Since 2.6.25-rc7, I've been seeing an occasional livelock on one x86_64
machine, copying kernel trees to tmpfs, paging out to swap.
Signature: 6000 pages under writeback but never getting written; most
tasks of interest trying to reclaim, but each get_swap_bio waiting for a
bio in mempool_alloc's io_schedule_timeout(5*HZ); every five seconds an
atomic page allocation failure report from kblockd failing to allocate a
sense_buffer in __scsi_get_command.
__scsi_get_command has a (one item) free_list to protect against this,
but rc1's [SCSI] use dynamically allocated sense buffer
de25deb180 upset that slightly. When it
fails to allocate from the separate sense_slab, instead of giving up, it
must fall back to the command free_list, which is sure to have a
sense_buffer attached.
Either my earlier -rc testing missed this, or there's some recent
contributory factor. One very significant factor is SLUB, which merges
slab caches when it can, and on 64-bit happens to merge both bio cache
and sense_slab cache into kmalloc's 128-byte cache: so that under this
swapping load, bios above are liable to gobble up all the slots needed
for scsi_cmnd sense_buffers below.
That's disturbing behaviour, and I tried a few things to fix it. Adding
a no-op constructor to the sense_slab inhibits SLUB from merging it, and
stops all the allocation failures I was seeing; but it's rather a hack,
and perhaps in different configurations we have other caches on the
swapout path which are ill-merged.
Another alternative is to revert the separate sense_slab, using
cache-line-aligned sense_buffer allocated beyond scsi_cmnd from the one
kmem_cache; but that might waste more memory, and is only a way of
diverting around the known problem.
While I don't like seeing the allocation failures, and hate the idea of
all those bios piled up above a scsi host working one by one, it does
seem to emerge fairly soon with the livelock fix. So lacking better
ideas, stick with that one clear fix for now.
Signed-off-by: Hugh Dickins <hugh@veritas.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Jens Axboe <jens.axboe@oracle.com>
Cc: Christoph Lameter <clameter@sgi.com>
Cc: Pekka Enberg <penberg@cs.helsinki.fi>
Cc: Peter Zijlstra <a.p.ziljstra@chello.nl>
Cc: Rafael J. Wysocki <rjw@sisk.pl>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
add support for mv6480 chip which subsystem id is 6480 in spite of device id is 6440.
Signed-off-by: Ke Wei <kewei@marvell.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
removed unused code and attached SATA address makes use of port id.
enable HBA interrupt after calling sas_register_ha();
Signed-off-by: Ke Wei <kewei@marvell.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
fix the buffer of rx DMA overflow bug.
fix default queue depth.
Signed-off-by: Ke Wei <kewei@marvell.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
if unplugged, driver's queuecommand function will return SAS_PHY_DOWN.
task->lldd_task is used for saving its slot info.
Signed-off-by: Ke Wei <kewei@marvell.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
When a slot is busy, we will not free this slot until slot reset is
completed. When unplugged the disk, we should release all command
tasks with unplugged port that have been sent.
If MVS_USE_TASKLET is defined, we can enable tasklet. Default is off.
Signed-off-by: Ke Wei <kewei@marvell.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
add a new tag handler to create slot num. When a slot num is busy, new
task can't hit this bit which was already used. plumb in phy speeds.
Signed-off-by: Ke Wei <kewei@marvell.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
and remove some unused members from struct.
Signed-off-by: Ke Wei <kewei@marvell.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
We give a very cryptic error if an ATA device is seen on a SAS port
but libsas isn't compiled to include libata to handle them. Add an
extra warning to explain specifically what the problem is.
Acked-by: Darrick J. Wong <djwong@us.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
This patch corrects some cases in scsi_add_host() that fail, but the "error"
return code was not reset after a prior use which set it to a non-error value.
Patch cut against scsi-rc-fixes-2.6
Signed-off-by: James Smart <james.smart@emulex.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
This reverts commit 4b6f5b3a99.
bsg takes a reference to the underlying generic device, so it's
impossible to unregister bsg in the device release routine.
Acked-by: FUJITA Tomonori <tomof@acm.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
This patch is only compile tested.
It seems that bitmap lookup routine for allocation_map in
a100u2w driver is simply wrong.
It cannot lookup more than first 32 bits. If all first 32 bits
are set, it just returns 33-th orc_scb even though the 33-th bit
is not set.
Signed-off-by: Akinobu Mita <akinobu.mita@gmail.com>
Tested-by: Alan Cox <alan@redhat.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Commit:
a341cd0f (SCSI: add asynchronous event notification API)
breaks:
285e9670 (sr,sd: send media state change modification events)
by introducing an event filter, which is removed here, to make
events, we are depending on, happen again.
Fix this by removing the event filter. It's pretty much broken at the
moment, since a user can't set it (the attribute being read only). A
proper fix will be to make the event discriminator distinguish between
AN and Polled media change events.
Cc: David Zeuthen <david@fubar.dk>
Cc: kristen accardi <kaccardi@gmail.com>
Cc: Jeff Garzik <jeff@garzik.org>
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Initialize the "state changed" flag, so we do not send a change event
immediately after registering a new device.
Signed-off-by: Kay Sievers <kay.sievers@vrfy.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Fix NULL pointer dereference during execution of Internal commands,
where gdth only allocates scp, but not scp->sense_buffer. The rest of
the code assumes that sense_buffer is allocated, which leads to a kernel
oops e.g. on reboot (during cache flush).
Signed-off-by: Sven Schnelle <svens@stackframe.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The Type B Adapter teardown does iounmap on pointers subtracted by a
constant offset. Since the offset is in bytes, we need the pointers to
be of type void * not uint32_t * so the subtraction is done in the
correct units and we iounmap the correct area.
Signed-off-by: Nick Cheng <nick.cheng@areca.com.tw>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Delay bsg unregistration, because we want to wait until all the request/cmds
have released their reference.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
We have several drivers that don't list SCSI as a dependency in
Kconfig. That leads to them potentially being selected as Y even if
SCSI is M (which will produce a build failure). Fix this by making
the if SCSI_LOWLEVEL that goes around all the drivers a tristate
forcing them all automatically to inherit the value of SCSI.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
buf[i] can be up to 0xfd, so doubling it and assigning the result to an
unsigned char truncates the value. Just use an unsigned int instead;
it's only a temporary.
Signed-off-by: Matthew Wilcox <willy@linux.intel.com>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
For qla4xxx, we could be starting a session, but some error (network,
target, IO from a device that got started, etc) could cause the session
to fail and curring the block/unblock and state manipulation could race
with each other. This patch just has those operations done in the
single threaded iscsi eh work queue, so that way they are serialized.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
We are seeing EXIST errors from sysfs during device addition.
We need a start scan callout so we do not start scanning sessions
found during hba setup, before the async scsi scan code is ready.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: David C Somayajulu <david.somayajulu@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The host reset callout could be starting to reset the hba at the same
time the dpc thread is. This creates lots of problems because they both
want to do wierd things with the firmware and interrupts, etc.
This patch just has the host reset function fully shutdown the dpc
thread before resetting the hba.
This patch also moves the setting of the session online bit to fix
a potential race with the dpc thread and iscsi recovery thread.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: David C Somayajulu <david.somayajulu@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Original patch from Tejun Heo <htejun@gmail.com> but should use ->extra_len
and not ->data_len, as we would then overshoot the original request size.
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
drivers/scsi/ibmvscsi/ibmvstgt.c: In function 'ibmvstgt_cmd_done':
drivers/scsi/ibmvscsi/ibmvstgt.c:292: error: 'cmd' undeclared (first use in this function)
drivers/scsi/ibmvscsi/ibmvstgt.c:292: error: (Each undeclared identifier is reported only once
drivers/scsi/ibmvscsi/ibmvstgt.c:292: error: for each function it appears in.)
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
scsi_tgt uses REQ_TYPE_BLOCK_PC so scsi_init_io doesn't set the length
for us. scsi_tgt needs to do it by itself.
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The scsi midlayer allocates scsi_cmnd->sense_buffer dynamically so we
can't initialize scsi_cmnd (the midlyaer does for us).
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Before calling __scsi_alloc_queue, scsi_host->shost_gendev.parent must
be initialized properly.
This patch moves __scsi_alloc_queue after scsi_add_host (like
initiator drivers do).
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Fix docbook problem in SCSI source files.
These cause the generated docbook to be incorrect.
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The original code would incorrectly hardcode ELS timeout values
rather than using the traditional '2 * r_a_tov' value. In some
cases, the hardcoded values would be larger than the
mailbox-command-timeout and result in a needless BIG_HAMMER (ISP
reset), the typical recovery mechanism employed in such cases.
The second defect in the original code was in the assignment of
the default 'ha->r_a_tov' to twice the traditional value.
Correct this by setting the value to 10 seconds.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
For recent ISPs, software must detect OVERRUN conditions by
checking the SS_RESIDUAL_OVER bit during CS_COMPLETE handling.
Update the driver to perform this check, which is consistent with
what earlier firmwares did by explicitly cracking open the
FCP_RSP statuses and returning an CS_DATA_OVERRUN.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
There's no point in hitting the RISC with what will most
assuredly be an unsucessful reset of the RISC hardware if the
initial stop-firmware mailbox command fails with a time-out
status. Instead, to avoid what could amount to a lengthy
stop-firmware/detect-failure/reset-risc loop, continue with
driver unloading and discard the stop-firmware requirement.
Signed-off-by: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The fix up from Daniel Drake for replacing GFP_DMA with something
more sensible has gone in here:
commit 69e562c234
Author: Daniel Drake <dsd@gentoo.org>
Date: Wed Feb 20 13:29:05 2008 +0000
[SCSI] arcmsr: fix message allocation
add a change log and update the version for this.
Signed-off-by: Nick Cheng <nick.cheng@areca.com.tw>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
ps3rom does:
scsi_for_each_sg(cmd, sgpnt, scsi_sg_count(cmd), k) {
kaddr = kmap_atomic(sg_page(sgpnt), KM_IRQ0);
We cannot do something like that with the clustering enabled (or we
can use scsi_kmap_atomic_sg).
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
sg driver rounds up the length in struct scatterlist to be a multiple
of 512 in some conditions. So LLDs can't use the data length in a sg
list to calculate residual. Instead, the length in struct scsi_cmnd
should be used.
[Geert: the variable buflen already contains scsi_bufflen(cmd)]
Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Signed-off-by: Geert Uytterhoeven <Geert.Uytterhoeven@sonycom.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Fix docbook problem in SCSI source files.
These cause the generated docbook to be incorrect.
Signed-off-by: Randy Dunlap <randy.dunlap@oracle.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The phy sas address is showing wrongly (wrong endianness). Fix up the
endian transforms to make this correct.
Signed-off-by: Ke Wei <kewei@marvell.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The recent patch named:
[SCSI] gdth: !use_sg cleanup and use of scsi accessors
has done a bad job in handling internal commands issued by gdth_execute().
Internal commands are issued with device gdth_cmd_str ready made directly
to the card, without any mapping or translations of scsi commands. So here
I added a gdth_cmd_str pointer to the gdth_cmndinfo private structure which
is then copied directly to host.
following this patch is a cleanup that removes the home cooked accessors
and reverts them to regular scsi_cmnd accessors. Since they are not used
anymore. After review maybe the 2 patches should be squashed together.
FIXME: There is still a problem with gdth_get_info(). as reported there
is a WARN_ON trigerd in dma_free_coherent() when doing:
$ cat /proc/sys/gdth/0
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Tested-by: Joerg Dorchain: <joerg@dorchain.net>
Tested-by: Stefan Priebe <s.priebe@allied-internet.ag>
Tested-by: Jon Chelton <jchelton@ffpglobal.com>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
gdth_exit would first remove all cards then stop the timer
and would not sync with the timer function. This caused a crash
in gdth_timer() when module was unloaded.
So del_timer_sync the timer before we delete the cards.
also the reboot notifier function would crash. So clean
that up and fix the crashes.
Signed-off-by: Boaz Harrosh <bharrosh@panasas.com>
Tested-by: Joerg Dorchain: <joerg@dorchain.net>
Tested-by: Stefan Priebe <s.priebe@allied-internet.ag>
Tested-by: Jon Chelton <jchelton@ffpglobal.com>
Cc: Stable Tree <stable@kernel.org>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
The old tools did not set max session cmds. This is a regression.
I removed the check when merging the power of 2 patch.
Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
This is a particularly nasty bug. The problem is that if any internal
ascb times out, currently we free it even though it's pending at the
sequencer. This results in the sequencer getting terminally confused
and the error message:
BUG:sequencer:dl:no ascb
Being returned when it comes back. The way to fix this is to manage
freeing the ascb from the tasklet completion routine, so that we only
free it when the sequencer actually returns it. The code is also
altered to use on stack completions and transfer variables.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
- Correct one use after free of the sas task
- update the reset required path to move straight to LUN reset
- make the bigger hammer actually reset something instead of just trying
to clear all the tasks.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Once the phy reset is plumbed in properly, SATA error handling fails
nastily because we change the port attached_sas_address using the WWN
field of the IDENTIFY message. This is a nice thing to do in theory,
but it really destroys hotplug because any event on the port causes an
automatic mismatch between the sas_address the phy just picked up and
the one we propagate into the port. However ugly they are, we have to
stick with the sas addresses made up by the phys and expanders.
Also does a few cosmetic changes to the way port printing is done to
make it clearer how a port is formed.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
Currently aic94xx has no exported I_T_nexus_reset function. This is a
bit of a huge problem, since sas_ata relies on this function to
perform an ATA phy reset and also it means that if abort fails, we
really have no bigger hammer to hit everything with.
Plumb in the I_T_nexus_reset by quiescing the sequencer, sending the
correct phy reset (link for ATA and hard for SAS) and then carefully
resuming the sequencer again.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
This is needed by the to be added I_T reset function in aic94xx. It
needs to know the local phy so it can send a link or hard reset along
the path.
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
mvsas.c picked up execute permissions. Move it back to being a plane
old file.
James Bottomley <James.Bottomley@HansenPartnership.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
During the last step of hibernation in the "platform" mode (with the
help of ACPI) we use the suspend code, including the devices'
->suspend() methods, to prepare the system for entering the ACPI S4
system sleep state.
But at least for some devices the operations performed by the
->suspend() callback in that case must be different from its operations
during regular suspend.
For this reason, introduce the new PM event type PM_EVENT_HIBERNATE and
pass it to the device drivers' ->suspend() methods during the last phase
of hibernation, so that they can distinguish this case and handle it as
appropriate. Modify the drivers that handle PM_EVENT_SUSPEND in a
special way and need to handle PM_EVENT_HIBERNATE in the same way.
These changes are necessary to fix a hibernation regression related
to the i915 driver (ref. http://lkml.org/lkml/2008/2/22/488).
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Acked-by: Pavel Machek <pavel@ucw.cz>
Tested-by: Jeff Chua <jeff.chua.linux@gmail.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Hugh Dickins