CONSUMER_INDEX]=
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_HOST_NORM_RESP_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_HOST_HIGH_RESP_QUEUE][AAC_PRODUCER_INDEX]=
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_HOST_HIGH_RESP_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_HOST_HIGH_RESP_QUEUE][AAC_CONSUMER_INDEX]=
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_HOST_HIGH_RESP_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_ADAP_NORM_RESP_QUEUE][AAC_PRODUCER_INDEX]=
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_ADAP_NORM_RESP_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_ADAP_NORM_RESP_QUEUE][AAC_CONSUMER_INDEX]=
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_ADAP_NORM_RESP_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_ADAP_HIGH_RESP_QUEUE][AAC_PRODUCER_INDEX]=
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_ADAP_HIGH_RESP_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_queues->qt_qindex[AAC_ADAP_HIGH_RESP_QUEUE][AAC_CONSUMER_INDEX]=
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
AAC_ADAP_HIGH_RESP_ENTRIES;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_qentries[AAC_HOST_NORM_CMD_QUEUE] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
&sc->aac_queues->qt_HostNormCmdQueue[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_qentries[AAC_HOST_HIGH_CMD_QUEUE] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
&sc->aac_queues->qt_HostHighCmdQueue[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_qentries[AAC_ADAP_NORM_CMD_QUEUE] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
&sc->aac_queues->qt_AdapNormCmdQueue[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_qentries[AAC_ADAP_HIGH_CMD_QUEUE] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
&sc->aac_queues->qt_AdapHighCmdQueue[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_qentries[AAC_HOST_NORM_RESP_QUEUE] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
&sc->aac_queues->qt_HostNormRespQueue[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_qentries[AAC_HOST_HIGH_RESP_QUEUE] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
&sc->aac_queues->qt_HostHighRespQueue[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_qentries[AAC_ADAP_NORM_RESP_QUEUE] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
&sc->aac_queues->qt_AdapNormRespQueue[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_qentries[AAC_ADAP_HIGH_RESP_QUEUE] =
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
&sc->aac_queues->qt_AdapHighRespQueue[0];
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Do controller-type-specific initialisation
*/
switch (sc->aac_hwif) {
case AAC_HWIF_I960RX:
AAC_SETREG4(sc, AAC_RX_ODBR, ~0);
break;
Add support for the Adaptec RAID-On-Chip architecture. This in turn provides support for the Adaptec 2130S adapter. Thanks to Adaptec for providing hardware for this.
2004-08-13 01:44:09 +00:00
case AAC_HWIF_RKT:
AAC_SETREG4(sc, AAC_RKT_ODBR, ~0);
break;
default:
break;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
/*
* Give the init structure to the controller.
*/
if (aac_sync_command(sc, AAC_MONKER_INITSTRUCT,
sc->aac_common_busaddr +
offsetof(struct aac_common, ac_init), 0, 0, 0,
NULL)) {
device_printf(sc->aac_dev,
"error establishing init structure\n");
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
error = EIO;
goto out;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
error = 0;
out:
return(error);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Send a synchronous command to the controller and wait for a result.
*/
static int
aac_sync_command(struct aac_softc *sc, u_int32_t command,
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
u_int32_t arg0, u_int32_t arg1, u_int32_t arg2, u_int32_t arg3,
u_int32_t *sp)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
time_t then;
u_int32_t status;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* populate the mailbox */
AAC_SET_MAILBOX(sc, command, arg0, arg1, arg2, arg3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* ensure the sync command doorbell flag is cleared */
AAC_CLEAR_ISTATUS(sc, AAC_DB_SYNC_COMMAND);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* then set it to signal the adapter */
AAC_QNOTIFY(sc, AAC_DB_SYNC_COMMAND);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* spin waiting for the command to complete */
then = time_second;
do {
if (time_second > (then + AAC_IMMEDIATE_TIMEOUT)) {
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
debug(1, "timed out");
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(EIO);
}
} while (!(AAC_GET_ISTATUS(sc) & AAC_DB_SYNC_COMMAND));
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* clear the completion flag */
AAC_CLEAR_ISTATUS(sc, AAC_DB_SYNC_COMMAND);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* get the command status */
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
status = AAC_GET_MAILBOX(sc, 0);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (sp != NULL)
*sp = status;
return(0);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
int
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
aac_sync_fib(struct aac_softc *sc, u_int32_t command, u_int32_t xferstate,
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
struct aac_fib *fib, u_int16_t datasize)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
if (datasize > AAC_FIB_DATASIZE)
return(EINVAL);
/*
* Set up the sync FIB
*/
fib->Header.XferState = AAC_FIBSTATE_HOSTOWNED |
AAC_FIBSTATE_INITIALISED |
AAC_FIBSTATE_EMPTY;
fib->Header.XferState |= xferstate;
fib->Header.Command = command;
fib->Header.StructType = AAC_FIBTYPE_TFIB;
fib->Header.Size = sizeof(struct aac_fib) + datasize;
fib->Header.SenderSize = sizeof(struct aac_fib);
Make aac(4) compile cleanly on 64-bit machines. The code was already 64-bit safe, but some (unneeded and/or harmless) downcasts were generating warnings. The driver still is not endian-clean.
2003-08-19 21:31:45 +00:00
fib->Header.SenderFibAddress = 0; /* Not needed */
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
fib->Header.ReceiverFibAddress = sc->aac_common_busaddr +
offsetof(struct aac_common,
ac_sync_fib);
/*
* Give the FIB to the controller, wait for a response.
*/
if (aac_sync_command(sc, AAC_MONKER_SYNCFIB,
fib->Header.ReceiverFibAddress, 0, 0, 0, NULL)) {
debug(2, "IO error");
return(EIO);
}
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
return (0);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Adapter-space FIB queue manipulation
*
* Note that the queue implementation here is a little funky; neither the PI or
* CI will ever be zero. This behaviour is a controller feature.
*/
static struct {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
int size;
int notify;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
} aac_qinfo[] = {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
{AAC_HOST_NORM_CMD_ENTRIES, AAC_DB_COMMAND_NOT_FULL},
{AAC_HOST_HIGH_CMD_ENTRIES, 0},
{AAC_ADAP_NORM_CMD_ENTRIES, AAC_DB_COMMAND_READY},
{AAC_ADAP_HIGH_CMD_ENTRIES, 0},
{AAC_HOST_NORM_RESP_ENTRIES, AAC_DB_RESPONSE_NOT_FULL},
{AAC_HOST_HIGH_RESP_ENTRIES, 0},
{AAC_ADAP_NORM_RESP_ENTRIES, AAC_DB_RESPONSE_READY},
{AAC_ADAP_HIGH_RESP_ENTRIES, 0}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
};
/*
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
* Atomically insert an entry into the nominated queue, returns 0 on success or
* EBUSY if the queue is full.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
* Note: it would be more efficient to defer notifying the controller in
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
* the case where we may be inserting several entries in rapid succession,
* but implementing this usefully may be difficult (it would involve a
* separate queue/notify interface).
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*/
static int
Bugfixes. Close a race and logic bug in the timeout handling, don't call the interrupt handler from the upper half, etc. This fixes some serious stability problems that we were seeing on our production server. These patches have been tested for almost 6 months and are a highly recommended MFC candidate. Reviewed by: gibbs, merry, msmith MFC after: 4 days
2001-08-05 06:20:13 +00:00
aac_enqueue_fib(struct aac_softc *sc, int queue, struct aac_command *cm)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
u_int32_t pi, ci;
The aac driver has evolved enough over the last few months that it no longer resembles the 4.x version very much. Garbage collect the legacy bits.
2003-03-01 05:07:19 +00:00
int error;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
u_int32_t fib_size;
u_int32_t fib_addr;
Bugfixes. Close a race and logic bug in the timeout handling, don't call the interrupt handler from the upper half, etc. This fixes some serious stability problems that we were seeing on our production server. These patches have been tested for almost 6 months and are a highly recommended MFC candidate. Reviewed by: gibbs, merry, msmith MFC after: 4 days
2001-08-05 06:20:13 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
fib_size = cm->cm_fib->Header.Size;
fib_addr = cm->cm_fib->Header.ReceiverFibAddress;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* get the producer/consumer indices */
pi = sc->aac_queues->qt_qindex[queue][AAC_PRODUCER_INDEX];
ci = sc->aac_queues->qt_qindex[queue][AAC_CONSUMER_INDEX];
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* wrap the queue? */
if (pi >= aac_qinfo[queue].size)
pi = 0;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* check for queue full */
if ((pi + 1) == ci) {
error = EBUSY;
goto out;
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Commit the correct version of the patch from last night. This fixes an immediate panic when doing any i/o, and it closes a completion race.
2004-06-01 15:50:11 +00:00
/*
* To avoid a race with its completion interrupt, place this command on
* the busy queue prior to advertising it to the controller.
*/
aac_enqueue_busy(cm);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* populate queue entry */
(sc->aac_qentries[queue] + pi)->aq_fib_size = fib_size;
(sc->aac_qentries[queue] + pi)->aq_fib_addr = fib_addr;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* update producer index */
sc->aac_queues->qt_qindex[queue][AAC_PRODUCER_INDEX] = pi + 1;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* notify the adapter if we know how */
if (aac_qinfo[queue].notify != 0)
AAC_QNOTIFY(sc, aac_qinfo[queue].notify);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
error = 0;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
out:
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(error);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
/*
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
* Atomically remove one entry from the nominated queue, returns 0 on
* success or ENOENT if the queue is empty.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*/
static int
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
aac_dequeue_fib(struct aac_softc *sc, int queue, u_int32_t *fib_size,
struct aac_fib **fib_addr)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
u_int32_t pi, ci;
Don't pass pointers to kernel data structures through 32-bit fields as 64-bit architectures won't like this. Use virtual array indexes instead. This *should* allow the driver to work on 64-bit platforms, though it's still not endian clean.
2003-04-28 06:16:20 +00:00
u_int32_t fib_index;
The aac driver has evolved enough over the last few months that it no longer resembles the 4.x version very much. Garbage collect the legacy bits.
2003-03-01 05:07:19 +00:00
int error;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
int notify;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* get the producer/consumer indices */
pi = sc->aac_queues->qt_qindex[queue][AAC_PRODUCER_INDEX];
ci = sc->aac_queues->qt_qindex[queue][AAC_CONSUMER_INDEX];
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* check for queue empty */
if (ci == pi) {
error = ENOENT;
goto out;
}
Correctly wrap the producer queue index when dequeuing commands. This wasn't a problem for command responses since we rarely ever filled the queue. However, adapter-initiated commands have a much smaller queue and could tickle this bug. It's possible that this might fix the recently reported problems with the aac-2120s, though I haven't been able to reproduce the problem locally. MFC-After: 1 day
2003-09-16 16:07:15 +00:00
/* wrap the pi so the following test works */
if (pi >= aac_qinfo[queue].size)
pi = 0;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
notify = 0;
if (ci == pi + 1)
notify++;
Bugfixes. Close a race and logic bug in the timeout handling, don't call the interrupt handler from the upper half, etc. This fixes some serious stability problems that we were seeing on our production server. These patches have been tested for almost 6 months and are a highly recommended MFC candidate. Reviewed by: gibbs, merry, msmith MFC after: 4 days
2001-08-05 06:20:13 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* wrap the queue? */
if (ci >= aac_qinfo[queue].size)
ci = 0;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* fetch the entry */
*fib_size = (sc->aac_qentries[queue] + ci)->aq_fib_size;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Don't pass pointers to kernel data structures through 32-bit fields as 64-bit architectures won't like this. Use virtual array indexes instead. This *should* allow the driver to work on 64-bit platforms, though it's still not endian clean.
2003-04-28 06:16:20 +00:00
switch (queue) {
case AAC_HOST_NORM_CMD_QUEUE:
case AAC_HOST_HIGH_CMD_QUEUE:
/*
* The aq_fib_addr is only 32 bits wide so it can't be counted
* on to hold an address. For AIF's, the adapter assumes
* that it's giving us an address into the array of AIF fibs.
* Therefore, we have to convert it to an index.
*/
fib_index = (sc->aac_qentries[queue] + ci)->aq_fib_addr /
sizeof(struct aac_fib);
*fib_addr = &sc->aac_common->ac_fibs[fib_index];
break;
case AAC_HOST_NORM_RESP_QUEUE:
case AAC_HOST_HIGH_RESP_QUEUE:
{
struct aac_command *cm;
/*
* As above, an index is used instead of an actual address.
* Gotta shift the index to account for the fast response
* bit. No other correction is needed since this value was
* originally provided by the driver via the SenderFibAddress
* field.
*/
fib_index = (sc->aac_qentries[queue] + ci)->aq_fib_addr;
cm = sc->aac_commands + (fib_index >> 1);
*fib_addr = cm->cm_fib;
/*
* Is this a fast response? If it is, update the fib fields in
* local memory since the whole fib isn't DMA'd back up.
*/
if (fib_index & 0x01) {
(*fib_addr)->Header.XferState |= AAC_FIBSTATE_DONEADAP;
*((u_int32_t*)((*fib_addr)->data)) = AAC_ERROR_NORMAL;
}
break;
}
default:
panic("Invalid queue in aac_dequeue_fib()");
break;
Major bugfixes for large memory and fast systems. aac.c: Re-arrange the interrupt handler to optimize the common case of the adapter interrupting us because one or more commands are complete, and do a read across the pci bus to ensure that all posted status writes are flushed. This should close a race that could cause command completion interrupts to be lost. Follow the spec a bit closer when filling out command structures. Enable the Fast Response feature to eliminate the need for the card to DMA successfull command completions back into host memory. Tell the controller how much physical memory we have. Without this there was a chance that our DMA regions would collide with the memory window used by the cache on the controller. The result would be massive data corruption. This seemed to mainly affect systems with >2GB of memory. Fix a few whitespace problems. aac_debug.c: Add an extra diagnostic when printing out commands. aac_disk.c: Add extra sanity checks. aacreg.h: Prepare for making this 64-bit clean by reducing the use of enumeration types in structures. Many thanks to Justin Gibbs for helping track these down.
2003-01-11 01:59:21 +00:00
}
Don't pass pointers to kernel data structures through 32-bit fields as 64-bit architectures won't like this. Use virtual array indexes instead. This *should* allow the driver to work on 64-bit platforms, though it's still not endian clean.
2003-04-28 06:16:20 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* update consumer index */
sc->aac_queues->qt_qindex[queue][AAC_CONSUMER_INDEX] = ci + 1;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* if we have made the queue un-full, notify the adapter */
if (notify && (aac_qinfo[queue].notify != 0))
AAC_QNOTIFY(sc, aac_qinfo[queue].notify);
error = 0;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
out:
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(error);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
* Put our response to an Adapter Initialed Fib on the response queue
*/
static int
aac_enqueue_response(struct aac_softc *sc, int queue, struct aac_fib *fib)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
u_int32_t pi, ci;
The aac driver has evolved enough over the last few months that it no longer resembles the 4.x version very much. Garbage collect the legacy bits.
2003-03-01 05:07:19 +00:00
int error;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
u_int32_t fib_size;
u_int32_t fib_addr;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(1);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* Tell the adapter where the FIB is */
fib_size = fib->Header.Size;
fib_addr = fib->Header.SenderFibAddress;
fib->Header.ReceiverFibAddress = fib_addr;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* get the producer/consumer indices */
pi = sc->aac_queues->qt_qindex[queue][AAC_PRODUCER_INDEX];
ci = sc->aac_queues->qt_qindex[queue][AAC_CONSUMER_INDEX];
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* wrap the queue? */
if (pi >= aac_qinfo[queue].size)
pi = 0;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* check for queue full */
if ((pi + 1) == ci) {
error = EBUSY;
goto out;
}
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* populate queue entry */
(sc->aac_qentries[queue] + pi)->aq_fib_size = fib_size;
(sc->aac_qentries[queue] + pi)->aq_fib_addr = fib_addr;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* update producer index */
sc->aac_queues->qt_qindex[queue][AAC_PRODUCER_INDEX] = pi + 1;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* notify the adapter if we know how */
if (aac_qinfo[queue].notify != 0)
AAC_QNOTIFY(sc, aac_qinfo[queue].notify);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
error = 0;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
out:
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(error);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
* Check for commands that have been outstanding for a suspiciously long time,
* and complain about them.
*/
static void
aac_timeout(struct aac_softc *sc)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_command *cm;
time_t deadline;
If the timeout handler runs and notices that commands are timed out, check the firmware status register on the card to see if the firmware is still running. There is no way to recover from this, but at least it can give a hint as whether the car has crashed (which happens all too often). MFC after: 3 days
2004-09-16 02:37:40 +00:00
int timedout, code;
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
* Traverse the busy command list, bitch about late commands once
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
* only.
*/
If the timeout handler runs and notices that commands are timed out, check the firmware status register on the card to see if the firmware is still running. There is no way to recover from this, but at least it can give a hint as whether the car has crashed (which happens all too often). MFC after: 3 days
2004-09-16 02:37:40 +00:00
timedout = 0;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
deadline = time_second - AAC_CMD_TIMEOUT;
TAILQ_FOREACH(cm, &sc->aac_busy, cm_link) {
if ((cm->cm_timestamp < deadline)
/* && !(cm->cm_flags & AAC_CMD_TIMEDOUT) */) {
cm->cm_flags |= AAC_CMD_TIMEDOUT;
device_printf(sc->aac_dev,
"COMMAND %p TIMEOUT AFTER %d SECONDS\n",
cm, (int)(time_second-cm->cm_timestamp));
AAC_PRINT_FIB(sc, cm->cm_fib);
If the timeout handler runs and notices that commands are timed out, check the firmware status register on the card to see if the firmware is still running. There is no way to recover from this, but at least it can give a hint as whether the car has crashed (which happens all too often). MFC after: 3 days
2004-09-16 02:37:40 +00:00
timedout++;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
}
If the timeout handler runs and notices that commands are timed out, check the firmware status register on the card to see if the firmware is still running. There is no way to recover from this, but at least it can give a hint as whether the car has crashed (which happens all too often). MFC after: 3 days
2004-09-16 02:37:40 +00:00
if (timedout) {
code = AAC_GET_FWSTATUS(sc);
if (code != AAC_UP_AND_RUNNING) {
device_printf(sc->aac_dev, "WARNING! Controller is no "
"longer running! code= 0x%x\n", code);
}
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return;
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Interface Function Vectors
*/
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Read the current firmware status word.
*/
static int
aac_sa_get_fwstatus(struct aac_softc *sc)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(AAC_GETREG4(sc, AAC_SA_FWSTATUS));
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
static int
aac_rx_get_fwstatus(struct aac_softc *sc)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(AAC_GETREG4(sc, AAC_RX_FWSTATUS));
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static int
aac_fa_get_fwstatus(struct aac_softc *sc)
{
int val;
debug_called(3);
val = AAC_GETREG4(sc, AAC_FA_FWSTATUS);
return (val);
}
Add support for the Adaptec RAID-On-Chip architecture. This in turn provides support for the Adaptec 2130S adapter. Thanks to Adaptec for providing hardware for this.
2004-08-13 01:44:09 +00:00
static int
aac_rkt_get_fwstatus(struct aac_softc *sc)
{
debug_called(3);
return(AAC_GETREG4(sc, AAC_RKT_FWSTATUS));
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Notify the controller of a change in a given queue
*/
static void
aac_sa_qnotify(struct aac_softc *sc, int qbit)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
AAC_SETREG2(sc, AAC_SA_DOORBELL1_SET, qbit);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
static void
aac_rx_qnotify(struct aac_softc *sc, int qbit)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
AAC_SETREG4(sc, AAC_RX_IDBR, qbit);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static void
aac_fa_qnotify(struct aac_softc *sc, int qbit)
{
debug_called(3);
AAC_SETREG2(sc, AAC_FA_DOORBELL1, qbit);
AAC_FA_HACK(sc);
}
Add support for the Adaptec RAID-On-Chip architecture. This in turn provides support for the Adaptec 2130S adapter. Thanks to Adaptec for providing hardware for this.
2004-08-13 01:44:09 +00:00
static void
aac_rkt_qnotify(struct aac_softc *sc, int qbit)
{
debug_called(3);
AAC_SETREG4(sc, AAC_RKT_IDBR, qbit);
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Get the interrupt reason bits
*/
static int
aac_sa_get_istatus(struct aac_softc *sc)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(AAC_GETREG2(sc, AAC_SA_DOORBELL0));
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
static int
aac_rx_get_istatus(struct aac_softc *sc)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(AAC_GETREG4(sc, AAC_RX_ODBR));
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static int
aac_fa_get_istatus(struct aac_softc *sc)
{
int val;
debug_called(3);
val = AAC_GETREG2(sc, AAC_FA_DOORBELL0);
return (val);
}
Add support for the Adaptec RAID-On-Chip architecture. This in turn provides support for the Adaptec 2130S adapter. Thanks to Adaptec for providing hardware for this.
2004-08-13 01:44:09 +00:00
static int
aac_rkt_get_istatus(struct aac_softc *sc)
{
debug_called(3);
return(AAC_GETREG4(sc, AAC_RKT_ODBR));
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Clear some interrupt reason bits
*/
static void
aac_sa_clear_istatus(struct aac_softc *sc, int mask)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
AAC_SETREG2(sc, AAC_SA_DOORBELL0_CLEAR, mask);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
static void
aac_rx_clear_istatus(struct aac_softc *sc, int mask)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
AAC_SETREG4(sc, AAC_RX_ODBR, mask);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static void
aac_fa_clear_istatus(struct aac_softc *sc, int mask)
{
debug_called(3);
AAC_SETREG2(sc, AAC_FA_DOORBELL0_CLEAR, mask);
AAC_FA_HACK(sc);
}
Add support for the Adaptec RAID-On-Chip architecture. This in turn provides support for the Adaptec 2130S adapter. Thanks to Adaptec for providing hardware for this.
2004-08-13 01:44:09 +00:00
static void
aac_rkt_clear_istatus(struct aac_softc *sc, int mask)
{
debug_called(3);
AAC_SETREG4(sc, AAC_RKT_ODBR, mask);
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Populate the mailbox and set the command word
*/
static void
aac_sa_set_mailbox(struct aac_softc *sc, u_int32_t command,
u_int32_t arg0, u_int32_t arg1, u_int32_t arg2, u_int32_t arg3)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(4);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
AAC_SETREG4(sc, AAC_SA_MAILBOX, command);
AAC_SETREG4(sc, AAC_SA_MAILBOX + 4, arg0);
AAC_SETREG4(sc, AAC_SA_MAILBOX + 8, arg1);
AAC_SETREG4(sc, AAC_SA_MAILBOX + 12, arg2);
AAC_SETREG4(sc, AAC_SA_MAILBOX + 16, arg3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
static void
aac_rx_set_mailbox(struct aac_softc *sc, u_int32_t command,
u_int32_t arg0, u_int32_t arg1, u_int32_t arg2, u_int32_t arg3)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(4);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
AAC_SETREG4(sc, AAC_RX_MAILBOX, command);
AAC_SETREG4(sc, AAC_RX_MAILBOX + 4, arg0);
AAC_SETREG4(sc, AAC_RX_MAILBOX + 8, arg1);
AAC_SETREG4(sc, AAC_RX_MAILBOX + 12, arg2);
AAC_SETREG4(sc, AAC_RX_MAILBOX + 16, arg3);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static void
aac_fa_set_mailbox(struct aac_softc *sc, u_int32_t command,
u_int32_t arg0, u_int32_t arg1, u_int32_t arg2, u_int32_t arg3)
{
debug_called(4);
AAC_SETREG4(sc, AAC_FA_MAILBOX, command);
AAC_FA_HACK(sc);
AAC_SETREG4(sc, AAC_FA_MAILBOX + 4, arg0);
AAC_FA_HACK(sc);
AAC_SETREG4(sc, AAC_FA_MAILBOX + 8, arg1);
AAC_FA_HACK(sc);
AAC_SETREG4(sc, AAC_FA_MAILBOX + 12, arg2);
AAC_FA_HACK(sc);
AAC_SETREG4(sc, AAC_FA_MAILBOX + 16, arg3);
AAC_FA_HACK(sc);
}
Add support for the Adaptec RAID-On-Chip architecture. This in turn provides support for the Adaptec 2130S adapter. Thanks to Adaptec for providing hardware for this.
2004-08-13 01:44:09 +00:00
static void
aac_rkt_set_mailbox(struct aac_softc *sc, u_int32_t command, u_int32_t arg0,
u_int32_t arg1, u_int32_t arg2, u_int32_t arg3)
{
debug_called(4);
AAC_SETREG4(sc, AAC_RKT_MAILBOX, command);
AAC_SETREG4(sc, AAC_RKT_MAILBOX + 4, arg0);
AAC_SETREG4(sc, AAC_RKT_MAILBOX + 8, arg1);
AAC_SETREG4(sc, AAC_RKT_MAILBOX + 12, arg2);
AAC_SETREG4(sc, AAC_RKT_MAILBOX + 16, arg3);
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Fetch the immediate command status word
*/
static int
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
aac_sa_get_mailbox(struct aac_softc *sc, int mb)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(4);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
return(AAC_GETREG4(sc, AAC_SA_MAILBOX + (mb * 4)));
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
static int
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
aac_rx_get_mailbox(struct aac_softc *sc, int mb)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(4);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
return(AAC_GETREG4(sc, AAC_RX_MAILBOX + (mb * 4)));
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static int
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
aac_fa_get_mailbox(struct aac_softc *sc, int mb)
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
{
int val;
debug_called(4);
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
val = AAC_GETREG4(sc, AAC_FA_MAILBOX + (mb * 4));
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
return (val);
}
Add support for the Adaptec RAID-On-Chip architecture. This in turn provides support for the Adaptec 2130S adapter. Thanks to Adaptec for providing hardware for this.
2004-08-13 01:44:09 +00:00
static int
aac_rkt_get_mailbox(struct aac_softc *sc, int mb)
{
debug_called(4);
return(AAC_GETREG4(sc, AAC_RKT_MAILBOX + (mb * 4)));
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Set/clear interrupt masks
*/
static void
aac_sa_set_interrupts(struct aac_softc *sc, int enable)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug(2, "%sable interrupts", enable ? "en" : "dis");
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (enable) {
AAC_SETREG2((sc), AAC_SA_MASK0_CLEAR, AAC_DB_INTERRUPTS);
} else {
AAC_SETREG2((sc), AAC_SA_MASK0_SET, ~0);
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
static void
aac_rx_set_interrupts(struct aac_softc *sc, int enable)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug(2, "%sable interrupts", enable ? "en" : "dis");
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (enable) {
AAC_SETREG4(sc, AAC_RX_OIMR, ~AAC_DB_INTERRUPTS);
} else {
AAC_SETREG4(sc, AAC_RX_OIMR, ~0);
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static void
aac_fa_set_interrupts(struct aac_softc *sc, int enable)
{
debug(2, "%sable interrupts", enable ? "en" : "dis");
if (enable) {
AAC_SETREG2((sc), AAC_FA_MASK0_CLEAR, AAC_DB_INTERRUPTS);
AAC_FA_HACK(sc);
} else {
AAC_SETREG2((sc), AAC_FA_MASK0, ~0);
AAC_FA_HACK(sc);
}
}
Add support for the Adaptec RAID-On-Chip architecture. This in turn provides support for the Adaptec 2130S adapter. Thanks to Adaptec for providing hardware for this.
2004-08-13 01:44:09 +00:00
static void
aac_rkt_set_interrupts(struct aac_softc *sc, int enable)
{
debug(2, "%sable interrupts", enable ? "en" : "dis");
if (enable) {
AAC_SETREG4(sc, AAC_RKT_OIMR, ~AAC_DB_INTERRUPTS);
} else {
AAC_SETREG4(sc, AAC_RKT_OIMR, ~0);
}
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Debugging and Diagnostics
*/
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Print some information about the controller.
*/
static void
aac_describe_controller(struct aac_softc *sc)
{
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
struct aac_fib *fib;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_adapter_info *info;
debug_called(2);
Collapse sync fib locking into normal i/o locking. The former didn't protect the registers so it was trivially possible for a sync command and i/o command to fight each other and confuse the controller. Make the sync fib alloc/release functions inline and remove the somewhat worthless AAC_SYNC_LOCK_FORCE flag. Thanks to Adil Katchi for helping me to track this down in RELENG_4.
2004-06-02 18:15:48 +00:00
aac_alloc_sync_fib(sc, &fib);
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
fib->data[0] = 0;
if (aac_sync_fib(sc, RequestAdapterInfo, 0, fib, 1)) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
device_printf(sc->aac_dev, "RequestAdapterInfo failed\n");
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
aac_release_sync_fib(sc);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return;
}
/* save the kernel revision structure for later use */
Put some of the probe messages under bootverbose so to lessen the noise.
2004-09-12 03:19:32 +00:00
info = (struct aac_adapter_info *)&fib->data[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc->aac_revision = info->KernelRevision;
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
Put some of the probe messages under bootverbose so to lessen the noise.
2004-09-12 03:19:32 +00:00
if (bootverbose) {
device_printf(sc->aac_dev, "%s %dMHz, %dMB cache memory, %s\n",
aac_describe_code(aac_cpu_variant, info->CpuVariant),
info->ClockSpeed, info->BufferMem / (1024 * 1024),
aac_describe_code(aac_battery_platform,
info->batteryPlatform));
device_printf(sc->aac_dev,
"Kernel %d.%d-%d, Build %d, S/N %6X\n",
info->KernelRevision.external.comp.major,
info->KernelRevision.external.comp.minor,
info->KernelRevision.external.comp.dash,
info->KernelRevision.buildNumber,
(u_int32_t)(info->SerialNumber & 0xffffff));
Begin support for 64-bit address support and workarounds for newer cards: - Add data structuress for doing 64-bit scatter/gather - Move busdma tag creations around so that only the parent is created in aac_pci.c. - Retrieve the capabilities word from the firmware before setting up command structures and tags. This allows the driver to decide whether to do 64-bit commands, and if work-arounds are needed for systems with >2GB of RAM. - Only enable the SCSI passthrough if it's enabled in the capabilities word in the firmware. This should fix problems with the 2120S and 2200S cards in systems with more than 2GB of RAM. Full 64-bit support is forthcoming. MFC-After: 1 week
2003-03-26 17:50:11 +00:00
device_printf(sc->aac_dev, "Supported Options=%b\n",
sc->supported_options,
"\20"
"\1SNAPSHOT"
"\2CLUSTERS"
"\3WCACHE"
"\4DATA64"
"\5HOSTTIME"
"\6RAID50"
"\7WINDOW4GB"
"\10SCSIUPGD"
"\11SOFTERR"
"\12NORECOND"
"\13SGMAP64"
"\14ALARM"
"\15NONDASD");
}
Put some of the probe messages under bootverbose so to lessen the noise.
2004-09-12 03:19:32 +00:00
aac_release_sync_fib(sc);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Look up a text description of a numeric error code and return a pointer to
* same.
*/
static char *
aac_describe_code(struct aac_code_lookup *table, u_int32_t code)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
int i;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
for (i = 0; table[i].string != NULL; i++)
if (table[i].code == code)
return(table[i].string);
return(table[i + 1].string);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Management Interface
*/
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
aac_open(struct cdev *dev, int flags, int fmt, d_thread_t *td)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc = dev->si_drv1;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* Check to make sure the device isn't already open */
if (sc->aac_state & AAC_STATE_OPEN) {
return EBUSY;
}
sc->aac_state |= AAC_STATE_OPEN;
return 0;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
aac_close(struct cdev *dev, int flags, int fmt, d_thread_t *td)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_softc *sc;
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
sc = dev->si_drv1;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* Mark this unit as no longer open */
sc->aac_state &= ~AAC_STATE_OPEN;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return 0;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
aac_ioctl(struct cdev *dev, u_long cmd, caddr_t arg, int flag, d_thread_t *td)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
union aac_statrequest *as;
struct aac_softc *sc;
int error = 0;
Make aac(4) compile cleanly on 64-bit machines. The code was already 64-bit safe, but some (unneeded and/or harmless) downcasts were generating warnings. The driver still is not endian-clean.
2003-08-19 21:31:45 +00:00
uint32_t cookie;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
as = (union aac_statrequest *)arg;
sc = dev->si_drv1;
switch (cmd) {
case AACIO_STATS:
switch (as->as_item) {
case AACQ_FREE:
case AACQ_BIO:
case AACQ_READY:
case AACQ_BUSY:
bcopy(&sc->aac_qstat[as->as_item], &as->as_qstat,
sizeof(struct aac_qstat));
break;
default:
error = ENOENT;
break;
}
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
break;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
case FSACTL_SENDFIB:
arg = *(caddr_t*)arg;
case FSACTL_LNX_SENDFIB:
debug(1, "FSACTL_SENDFIB");
error = aac_ioctl_sendfib(sc, arg);
break;
case FSACTL_AIF_THREAD:
case FSACTL_LNX_AIF_THREAD:
debug(1, "FSACTL_AIF_THREAD");
error = EINVAL;
break;
case FSACTL_OPEN_GET_ADAPTER_FIB:
arg = *(caddr_t*)arg;
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
case FSACTL_LNX_OPEN_GET_ADAPTER_FIB:
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug(1, "FSACTL_OPEN_GET_ADAPTER_FIB");
/*
* Pass the caller out an AdapterFibContext.
*
* Note that because we only support one opener, we
* basically ignore this. Set the caller's context to a magic
* number just in case.
*
* The Linux code hands the driver a pointer into kernel space,
* and then trusts it when the caller hands it back. Aiee!
* Here, we give it the proc pointer of the per-adapter aif
* thread. It's only used as a sanity check in other calls.
*/
Make aac(4) compile cleanly on 64-bit machines. The code was already 64-bit safe, but some (unneeded and/or harmless) downcasts were generating warnings. The driver still is not endian-clean.
2003-08-19 21:31:45 +00:00
cookie = (uint32_t)(uintptr_t)sc->aifthread;
error = copyout(&cookie, arg, sizeof(cookie));
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
break;
case FSACTL_GET_NEXT_ADAPTER_FIB:
arg = *(caddr_t*)arg;
case FSACTL_LNX_GET_NEXT_ADAPTER_FIB:
debug(1, "FSACTL_GET_NEXT_ADAPTER_FIB");
error = aac_getnext_aif(sc, arg);
break;
case FSACTL_CLOSE_GET_ADAPTER_FIB:
case FSACTL_LNX_CLOSE_GET_ADAPTER_FIB:
debug(1, "FSACTL_CLOSE_GET_ADAPTER_FIB");
/* don't do anything here */
break;
case FSACTL_MINIPORT_REV_CHECK:
arg = *(caddr_t*)arg;
case FSACTL_LNX_MINIPORT_REV_CHECK:
debug(1, "FSACTL_MINIPORT_REV_CHECK");
error = aac_rev_check(sc, arg);
break;
case FSACTL_QUERY_DISK:
arg = *(caddr_t*)arg;
case FSACTL_LNX_QUERY_DISK:
debug(1, "FSACTL_QUERY_DISK");
error = aac_query_disk(sc, arg);
break;
case FSACTL_DELETE_DISK:
case FSACTL_LNX_DELETE_DISK:
/*
* We don't trust the underland to tell us when to delete a
* container, rather we rely on an AIF coming from the
* controller
*/
error = 0;
break;
default:
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
debug(1, "unsupported cmd 0x%lx\n", cmd);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
error = EINVAL;
break;
}
return(error);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
static int
Do the dreaded s/dev_t/struct cdev */ Bump __FreeBSD_version accordingly.
2004-06-16 09:47:26 +00:00
aac_poll(struct cdev *dev, int poll_events, d_thread_t *td)
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
{
struct aac_softc *sc;
int revents;
sc = dev->si_drv1;
revents = 0;
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_aifq_lock);
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
if ((poll_events & (POLLRDNORM | POLLIN)) != 0) {
if (sc->aac_aifq_tail != sc->aac_aifq_head)
revents |= poll_events & (POLLIN | POLLRDNORM);
}
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_aifq_lock);
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
if (revents == 0) {
if (poll_events & (POLLIN | POLLRDNORM))
selrecord(td, &sc->rcv_select);
}
return (revents);
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Send a FIB supplied from userspace
*/
static int
aac_ioctl_sendfib(struct aac_softc *sc, caddr_t ufib)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_command *cm;
int size, error;
debug_called(2);
cm = NULL;
/*
* Get a command
*/
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_io_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (aac_alloc_command(sc, &cm)) {
error = EBUSY;
goto out;
}
/*
* Fetch the FIB header, then re-copy to get data as well.
*/
if ((error = copyin(ufib, cm->cm_fib,
sizeof(struct aac_fib_header))) != 0)
goto out;
size = cm->cm_fib->Header.Size + sizeof(struct aac_fib_header);
if (size > sizeof(struct aac_fib)) {
Make aac(4) compile cleanly on 64-bit machines. The code was already 64-bit safe, but some (unneeded and/or harmless) downcasts were generating warnings. The driver still is not endian-clean.
2003-08-19 21:31:45 +00:00
device_printf(sc->aac_dev, "incoming FIB oversized (%d > %zd)\n",
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
size, sizeof(struct aac_fib));
size = sizeof(struct aac_fib);
}
if ((error = copyin(ufib, cm->cm_fib, size)) != 0)
goto out;
cm->cm_fib->Header.Size = size;
cm->cm_timestamp = time_second;
/*
* Pass the FIB to the controller, wait for it to complete.
*/
Remove the 'timeout' argument from aac_wait_command() as it isn't used and never will be. Update the XXX comment for this function to accurately reflect why things are the way they are.
2004-04-14 19:11:29 +00:00
if ((error = aac_wait_command(cm)) != 0) {
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
device_printf(sc->aac_dev,
"aac_wait_command return %d\n", error);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
goto out;
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Copy the FIB and data back out to the caller.
*/
size = cm->cm_fib->Header.Size;
if (size > sizeof(struct aac_fib)) {
Make aac(4) compile cleanly on 64-bit machines. The code was already 64-bit safe, but some (unneeded and/or harmless) downcasts were generating warnings. The driver still is not endian-clean.
2003-08-19 21:31:45 +00:00
device_printf(sc->aac_dev, "outbound FIB oversized (%d > %zd)\n",
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
size, sizeof(struct aac_fib));
size = sizeof(struct aac_fib);
}
error = copyout(cm->cm_fib, ufib, size);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
out:
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (cm != NULL) {
aac_release_command(cm);
}
Bring aac out from under Giant: - the mutex aac_io_lock protects the main codepaths which handle queues and hardware registers. Only one acquire/release is done in the top-half and the taskqueue. This mutex also applies to the userland command path and CAM data path. - Move the taskqueue to the new Giant-free version. - Register the disk device with DISKFLAG_NOGIANT so the top-half processing runs without Giant. - Move the dynamic command allocator to the worker thread to avoid locking issues with bus_dmamem_alloc(). This gives about 20% improvement in most of my benchmarks.
2003-02-26 04:46:21 +00:00
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_io_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(error);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Handle an AIF sent to us by the controller; queue it for later reference.
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
* If the queue fills up, then drop the older entries.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*/
static void
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
aac_handle_aif(struct aac_softc *sc, struct aac_fib *fib)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_aif_command *aif;
struct aac_container *co, *co_next;
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
struct aac_mntinfo *mi;
struct aac_mntinforesp *mir = NULL;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
u_int16_t rsize;
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
int next, found;
When scanning for changed containers, don't assume that the controller will respond in a sane manner. Thanks to Petri Helenius <pete@he.iki.fi> for spotting this and pestering me to fix it.
2003-06-03 02:10:55 +00:00
int count = 0, added = 0, i = 0;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
aif = (struct aac_aif_command*)&fib->data[0];
aac_print_aif(sc, aif);
/* Is it an event that we should care about? */
switch (aif->command) {
case AifCmdEventNotify:
switch (aif->data.EN.type) {
case AifEnAddContainer:
case AifEnDeleteContainer:
/*
* A container was added or deleted, but the message
* doesn't tell us anything else! Re-enumerate the
* containers and sort things out.
*/
Collapse sync fib locking into normal i/o locking. The former didn't protect the registers so it was trivially possible for a sync command and i/o command to fight each other and confuse the controller. Make the sync fib alloc/release functions inline and remove the somewhat worthless AAC_SYNC_LOCK_FORCE flag. Thanks to Adil Katchi for helping me to track this down in RELENG_4.
2004-06-02 18:15:48 +00:00
aac_alloc_sync_fib(sc, &fib);
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
mi = (struct aac_mntinfo *)&fib->data[0];
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
do {
/*
* Ask the controller for its containers one at
* a time.
* XXX What if the controller's list changes
* midway through this enumaration?
* XXX This should be done async.
*/
bzero out the sync command buffer when sending commands. This was causing problems when enumerating multiple arrays. This is an MFC candidate.
2002-05-03 00:07:50 +00:00
bzero(mi, sizeof(struct aac_mntinfo));
mi->Command = VM_NameServe;
mi->MntType = FT_FILESYS;
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
mi->MntCount = i;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
rsize = sizeof(mir);
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
if (aac_sync_fib(sc, ContainerCommand, 0, fib,
sizeof(struct aac_mntinfo))) {
When scanning for changed containers, don't assume that the controller will respond in a sane manner. Thanks to Petri Helenius <pete@he.iki.fi> for spotting this and pestering me to fix it.
2003-06-03 02:10:55 +00:00
printf("Error probing container %d\n",
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
i);
continue;
}
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
mir = (struct aac_mntinforesp *)&fib->data[0];
When scanning for changed containers, don't assume that the controller will respond in a sane manner. Thanks to Petri Helenius <pete@he.iki.fi> for spotting this and pestering me to fix it.
2003-06-03 02:10:55 +00:00
/* XXX Need to check if count changed */
count = mir->MntRespCount;
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Check the container against our list.
* co->co_found was already set to 0 in a
* previous run.
*/
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
if ((mir->Status == ST_OK) &&
(mir->MntTable[0].VolType != CT_NONE)) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
found = 0;
TAILQ_FOREACH(co,
&sc->aac_container_tqh,
co_link) {
if (co->co_mntobj.ObjectId ==
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
mir->MntTable[0].ObjectId) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
co->co_found = 1;
found = 1;
break;
}
}
/*
* If the container matched, continue
* in the list.
*/
if (found) {
i++;
continue;
}
/*
* This is a new container. Do all the
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
* appropriate things to set it up.
*/
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
aac_add_container(sc, mir, 1);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
added = 1;
}
i++;
When scanning for changed containers, don't assume that the controller will respond in a sane manner. Thanks to Petri Helenius <pete@he.iki.fi> for spotting this and pestering me to fix it.
2003-06-03 02:10:55 +00:00
} while ((i < count) && (i < AAC_MAX_CONTAINERS));
Prepare for a major update to the aac driver: Update the aac driver with the new crashdump api. Protect sync fibs with a mutex. Align all DMA buffers on a PAGE_SIZE boundary. MFC after: 3 days
2002-04-24 05:12:50 +00:00
aac_release_sync_fib(sc);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Go through our list of containers and see which ones
* were not marked 'found'. Since the controller didn't
* list them they must have been deleted. Do the
* appropriate steps to destroy the device. Also reset
* the co->co_found field.
*/
co = TAILQ_FIRST(&sc->aac_container_tqh);
while (co != NULL) {
if (co->co_found == 0) {
device_delete_child(sc->aac_dev,
co->co_disk);
co_next = TAILQ_NEXT(co, co_link);
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_container_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
TAILQ_REMOVE(&sc->aac_container_tqh, co,
co_link);
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_container_lock);
Change FREE() to free()
2004-08-12 05:09:17 +00:00
free(co, M_AACBUF);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
co = co_next;
} else {
co->co_found = 0;
co = TAILQ_NEXT(co, co_link);
}
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* Attach the newly created containers */
if (added)
bus_generic_attach(sc->aac_dev);
Fix mis-indentation. Approved by: scottl
2002-10-20 16:00:43 +00:00
break;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
default:
break;
}
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
default:
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
break;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/* Copy the AIF data to the AIF queue for ioctl retrieval */
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_aifq_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
next = (sc->aac_aifq_head + 1) % AAC_AIFQ_LENGTH;
if (next != sc->aac_aifq_tail) {
bcopy(aif, &sc->aac_aifq[next], sizeof(struct aac_aif_command));
Add functionality and fix bugs so the driver will work with soon-to-be released management apps. 1. Implement poll(). This will check for queued aif's so that a subsequent ioctl call to retrieve the next aif will not block. 2. Don't catch signals when sleeping on a fib sent from userland. This causes a race and panic due to the pthread context switcher waking up the tsleep at inopportune times. 3. Fix some whitespace nits. MFC after: 3 days
2001-12-02 03:47:33 +00:00
sc->aac_aifq_head = next;
/* On the off chance that someone is sleeping for an aif... */
if (sc->aac_state & AAC_STATE_AIF_SLEEPER)
wakeup(sc->aac_aifq);
/* Wakeup any poll()ers */
- Implement selwakeuppri() which allows raising the priority of a thread being waken up. The thread waken up can run at a priority as high as after tsleep(). - Replace selwakeup()s with selwakeuppri()s and pass appropriate priorities. - Add cv_broadcastpri() which raises the priority of the broadcast threads. Used by selwakeuppri() if collision occurs. Not objected in: -arch, -current
2003-11-09 09:17:26 +00:00
selwakeuppri(&sc->rcv_select, PRIBIO);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_aifq_lock);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
* Return the Revision of the driver to userspace and check to see if the
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
* userspace app is possibly compatible. This is extremely bogus since
* our driver doesn't follow Adaptec's versioning system. Cheat by just
* returning what the card reported.
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
*/
static int
Put the linux compat ioctl interface on death row. Add FreeBSD definitions of all of the ioctls, plus round out all ioctl definitions to match what exists for linux. Allow ioctls to be called through either the native or linux interface, though no apps exist (yet) that can take advantage of native calling.
2001-08-06 04:34:47 +00:00
aac_rev_check(struct aac_softc *sc, caddr_t udata)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_rev_check rev_check;
struct aac_rev_check_resp rev_check_resp;
int error = 0;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Copyin the revision struct from userspace
*/
if ((error = copyin(udata, (caddr_t)&rev_check,
Reformat for 80 columns. Sorry, but I had to do it.
2001-08-03 00:50:30 +00:00
sizeof(struct aac_rev_check))) != 0) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return error;
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug(2, "Userland revision= %d\n",
rev_check.callingRevision.buildNumber);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Doctor up the response struct.
*/
rev_check_resp.possiblyCompatible = 1;
rev_check_resp.adapterSWRevision.external.ul =
sc->aac_revision.external.ul;
rev_check_resp.adapterSWRevision.buildNumber =
sc->aac_revision.buildNumber;
return(copyout((caddr_t)&rev_check_resp, udata,
sizeof(struct aac_rev_check_resp)));
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
* Pass the caller the next AIF in their queue
*/
static int
Put the linux compat ioctl interface on death row. Add FreeBSD definitions of all of the ioctls, plus round out all ioctl definitions to match what exists for linux. Allow ioctls to be called through either the native or linux interface, though no apps exist (yet) that can take advantage of native calling.
2001-08-06 04:34:47 +00:00
aac_getnext_aif(struct aac_softc *sc, caddr_t arg)
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct get_adapter_fib_ioctl agf;
The aac driver has evolved enough over the last few months that it no longer resembles the 4.x version very much. Garbage collect the legacy bits.
2003-03-01 05:07:19 +00:00
int error;
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if ((error = copyin(arg, &agf, sizeof(agf))) == 0) {
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
* Check the magic number that we gave the caller.
*/
Make aac(4) compile cleanly on 64-bit machines. The code was already 64-bit safe, but some (unneeded and/or harmless) downcasts were generating warnings. The driver still is not endian-clean.
2003-08-19 21:31:45 +00:00
if (agf.AdapterFibContext != (int)(uintptr_t)sc->aifthread) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
error = EFAULT;
} else {
Put the linux compat ioctl interface on death row. Add FreeBSD definitions of all of the ioctls, plus round out all ioctl definitions to match what exists for linux. Allow ioctls to be called through either the native or linux interface, though no apps exist (yet) that can take advantage of native calling.
2001-08-06 04:34:47 +00:00
error = aac_return_aif(sc, agf.AifFib);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if ((error == EAGAIN) && (agf.Wait)) {
sc->aac_state |= AAC_STATE_AIF_SLEEPER;
while (error == EAGAIN) {
error = tsleep(sc->aac_aifq, PRIBIO |
PCATCH, "aacaif", 0);
if (error == 0)
error = aac_return_aif(sc,
agf.AifFib);
}
sc->aac_state &= ~AAC_STATE_AIF_SLEEPER;
}
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(error);
A new driver for PCI:SCSI RAID controllers based on the Adaptec FSA design. This includes integrated Dell RAID controllers, the Dell PERC 2/QC and the HP NetRAID-4M.
2000-09-13 03:20:35 +00:00
}
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
* Hand the next AIF off the top of the queue out to userspace.
*/
static int
Put the linux compat ioctl interface on death row. Add FreeBSD definitions of all of the ioctls, plus round out all ioctl definitions to match what exists for linux. Allow ioctls to be called through either the native or linux interface, though no apps exist (yet) that can take advantage of native calling.
2001-08-06 04:34:47 +00:00
aac_return_aif(struct aac_softc *sc, caddr_t uptr)
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
{
Fix a couple of bugs with AIF handling: - Correct the logic for the AIF array index pointers so that correct slot is always looked at. - Copy the full FIB payload size when copying AIF's, not just the first 64 bytes. Thanks to Mirapoint, Inc, for pointing these problems out and offering a solution.
2003-10-17 21:44:06 +00:00
int next, error;
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_aifq_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
if (sc->aac_aifq_tail == sc->aac_aifq_head) {
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_aifq_lock);
Fix a couple of bugs with AIF handling: - Correct the logic for the AIF array index pointers so that correct slot is always looked at. - Copy the full FIB payload size when copying AIF's, not just the first 64 bytes. Thanks to Mirapoint, Inc, for pointing these problems out and offering a solution.
2003-10-17 21:44:06 +00:00
return (EAGAIN);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
}
Fix a couple of bugs with AIF handling: - Correct the logic for the AIF array index pointers so that correct slot is always looked at. - Copy the full FIB payload size when copying AIF's, not just the first 64 bytes. Thanks to Mirapoint, Inc, for pointing these problems out and offering a solution.
2003-10-17 21:44:06 +00:00
next = (sc->aac_aifq_tail + 1) % AAC_AIFQ_LENGTH;
error = copyout(&sc->aac_aifq[next], uptr,
sizeof(struct aac_aif_command));
if (error)
device_printf(sc->aac_dev,
"aac_return_aif: copyout returned %d\n", error);
else
sc->aac_aifq_tail = next;
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_aifq_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
return(error);
Major bugfix and minor update. This should resolve the current issues with the driver locking up under load. - Restructure so that we use a static pool of commands/FIBs, rather than allocating them in clusters. The cluster allocation just made things more complicated, and allowed us to waste more memory in peak load situations. - Make queueing macros more like my other drivers. This adds queue stats for free. Add some debugging to take advantage of this. - Reimplement the periodic timeout scan. Kick the interrupt handler and the start routine every scan as well, just to be safe. Track busy commands properly. - Bring resource cleanup into line with resource allocation. We should now clean up correctly after a failed probe/unload/etc. - Try to start new commands when old ones are completed. We weren't doing this before, which could lead to deadlock when the controller was full. - Don't try to build a new command if we have found a deferred command. This could cause us to lose the deferred command. - Use diskerr() to report I/O errors. - Don't bail if the AdapterInfo structure is the wrong size. Some variation seems to be normal. We need to improve our handing of 2.x firmware sets. - Improve some comments in an attempt to try to make things clearer. - Restructure to avoid some warnings.
2000-12-27 13:14:56 +00:00
}
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
/*
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
* Give the userland some information about the container. The AAC arch
* expects the driver to be a SCSI passthrough type driver, so it expects
* the containers to have b:t:l numbers. Fake it.
*/
static int
aac_query_disk(struct aac_softc *sc, caddr_t uptr)
{
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
struct aac_query_disk query_disk;
struct aac_container *co;
struct aac_disk *disk;
int error, id;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
debug_called(2);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
disk = NULL;
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
error = copyin(uptr, (caddr_t)&query_disk,
sizeof(struct aac_query_disk));
if (error)
return (error);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
id = query_disk.ContainerNumber;
if (id == -1)
return (EINVAL);
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_lock(&sc->aac_container_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
TAILQ_FOREACH(co, &sc->aac_container_tqh, co_link) {
if (co->co_mntobj.ObjectId == id)
break;
}
Fix mis-indentation. Approved by: scottl
2002-10-20 16:00:43 +00:00
if (co == NULL) {
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
query_disk.Valid = 0;
query_disk.Locked = 0;
query_disk.Deleted = 1; /* XXX is this right? */
Fix mis-indentation. Approved by: scottl
2002-10-20 16:00:43 +00:00
} else {
disk = device_get_softc(co->co_disk);
query_disk.Valid = 1;
query_disk.Locked =
(disk->ad_flags & AAC_DISK_OPEN) ? 1 : 0;
query_disk.Deleted = 0;
query_disk.Bus = device_get_unit(sc->aac_dev);
query_disk.Target = disk->unit;
query_disk.Lun = 0;
query_disk.UnMapped = 0;
Move to 'struct disk*' API Submitted by: phk
2003-02-26 02:50:27 +00:00
sprintf(&query_disk.diskDeviceName[0], "%s%d",
Change the disk(9) API in order to make device removal more robust. Previously the "struct disk" were owned by the device driver and this gave us problems when the device disappared and the users of that device were not immediately disappearing. Now the struct disk is allocate with a new call, disk_alloc() and owned by geom_disk and just abandonned by the device driver when disk_create() is called. Unfortunately, this results in a ton of "s/\./->/" changes to device drivers. Since I'm doing the sweep anyway, a couple of other API improvements have been carried out at the same time: The Giant awareness flag has been flipped from DISKFLAG_NOGIANT to DISKFLAG_NEEDSGIANT A version number have been added to disk_create() so that we can detect, report and ignore binary drivers with old ABI in the future. Manual page update to follow shortly.
2004-02-18 21:36:53 +00:00
disk->ad_disk->d_name, disk->ad_disk->d_unit);
Fix mis-indentation. Approved by: scottl
2002-10-20 16:00:43 +00:00
}
Remove the AAC_LOCK macros. They no longer abstract anything and only obfuscate the code. No functional differences.
2004-08-12 05:05:06 +00:00
mtx_unlock(&sc->aac_container_lock);
Bring the aac driver *much* closer to style(9). Reviewed by: ken
2001-09-05 20:43:02 +00:00
error = copyout((caddr_t)&query_disk, uptr,
sizeof(struct aac_query_disk));
return (error);
Mega update to the aac driver. 1. Correctly handle commands initiated by the adapter. These commands are defered to a kthread responsible for their processing, then are properly returned to the controller. 2. Add/remove disk devices when notified by the card that a container was created/deleted/offline. 3. Implement crashdump functionality. 4. Support all ioctls needed for the management CLI to work. The linux version of this app can be found at the Dell or HP website. A native version will be forthcoming. MFC-after: 4.4-RELEASE
2001-08-29 23:34:05 +00:00
}
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
static void
aac_get_bus_info(struct aac_softc *sc)
{
struct aac_fib *fib;
struct aac_ctcfg *c_cmd;
struct aac_ctcfg_resp *c_resp;
struct aac_vmioctl *vmi;
struct aac_vmi_businf_resp *vmi_resp;
struct aac_getbusinf businfo;
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
struct aac_sim *caminf;
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
device_t child;
int i, found, error;
Collapse sync fib locking into normal i/o locking. The former didn't protect the registers so it was trivially possible for a sync command and i/o command to fight each other and confuse the controller. Make the sync fib alloc/release functions inline and remove the somewhat worthless AAC_SYNC_LOCK_FORCE flag. Thanks to Adil Katchi for helping me to track this down in RELENG_4.
2004-06-02 18:15:48 +00:00
aac_alloc_sync_fib(sc, &fib);
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
c_cmd = (struct aac_ctcfg *)&fib->data[0];
bzero out the sync command buffer when sending commands. This was causing problems when enumerating multiple arrays. This is an MFC candidate.
2002-05-03 00:07:50 +00:00
bzero(c_cmd, sizeof(struct aac_ctcfg));
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
c_cmd->Command = VM_ContainerConfig;
c_cmd->cmd = CT_GET_SCSI_METHOD;
c_cmd->param = 0;
error = aac_sync_fib(sc, ContainerCommand, 0, fib,
sizeof(struct aac_ctcfg));
if (error) {
device_printf(sc->aac_dev, "Error %d sending "
"VM_ContainerConfig command\n", error);
aac_release_sync_fib(sc);
return;
}
c_resp = (struct aac_ctcfg_resp *)&fib->data[0];
if (c_resp->Status != ST_OK) {
device_printf(sc->aac_dev, "VM_ContainerConfig returned 0x%x\n",
c_resp->Status);
aac_release_sync_fib(sc);
return;
}
sc->scsi_method_id = c_resp->param;
vmi = (struct aac_vmioctl *)&fib->data[0];
bzero out the sync command buffer when sending commands. This was causing problems when enumerating multiple arrays. This is an MFC candidate.
2002-05-03 00:07:50 +00:00
bzero(vmi, sizeof(struct aac_vmioctl));
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
vmi->Command = VM_Ioctl;
vmi->ObjType = FT_DRIVE;
vmi->MethId = sc->scsi_method_id;
vmi->ObjId = 0;
vmi->IoctlCmd = GetBusInfo;
error = aac_sync_fib(sc, ContainerCommand, 0, fib,
sizeof(struct aac_vmioctl));
if (error) {
device_printf(sc->aac_dev, "Error %d sending VMIoctl command\n",
error);
aac_release_sync_fib(sc);
return;
}
vmi_resp = (struct aac_vmi_businf_resp *)&fib->data[0];
if (vmi_resp->Status != ST_OK) {
device_printf(sc->aac_dev, "VM_Ioctl returned %d\n",
vmi_resp->Status);
aac_release_sync_fib(sc);
return;
}
bcopy(&vmi_resp->BusInf, &businfo, sizeof(struct aac_getbusinf));
aac_release_sync_fib(sc);
found = 0;
for (i = 0; i < businfo.BusCount; i++) {
if (businfo.BusValid[i] != AAC_BUS_VALID)
continue;
Free the container and sim objects on detach. Convert several uses of MALLOC to malloc.
2003-02-06 00:37:37 +00:00
caminf = (struct aac_sim *)malloc( sizeof(struct aac_sim),
M_AACBUF, M_NOWAIT | M_ZERO);
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
if (caminf == NULL)
continue;
child = device_add_child(sc->aac_dev, "aacp", -1);
if (child == NULL) {
device_printf(sc->aac_dev, "device_add_child failed\n");
continue;
}
caminf->TargetsPerBus = businfo.TargetsPerBus;
caminf->BusNumber = i;
caminf->InitiatorBusId = businfo.InitiatorBusId[i];
caminf->aac_sc = sc;
Teach the CAM module how to deregister itself so it can be unloaded.
2003-02-06 03:31:02 +00:00
caminf->sim_dev = child;
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
device_set_ivars(child, caminf);
device_set_desc(child, "SCSI Passthrough Bus");
Various cleanups: - Move the command timeout check from a separate repeating timeout to the kthread since the kthread is already running periodically. - Move printing the hardware print buffer to the kthread. - Properly shut down the kernel thread on detach. - Detach the child array devices on detach. - Don't issue a controller halt command on detach. Doing so requires a PCI reset to wake the controller back up. The driver can now be unloaded as long as CAM support is not enabled.
2003-02-05 23:24:25 +00:00
TAILQ_INSERT_TAIL(&sc->aac_sim_tqh, caminf, sim_link);
Add a CAM interface to the aac driver. This is useful in case you should ever connect a SCSI Cdrom/Tape/Jukebox/Scanner/Printer/kitty-litter-scooper to your high-end RAID controller. The interface to the arrays is still via the block interface; this merely provides a way to circumvent the RAID functionality and access the SCSI buses directly. Note that for somewhat obvious reasons, hard drives are not exposed to the da driver through this interface, though you can still talk to them via the pass driver. Be the first on your block to low-level format unsuspecting drives that are part of an array! To enable this, add the 'aacp' device to your kernel config. MFC after: 3 days
2002-04-27 01:31:17 +00:00
found = 1;
}
if (found)
bus_generic_attach(sc->aac_dev);
return;
}
Reference in New Issue Copy Permalink