Dan Williams [Sat, 23 Jul 2011 18:44:25 +0000 (20:44 +0200)]
block: strict rq_affinity
Some systems benefit from completions always being steered to the strict
requester cpu rather than the looser "per-socket" steering that
blk_cpu_to_group() attempts by default. This is because the first
CPU in the group mask ends up being completely overloaded with work,
while the others (including the original submitter) has power left
to spare.
Allow the strict mode to be set by writing '2' to the sysfs control
file. This is identical to the scheme used for the nomerges file,
where '2' is a more aggressive setting than just being turned on.
echo 2 > /sys/block/<bdev>/queue/rq_affinity
Cc: Christoph Hellwig <hch@infradead.org> Cc: Roland Dreier <roland@purestorage.com> Tested-by: Dave Jiang <dave.jiang@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
backing-dev: use synchronize_rcu_expedited instead of synchronize_rcu
backing-dev: use synchronize_rcu_expedited instead of synchronize_rcu
synchronize_rcu sleeps several timer ticks. synchronize_rcu_expedited is
much faster.
With 100Hz timer frequency, when we remove 10000 block devices with
"dmsetup remove_all" command, it takes 27 minutes. With this patch,
removing 10000 block devices takes only 15 seconds.
Shaohua Li [Tue, 12 Jul 2011 12:24:56 +0000 (14:24 +0200)]
CFQ: add think time check for group
Currently when the last queue of a group has no request, we don't expire
the queue to hope request from the group comes soon, so the group doesn't
miss its share. But if the think time is big, the assumption isn't correct
and we just waste bandwidth. In such case, we don't do idle.
patched base
test1 64k 39k
test2 548k 540k
total 604k 578k
group1 gets much better throughput because it waits less time.
To check if the patch changes behavior of queue without think time. I also
tried to give test1 2ms think time or no think time. The test result is stable.
The thoughput doesn't change with/without the patch.
Shaohua Li [Tue, 12 Jul 2011 12:24:55 +0000 (14:24 +0200)]
CFQ: add think time check for service tree
Currently when the last queue of a service tree has no request, we don't
expire the queue to hope request from the service tree comes soon, so the
service tree doesn't miss its share. But if the think time is big, the
assumption isn't correct and we just waste bandwidth. In such case, we
don't do idle.
patched base
test1 41k/s 33k/s
test2 15868k/s 15789k/s
total 15902k/s 15817k/s
A slightly better
To check if the patch changes behavior of queue without think time. I also
tried to give test1 2ms think time or no think time. The test has variation
even without the patch, but the average throughput doesn't change with/without
the patch.
Shaohua Li [Tue, 12 Jul 2011 12:24:35 +0000 (14:24 +0200)]
CFQ: move think time check variables to a separate struct
Move the variables to do think time check to a sepatate struct. This is
to prepare adding think time check for service tree and group. No
functional change.
fs_excl is a poor man's priority inheritance for filesystems to hint to
the block layer that an operation is important. It was never clearly
specified, not widely adopted, and will not prevent starvation in many
cases (like across cgroups).
fs_excl was introduced with the time sliced CFQ IO scheduler, to
indicate when a process held FS exclusive resources and thus needed
a boost.
It doesn't cover all file systems, and it was never fully complete.
Lets kill it.
There is no consistency among filesystems from what bios (or requests)
are marked as being metadata. It's interesting to expose this in traces,
but we shouldn't schedule the requests differently based on whether or
not they're marked as being metadata.
Shaohua Li [Fri, 8 Jul 2011 06:19:21 +0000 (08:19 +0200)]
block: document blk_plug list access
I'm often confused why not disable preempt when changing blk_plug list. It
would be better to add comments here in case others have the similar concerns.
Signed-off-by: Shaohua Li <shaohua.li@intel.com> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Shaohua Li [Fri, 8 Jul 2011 06:19:20 +0000 (08:19 +0200)]
block: avoid building too big plug list
When I test fio script with big I/O depth, I found the total throughput drops
compared to some relative small I/O depth. The reason is the thread accumulates
big requests in its plug list and causes some delays (surely this depends
on CPU speed).
I thought we'd better have a threshold for requests. When a threshold reaches,
this means there is no request merge and queue lock contention isn't severe
when pushing per-task requests to queue, so the main advantages of blk plug
don't exist. We can force a plug list flush in this case.
With this, my test throughput actually increases and almost equals to small
I/O depth. Another side effect is irq off time decreases in blk_flush_plug_list()
for big I/O depth.
The BLK_MAX_REQUEST_COUNT is choosen arbitarily, but 16 is efficiently to
reduce lock contention to me. But I'm open here, 32 is ok in my test too.
Signed-off-by: Shaohua Li <shaohua.li@intel.com> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
Mike Snitzer [Wed, 6 Jul 2011 19:30:50 +0000 (21:30 +0200)]
block: eliminate potential for infinite loop in blkdev_issue_discard
Due to the recently identified overflow in read_capacity_16() it was
possible for max_discard_sectors to be zero but still have discards
enabled on the associated device's queue.
Eliminate the possibility for blkdev_issue_discard to infinitely loop.
Interestingly this issue wasn't identified until a device, whose
discard_granularity was 0 due to read_capacity_16 overflow, was consumed
by blk_stack_limits() to construct limits for a higher-level DM
multipath device. The multipath device's resulting limits never had the
discard limits stacked because blk_stack_limits() will only do so if
the bottom device's discard_granularity != 0. This resulted in the
multipath device's limits.max_discard_sectors being 0.
Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
On Linux x86_64 host with 32bit userspace, running
qemu or even just "qemu-img create -f qcow2 some.img 1G"
causes a kernel warning:
ioctl32(qemu-img:5296): Unknown cmd fd(3) cmd(00005326){t:'S';sz:0} arg(7fffffff) on some.img
ioctl32(qemu-img:5296): Unknown cmd fd(3) cmd(801c0204){t:02;sz:28} arg(fff77350) on some.img
ioctl 00005326 is CDROM_DRIVE_STATUS,
ioctl 801c0204 is FDGETPRM.
The warning appears because the Linux compat-ioctl handler for these
ioctls only applies to block devices, while qemu also uses the ioctls on
plain files.
block: flush MEDIA_CHANGE from drivers on close(2)
Currently, only open(2) is defined as the 'clearing' point. It has
two roles - first, it's an acknowledgement from userland indicating
that the event has been received and kernel can clear pending states
and proceed to generate more events. Secondly, it's passed on to
device drivers as a hint indicating that a synchronization point has
been reached and it might want to take a deeper look at the device.
The latter currently is only used by sr which uses two different
mechanisms - GET_EVENT_MEDIA_STATUS_NOTIFICATION and TEST_UNIT_READY
to discover events, where the former is lighter weight and safe to be
used repeatedly but may not provide full coverage. Among other
things, GET_EVENT can't detect media removal while TUR can.
This patch makes close(2) - blkdev_put() - indicate clearing hint for
MEDIA_CHANGE to drivers. disk_check_events() is renamed to
disk_flush_events() and updated to take @mask for events to flush
which is or'd to ev->clearing and will be passed to the driver on the
next ->check_events() invocation.
This change makes sr generate MEDIA_CHANGE when media is ejected from
userland - e.g. with eject(1).
Note: Given the current usage, it seems @clearing hint is needlessly
complex. disk_clear_events() can simply clear all events and the hint
can be boolean @flush.
Lars Ellenberg [Tue, 28 Jun 2011 11:22:48 +0000 (13:22 +0200)]
drbd: we should write meta data updates with FLUSH FUA
We used to write these with BIO_RW_BARRIER aka REQ_HARDBARRIER (unless
disabled in the configuration). The correct semantic now would be to
write with FLUSH/FUA.
For example, with activity log transactions, FUA alone is not enough, we
need the corresponding bitmap update (and all related application
updates) on stable storage as well.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com> Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Lars Ellenberg [Mon, 20 Jun 2011 12:44:45 +0000 (14:44 +0200)]
drbd: when receive times out on meta socket, also check last receive time on data socket
If we have an asymetrically congested network, we may send P_PING,
but due to congestion, the corresponding P_PING_ACK would time out,
and we would drop a (congested, but otherwise) healthy connection
("PingAck did not arrive in time.")
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com> Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Lars Ellenberg [Tue, 14 Jun 2011 12:18:23 +0000 (14:18 +0200)]
drbd: account bitmap IO during resync as resync-(related-)-io
If we have a good resync rate, we will frequently update the on-disk
bitmap, which, if not accounted for as resync io, may let an otherwise
idle device appear to be "busy", and cause us to throttle resync.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com> Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Lars Ellenberg [Mon, 6 Jun 2011 09:31:42 +0000 (11:31 +0200)]
drbd: don't cond_resched_lock with IRQs disabled
The last commit, drbd: add missing spinlock to bitmap receive,
introduced a cond_resched_lock(), where the lock in question is taken
with irqs disabled.
As we must not schedule with IRQs disabled,
and cond_resched_lock_irq() does not exist, yet,
we re-aquire the spin_lock_irq() for each bitmap page processed in turn.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com> Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Lars Ellenberg [Fri, 3 Jun 2011 19:18:13 +0000 (21:18 +0200)]
drbd: add missing spinlock to bitmap receive
During bitmap exchange, when using the RLE bitmap compression scheme,
we have a code path that can set the whole bitmap at once.
To avoid holding spin_lock_irq() for too long, we used to lock out other
bitmap modifications during bitmap exchange by other means, and then,
knowing we have exclusive access to the bitmap, modify it without
the spinlock, and with IRQs enabled.
Since we now allow local IO to continue, potentially setting additional
bits during the bitmap receive phase, this is no longer true, and we get
uncoordinated updates of bitmap members, causing bm_set to no longer
accurately reflect the total number of set bits.
To actually see this, you'd need to have a large bitmap, use RLE bitmap
compression, and have busy IO during sync handshake and bitmap exchange.
Fix this by taking the spin_lock_irq() in this code path as well, but
calling cond_resched_lock() after each page worth of bits processed.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com> Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>
Linus Torvalds [Mon, 27 Jun 2011 02:39:22 +0000 (19:39 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs-2.6:
cifs: propagate errors from cifs_get_root() to mount(2)
cifs: tidy cifs_do_mount() up a bit
cifs: more breakage on mount failures
cifs: close sget() races
cifs: pull freeing mountdata/dropping nls/freeing cifs_sb into cifs_umount()
cifs: move cifs_umount() call into ->kill_sb()
cifs: pull cifs_mount() call up
sanitize cifs_umount() prototype
cifs: initialize ->tlink_tree in cifs_setup_cifs_sb()
cifs: allocate mountdata earlier
cifs: leak on mount if we share superblock
cifs: don't pass superblock to cifs_mount()
cifs: don't leak nls on mount failure
cifs: double free on mount failure
take bdi setup/destruction into cifs_mount/cifs_umount
Linus Torvalds [Sat, 25 Jun 2011 14:23:27 +0000 (07:23 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6:
ALSA: Remove unneeded version.h includes from sound/
ASoC: pxa-ssp: Correct check for stream presence
ASoC: imx: add missing module informations
ASoC: imx: Remove unused Kconfig SND_MXC_SOC_SSI entry
ALSA: HDA: Pinfix quirk for HP Z200 Workstation
ALSA: VIA HDA: Create a master amplifier control for VT1718S.
ALSA: VIA HDA: Mute/unmute mixer conncted to Headphone for VT1718S.
ALSA: VIA HDA: Modify initial verbs list for VT1718S.
ALSA: hda - Remove ALC268 model override for CPR2000
ALSA: HDA: Remove quirk for an HP device
ASoC: Remove unused and about to be broken SND_SOC_CUSTOM I/O bus
Linus Torvalds [Sat, 25 Jun 2011 05:01:40 +0000 (22:01 -0700)]
Merge branch 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/keithp/linux-2.6
* 'drm-intel-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/keithp/linux-2.6:
drm/i915: save/resume forcewake lock fixes
Revert "drm/i915: Kill GTT mappings when moving from GTT domain"
drm/i915: Apply HWSTAM workaround for BSD ring on SandyBridge
drm/i915: Call intel_enable_plane from i9xx_crtc_mode_set (again)
Al Viro [Fri, 17 Jun 2011 13:50:44 +0000 (09:50 -0400)]
cifs: more breakage on mount failures
if cifs_get_root() fails, we end up with ->mount() returning NULL,
which is not what callers expect. Moreover, in case of superblock
reuse we end up leaking a superblock reference...
Acked-by: Pavel Shilovsky <piastryyy@gmail.com> Reviewed-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Al Viro [Fri, 17 Jun 2011 13:32:10 +0000 (09:32 -0400)]
cifs: move cifs_umount() call into ->kill_sb()
instead of calling it manually in case if cifs_read_super() fails
to set ->s_root, just call it from ->kill_sb(). cifs_put_super()
is gone now *and* we have cifs_sb shutdown and destruction done
after the superblock is gone from ->s_instances.
Acked-by: Pavel Shilovsky <piastryyy@gmail.com> Reviewed-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Al Viro [Fri, 17 Jun 2011 13:05:48 +0000 (09:05 -0400)]
cifs: don't pass superblock to cifs_mount()
To close sget() races we'll need to be able to set cifs_sb up before
we get the superblock, so we'll want to be able to do cifs_mount()
earlier. Fortunately, it's easy to do - setting ->s_maxbytes can
be done in cifs_read_super(), ditto for ->s_time_gran and as for
putting MS_POSIXACL into ->s_flags, we can mirror it in ->mnt_cifs_flags
until cifs_read_super() is called. Kill unused 'devname' argument,
while we are at it...
Acked-by: Pavel Shilovsky <piastryyy@gmail.com> Reviewed-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Al Viro [Fri, 17 Jun 2011 12:49:12 +0000 (08:49 -0400)]
cifs: don't leak nls on mount failure
if cifs_sb allocation fails, we still need to drop nls we'd stashed
into volume_info - the one we would've copied to cifs_sb if we could
allocate the latter.
Acked-by: Pavel Shilovsky <piastryyy@gmail.com> Reviewed-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Al Viro [Fri, 17 Jun 2011 12:34:57 +0000 (08:34 -0400)]
cifs: double free on mount failure
if we get to out_super with ->s_root already set (e.g. with
cifs_get_root() failure), we'll end up with cifs_put_super()
called and ->mountdata freed twice. We'll also get cifs_sb
freed twice and cifs_sb->local_nls dropped twice. The problem
is, we can get to out_super both with and without ->s_root,
which makes ->put_super() a bad place for such work.
Switch to ->kill_sb(), have all that work done there after
kill_anon_super(). Unlike ->put_super(), ->kill_sb() is
called by deactivate_locked_super() whether we have ->s_root
or not.
Acked-by: Pavel Shilovsky <piastryyy@gmail.com> Reviewed-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Jeff Layton [Tue, 21 Jun 2011 11:18:26 +0000 (07:18 -0400)]
cifs: mark CONFIG_CIFS_NFSD_EXPORT as BROKEN
This does not work properly with CIFS as current servers do not
enable support for the FILE_OPEN_BY_FILE_ID on SMB NTCreateX
and not all NFS clients handle ESTALE.
For now, it just plain doesn't work. Mark it BROKEN to discourage
distros from enabling it.
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com>
Jeff Layton [Mon, 20 Jun 2011 18:33:16 +0000 (14:33 -0400)]
cifs: free blkcipher in smbhash
This is currently leaked in the rc == 0 case.
Reported-by: J. Bruce Fields <bfields@fieldses.org> Signed-off-by: Jeff Layton <jlayton@redhat.com> Reviewed-by: Shirish Pargaonkar <shirishpargaonkar@gmail.com> Signed-off-by: Steve French <sfrench@us.ibm.com>
Linus Torvalds [Fri, 24 Jun 2011 15:42:35 +0000 (08:42 -0700)]
Merge branch 'for-linus' of git://git.kernel.dk/linux-block
* 'for-linus' of git://git.kernel.dk/linux-block:
block: add REQ_SECURE to REQ_COMMON_MASK
block: use the passed in @bdev when claiming if partno is zero
block: Add __attribute__((format(printf...) and fix fallout
block: make disk_block_events() properly wait for work cancellation
block: remove non-syncing __disk_block_events() and fold it into disk_block_events()
block: don't use non-syncing event blocking in disk_check_events()
cfq-iosched: fix locking around ioc->ioc_data assignment
Linus Torvalds [Fri, 24 Jun 2011 15:41:36 +0000 (08:41 -0700)]
Merge branch 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
pata_marvell: Add support for 88SE91A0, 88SE91A4
libata/sas: only set FROZEN flag if new EH is supported
libata: apply NOSETXFER horkage to the affected Pioneer drives regardless of firmware revision
drivers/ata/sata_dwc_460ex: Fix typo 'corrresponding'
Linus Torvalds [Fri, 24 Jun 2011 15:39:43 +0000 (08:39 -0700)]
Merge branch 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6
* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
drm/radeon/kms: handle special cases for vddc
drm/radeon/kms: fix num_banks tiling config for fusion
Linus Torvalds [Fri, 24 Jun 2011 15:36:16 +0000 (08:36 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jbarnes/pci-2.6:
x86/PCI/ACPI: fix type mismatch
PCI: fix new kernel-doc warning
PCI: Fix warning in drivers/pci/probe.c on sparc64
* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
cifs: fix wsize negotiation to respect max buffer size and active signing (try #4)
CIFS: Fix problem with 3.0-rc1 null user mount failure
Jesper Juhl [Thu, 23 Jun 2011 21:54:40 +0000 (23:54 +0200)]
ALSA: Remove unneeded version.h includes from sound/
In the sound/ directory there are two files (flagged by 'make
versioncheck'); sound/pci/asihpi/asihpi.c and
sound/soc/codecs/wm8991.c that include linux/version.h although they
don't need it. This patch removes the unneeded includes.
This patch fixes the following conversion specification warning for size_t
drivers/target/tcm_fc/tfc_io.c: In function ‘ft_queue_data_in’:
drivers/target/tcm_fc/tfc_io.c:209: warning: format ‘%x’ expects type ‘unsigned int’, but argument 5 has type ‘size_t’
Reported-by: Randy Dunlap <randy.dunlap@oracle.com> Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Kiran Patil [Wed, 22 Jun 2011 23:30:22 +0000 (16:30 -0700)]
tcm_fc: Fix ft_send_tm LUN lookup OOPs
This patch fixes a bug in ft_send_tm() that was incorrectly calling
ft_get_lun_for_cmd() -> transport_get_lun_for_cmd(), instead of using
transport_get_lun_for_tmr() for the proper struct se_lun lookup
that was triggering an OOPs in the se_cmd->tmr_req failure path.
This patch fixes the issue by re-arranging the codepath where
transport_get_lun_for_tmr() is called after tmr request is allocated and
made it available as part of se_cmd.
It also drops the now unnecessary ft_get_lun_for_cmd() unpacking code, and
uses scsilun_to_int() directly ahead of transport_get_lun_for_cmd() and
transport_get_lun_for_tmr() usage.
Signed-off-by: Patil, Kiran <kiran.patil@intel.com> Signed-off-by: Robert Love <robert.w.love@intel.com> Signed-off-by: Nicholas A. Bellinger <nab@linux-iscsi.org>
Dan Carpenter [Mon, 13 Jun 2011 20:10:49 +0000 (23:10 +0300)]
target: Drop bogus ERR_PTR usage in target_fabric_configfs_init
In the original code, there were several places inside the
target_fabric_configfs_init() function that returned NULL on error
and one place the returned an ERR_PTR. There are two places that
call this function and they only check for NULL returns; they don't
check for ERR_PTRs. So I've changed the ERR_PTR so now the function
only returns NULL on error.
Signed-off-by: Dan Carpenter <error27@gmail.com> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
Roland Dreier [Wed, 22 Jun 2011 08:02:21 +0000 (01:02 -0700)]
target: Convert transport_deregister_session_configfs nacl_sess_lock to save irq state
This patch converts transport_deregister_session_configfs() to save/restore
spinlock IRQ state for struct se_node_acl->nacl_sess_lock access as tcm_qla2xxx
logic expects to call transport_deregister_session_configfs() code with
irq save already held for struct qla_hw_data.
Reported-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org>
This patch fixes two possible NULL pointer dereferences in target v4.0
code where se_tmr release path in core_tmr_release_req() can OOPs upon
transport_get_lun_for_tmr() failure by attempting to access se_device or
se_tmr->tmr_list without a valid member of se_device->tmr_list during
transport_free_se_cmd() release. This patch moves the se_tmr->tmr_dev
pointer assignment in transport_get_lun_for_tmr() until after possible
-ENODEV failures during unpacked_lun lookup.
This addresses an OOPs originally reported with LIO v4.1 upstream on
.39 code here:
libata/sas: only set FROZEN flag if new EH is supported
On 16.06.2011 [08:28:39 -0500], Brian King wrote:
> On 06/16/2011 02:51 AM, Tejun Heo wrote:
> > On Wed, Jun 15, 2011 at 04:34:17PM -0700, Nishanth Aravamudan wrote:
> >>> That looks like the right thing to do. For ipr's usage of
> >>> libata, we don't have the concept of a port frozen state, so this flag
> >>> should really never get set. The alternate way to fix this would be to
> >>> only set ATA_PFLAG_FROZEN in ata_port_alloc if ap->ops->error_handler
> >>> is not NULL.
> >>
> >> It seemed like ipr is as you say, but I wasn't sure if it was
> >> appropriate to make the change above in the common libata-scis code or
> >> not. I don't want to break some other device on accident.
> >>
> >> Also, I tried your suggestion, but I don't think that can happen in
> >> ata_port_alloc? ata_port_alloc is allocated ap itself, and it seems like
> >> ap->ops typically gets set only after ata_port_alloc returns?
> >
> > Maybe we can test error_handler in ata_sas_port_start()?
>
> Good point. Since libsas is converted to the new eh now, we would need to have
> this test.
Commit 7b3a24c57d2eeda8dba9c205342b12689c4679f9 ("ahci: don't enable
port irq before handler is registered") caused a regression for CD-ROMs
attached to the IPR SATA bus on Power machines:
ata_port_alloc: ENTER
ata_port_probe: ata1: bus probe begin
ata1.00: ata_dev_read_id: ENTER
ata1.00: failed to IDENTIFY (I/O error, err_mask=0x40)
ata1.00: ata_dev_read_id: ENTER
ata1.00: failed to IDENTIFY (I/O error, err_mask=0x40)
ata1.00: limiting speed to UDMA7:PIO5
ata1.00: ata_dev_read_id: ENTER
ata1.00: failed to IDENTIFY (I/O error, err_mask=0x40)
ata1.00: disabled
ata_port_probe: ata1: bus probe end
scsi_alloc_sdev: Allocation failure during SCSI scanning, some SCSI devices might not be configured
The FROZEN flag added in that commit is only cleared by the new EH code,
which is not used by ipr. Clear this flag in the SAS code if we don't
support new EH.
Reported-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Signed-off-by: Jeff Garzik <jgarzik@pobox.com>
Tejun Heo [Wed, 22 Jun 2011 10:13:10 +0000 (12:13 +0200)]
libata: apply NOSETXFER horkage to the affected Pioneer drives
regardless of firmware revision
It's unlikely NOSETXFER works for a revision of drive but doesn't for
another and pioneer doesn't seem to be fixing firmwares for the
affected drives. Apply NOSETXFER to the affected pioneer drives
regardless of firmware revision.
http://article.gmane.org/gmane.linux.ide/49734
Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: fl-00@gmx.de Signed-off-by: Jeff Garzik <jgarzik@pobox.com>
Daniel Mack [Wed, 22 Jun 2011 18:48:25 +0000 (20:48 +0200)]
ASoC: pxa-ssp: Correct check for stream presence
Don't rely on the codec's channels_min information to decide wheter or
not allocate a substream's DMA buffer. Rather check if the substream
itself was allocated previously.
Signed-off-by: Daniel Mack <zonque@gmail.com> Acked-by: Liam Girdwood <lrg@ti.com> Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com> Cc: stable@kernel.org
CAP_LARGE_WRITEX is ignored when signing is active. Also, the maximum
size for a write without CAP_LARGE_WRITEX should be the maxBuf that
the server sent in the NEGOTIATE request.
Fix the wsize negotiation to take this into account. While we're at it,
alter the other wsize definitions to use sizeof(WRITE_REQ) to allow for
slightly larger amounts of data to potentially be written per request.
Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com>
Russell King [Wed, 22 Jun 2011 10:55:50 +0000 (11:55 +0100)]
Fix CPU spinlock lockups on secondary CPU bringup
Secondary CPU bringup typically calls calibrate_delay() during its
initialization. However, calibrate_delay() modifies a global variable
(loops_per_jiffy) used for udelay() and __delay().
A side effect of 71c696b1 ("calibrate: extract fall-back calculation
into own helper") introduced in the 2.6.39 merge window means that we
end up with a substantial period where loops_per_jiffy is zero. This
causes the spinlock debugging code to malfunction:
u64 loops = loops_per_jiffy * HZ;
for (;;) {
for (i = 0; i < loops; i++) {
if (arch_spin_trylock(&lock->raw_lock))
return;
__delay(1);
}
...
}
by never calling arch_spin_trylock() - resulting in the CPU locking
up in an infinite loop inside __spin_lock_debug().
Work around this by only writing to loops_per_jiffy only once we have
completed all the calibration decisions.
Tested-by: Santosh Shilimkar <santosh.shilimkar@ti.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Cc: <stable@kernel.org> (2.6.39-stable)
--
Better solutions (such as omitting the calibration for secondary CPUs,
or arranging for calibrate_delay() to return the LPJ value and leave
it to the caller to decide where to store it) are a possibility, but
would be much more invasive into each architecture.
I think this is the best solution for -rc and stable, but it should be
revisited for the next merge window.
The driver went to initialize its waitqueue at the start of the main
processing thread. However, it is possible that this thread is not
scheduled on a CPU before the write function is called which leads to a
following error:
William Douglas [Thu, 23 Jun 2011 12:38:36 +0000 (13:38 +0100)]
mrst_max3110: Change max missing message priority.
Change print message to notice instead of error to clean up non critical
messages showing on startup. The MAX3111 not being present is a normal
path for end user systems.
Signed-off-by: William Douglas <william.douglas@intel.com>
[rebased on 3.0, switched to dev_dbg()] Signed-off-by: Alan Cox <alan@linux.intel.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Thu, 23 Jun 2011 04:49:07 +0000 (21:49 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/shaggy/jfs-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/shaggy/jfs-2.6:
jfs: agstart field must be 64 bits
JFS: Don't save agno in the inode
jfs: Update agstart when resizing volume
jfs: old_agsize should be 64 bits in jfs_extendfs
David Rientjes [Thu, 23 Jun 2011 01:13:04 +0000 (18:13 -0700)]
mm, hotplug: protect zonelist building with zonelists_mutex
Commit 959ecc48fc75 ("mm/memory_hotplug.c: fix building of node hotplug
zonelist") does not protect the build_all_zonelists() call with
zonelists_mutex as needed. This can lead to races in constructing
zonelist ordering if a concurrent build is underway. Protecting this
with lock_memory_hotplug() is insufficient since zonelists can be
rebuild though sysfs as well.
David Rientjes [Thu, 23 Jun 2011 01:13:01 +0000 (18:13 -0700)]
mm, hotplug: fix error handling in mem_online_node()
The error handling in mem_online_node() is incorrect: hotadd_new_pgdat()
returns NULL if the new pgdat could not have been allocated and a pointer
to it otherwise.
mem_online_node() should fail if hotadd_new_pgdat() fails, not the
inverse. This fixes an issue when memoryless nodes are not onlined and
their sysfs interface is not registered when their first cpu is brought
up.
The bug was introduced by commit cf23422b9d76 ("cpu/mem hotplug: enable
CPUs online before local memory online") iow v2.6.35.
Pavel Shilovsky [Mon, 20 Jun 2011 08:33:16 +0000 (12:33 +0400)]
CIFS: Fix problem with 3.0-rc1 null user mount failure
Figured it out: it was broken by b946845a9dc523c759cae2b6a0f6827486c3221a commit - "cifs: cifs_parse_mount_options: do not tokenize mount options in-place". So, as a quick fix I suggest to apply this patch.
[PATCH] CIFS: Fix kfree() with constant string in a null user case
Signed-off-by: Pavel Shilovsky <piastryyy@gmail.com> Reviewed-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com>
Ben Widawsky [Wed, 22 Jun 2011 16:55:01 +0000 (09:55 -0700)]
drm/i915: save/resume forcewake lock fixes
The lock must be held for the saving and restoring of VGA state.
Signed-off-by: Ben Widawsky <ben@bwidawsk.net> CC: Alexander Zhaunerchyk <alex.vizor@gmail.com> CC: Andrey Rahmatullin <wrar@wrar.name> Reviewed-by: Chris Wilson <chris@chris-wilson.co.uk> Signed-off-by: Keith Packard <keithp@keithp.com>
David Howells [Tue, 21 Jun 2011 13:32:05 +0000 (14:32 +0100)]
KEYS: Fix error handling in construct_key_and_link()
Fix error handling in construct_key_and_link().
If construct_alloc_key() returns an error, it shouldn't pass out through
the normal path as the key_serial() called by the kleave() statement
will oops when it gets an error code in the pointer:
David Howells [Tue, 21 Jun 2011 09:29:51 +0000 (10:29 +0100)]
MN10300: asm/uaccess.h needs to #include linux/kernel.h for might_sleep()
MN10300's asm/uaccess.h needs to #include linux/kernel.h to get might_sleep()
otherwise it fails to build on MN10300 allyesconfig. This fails in a few
places with messages like the following:
In file included from security/keys/trusted.c:14:
include/linux/uaccess.h: In function '__copy_from_user_nocache':
include/linux/uaccess.h:52: error: implicit declaration of function 'might_sleep'
Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Wed, 22 Jun 2011 01:20:55 +0000 (18:20 -0700)]
Merge branch 'bugfixes' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6
* 'bugfixes' of git://git.linux-nfs.org/projects/trondmy/nfs-2.6:
NFS: Fix decode_secinfo_maxsz
NFSv4.1: Fix an off-by-one error in pnfs_generic_pg_test
NFSv4.1: Fix some issues with pnfs_generic_pg_test
NFSv4.1: file layout must consider pg_bsize for coalescing
pnfs-obj: No longer needed to take an extra ref at add_device
SUNRPC: Ensure the RPC client only quits on fatal signals
NFSv4: Fix a readdir regression
nfs4.1: mark layout as bad on error path in _pnfs_return_layout
nfs4.1: prevent race that allowed use of freed layout in _pnfs_return_layout
NFSv4.1: need to put_layout_hdr on _pnfs_return_layout error path
NFS: (d)printks should use %zd for ssize_t arguments
NFSv4.1: fix break condition in pnfs_find_lseg
nfs4.1: fix several problems with _pnfs_return_layout
NFSv4.1: allow zero fh array in filelayout decode layout
NFSv4.1: allow nfs_fhget to succeed with mounted on fileid
NFSv4.1: Fix a refcounting issue in the pNFS device id cache
NFSv4.1: deprecate headerpadsz in CREATE_SESSION
NFS41: do not update isize if inode needs layoutcommit
NLM: Don't hang forever on NLM unlock requests
NFS: fix umount of pnfs filesystems
John Stultz [Fri, 17 Jun 2011 01:47:37 +0000 (18:47 -0700)]
alarmtimers: Return -ENOTSUPP if no RTC device is present
Toralf Förster and Richard Weinberger noted that if there is
no RTC device, the alarm timers core prints out an annoying
"ALARM timers will not wake from suspend" message.
This warning has been removed in a previous patch, however
the issue still remains: The original idea was to support
alarm timers even if there was no rtc device, as long as the
system didn't go into suspend.
However, after further consideration, communicating to the application
that alarmtimers are not fully functional seems like the better
solution.
So this patch makes it so we return -ENOTSUPP to any posix _ALARM
clockid calls if there is no backing RTC device on the system.
Further this changes the behavior where when there is no rtc device
we will check for one on clock_getres, clock_gettime, timer_create,
and timer_nsleep instead of on suspend.
CC: Toralf Förster <toralf.foerster@gmx.de> CC: Richard Weinberger <richard@nod.at CC: Peter Zijlstra <peterz@infradead.org> CC: Thomas Gleixner <tglx@linutronix.de> Reported-by: Toralf Förster <toralf.foerster@gmx.de>
Reported by: Richard Weinberger <richard@nod.at> Signed-off-by: John Stultz <john.stultz@linaro.org>
John Stultz [Fri, 17 Jun 2011 01:27:09 +0000 (18:27 -0700)]
alarmtimers: Handle late rtc module loading
The alarmtimers code currently picks a rtc device to use at
late init time. However, if your rtc driver is loaded as a module,
it may be registered after the alarmtimers late init code, leaving
the alarmtimers nonfunctional.
This patch moves the the rtcdevice selection to when we actually try
to use it, allowing us to make use of rtc modules that may have been
loaded at any point since bootup.
CC: Thomas Gleixner <tglx@linutronix.de> CC: Meelis Roos <mroos@ut.ee> Reported-by: Meelis Roos <mroos@ut.ee> Signed-off-by: John Stultz <john.stultz@linaro.org>
PCI / PM: Block races between runtime PM and system sleep
After commit e8665002477f0278f84f898145b1f141ba26ee26
(PM: Allow pm_runtime_suspend() to succeed during system suspend) it
is possible that a device resumed by the pm_runtime_resume(dev) in
pci_pm_prepare() will be suspended immediately from a work item,
timer function or otherwise, defeating the very purpose of calling
pm_runtime_resume(dev) from there. To prevent that from happening
it is necessary to increment the runtime PM usage counter of the
device by replacing pm_runtime_resume() with pm_runtime_get_sync().
Moreover, the incremented runtime PM usage counter has to be
decremented by the corresponding pci_pm_complete(), via
pm_runtime_put_sync().
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Cc: stable@kernel.org Acked-by: Jesse Barnes <jbarnes@virtuousgeek.org>
Commit 4d27e9dcff00a6425d779b065ec8892e4f391661 (PM: Make power
domain callbacks take precedence over subsystem ones) forgot to
update the device power management documentation to take changes
made by it into account. Correct that mistake.
PM / Runtime: Handle clocks correctly if CONFIG_PM_RUNTIME is unset
Commit 85eb8c8d0b0900c073b0e6f89979ac9c439ade1a (PM / Runtime:
Generic clock manipulation rountines for runtime PM (v6)) converted
the shmobile platform to using generic code for runtime PM clock
management, but it changed the behavior for CONFIG_PM_RUNTIME unset
incorrectly.
Specifically, for CONFIG_PM_RUNTIME unset pm_runtime_clk_notify()
should enable clocks for action equal to BUS_NOTIFY_BIND_DRIVER and
it should disable them for action equal to BUS_NOTIFY_UNBOUND_DRIVER
(instead of BUS_NOTIFY_ADD_DEVICE and BUS_NOTIFY_DEL_DEVICE,
respectively). Make this function behave as appropriate.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Acked-by: Magnus Damm <damm@opensource.se>
Alan Stern [Sat, 18 Jun 2011 20:42:09 +0000 (22:42 +0200)]
PM: Fix async resume following suspend failure
The PM core doesn't handle suspend failures correctly when it comes to
asynchronously suspended devices. These devices are moved onto the
dpm_suspended_list as soon as the corresponding async thread is
started up, and they remain on the list even if they fail to suspend
or the sleep transition is cancelled before they get suspended. As a
result, when the PM core unwinds the transition, it tries to resume
the devices even though they were never suspended.
This patch (as1474) fixes the problem by adding a new "is_suspended"
flag to dev_pm_info. Devices are resumed only if the flag is set.
[rjw:
* Moved the dev->power.is_suspended check into device_resume(),
because we need to complete dev->power.completion and clear
dev->power.is_prepared too for devices whose
dev->power.is_suspended flags are unset.
* Fixed __device_suspend() to avoid setting dev->power.is_suspended
if async_error is different from zero.]
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Cc: stable@kernel.org
Michal Kubecek [Sat, 18 Jun 2011 18:34:01 +0000 (20:34 +0200)]
PM: Free memory bitmaps if opening /dev/snapshot fails
When opening /dev/snapshot device, snapshot_open() creates memory
bitmaps which are freed in snapshot_release(). But if any of the
callbacks called by pm_notifier_call_chain() returns NOTIFY_BAD, open()
fails, snapshot_release() is never called and bitmaps are not freed.
Next attempt to open /dev/snapshot then triggers BUG_ON() check in
create_basic_memory_bitmaps(). This happens e.g. when vmwatchdog module
is active on s390x.
Signed-off-by: Michal Kubecek <mkubecek@suse.cz> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Cc: stable@kernel.org
Alan Stern [Sat, 18 Jun 2011 18:22:23 +0000 (20:22 +0200)]
PM: Rename dev_pm_info.in_suspend to is_prepared
This patch (as1473) renames the "in_suspend" field in struct
dev_pm_info to "is_prepared", in preparation for an upcoming change.
The new name is more descriptive of what the field really means.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Cc: stable@kernel.org
projects / karo-tx-linux.git / log