Commit ae1b153962, block: reimplement
FLUSH/FUA to support merge, introduced a performance regression when
running any sort of fsyncing workload using dm-multipath and certain
storage (in our case, an HP EVA). The test I ran was fs_mark, and it
dropped from ~800 files/sec on ext4 to ~100 files/sec. It turns out
that dm-multipath always advertised flush+fua support, and passed
commands on down the stack, where those flags used to get stripped off.
The above commit changed that behavior:
static inline struct request *__elv_next_request(struct request_queue *q)
{
struct request *rq;
while (1) {
- while (!list_empty(&q->queue_head)) {
+ if (!list_empty(&q->queue_head)) {
rq = list_entry_rq(q->queue_head.next);
- if (!(rq->cmd_flags & (REQ_FLUSH | REQ_FUA)) ||
- (rq->cmd_flags & REQ_FLUSH_SEQ))
- return rq;
- rq = blk_do_flush(q, rq);
- if (rq)
- return rq;
+ return rq;
}
Note that previously, a command would come in here, have
REQ_FLUSH|REQ_FUA set, and then get handed off to blk_do_flush:
struct request *blk_do_flush(struct request_queue *q, struct request *rq)
{
unsigned int fflags = q->flush_flags; /* may change, cache it */
bool has_flush = fflags & REQ_FLUSH, has_fua = fflags & REQ_FUA;
bool do_preflush = has_flush && (rq->cmd_flags & REQ_FLUSH);
bool do_postflush = has_flush && !has_fua && (rq->cmd_flags &
REQ_FUA);
unsigned skip = 0;
...
if (blk_rq_sectors(rq) && !do_preflush && !do_postflush) {
rq->cmd_flags &= ~REQ_FLUSH;
if (!has_fua)
rq->cmd_flags &= ~REQ_FUA;
return rq;
}
So, the flush machinery was bypassed in such cases (q->flush_flags == 0
&& rq->cmd_flags & (REQ_FLUSH|REQ_FUA)).
Now, however, we don't get into the flush machinery at all. Instead,
__elv_next_request just hands a request with flush and fua bits set to
the scsi_request_fn, even if the underlying request_queue does not
support flush or fua.
The agreed upon approach is to fix the flush machinery to allow
stacking. While this isn't used in practice (since there is only one
request-based dm target, and that target will now reflect the flush
flags of the underlying device), it does future-proof the solution, and
make it function as designed.
In order to make this work, I had to add a field to the struct request,
inside the flush structure (to store the original req->end_io). Shaohua
had suggested overloading the union with rb_node and completion_data,
but the completion data is used by device mapper and can also be used by
other drivers. So, I didn't see a way around the additional field.
I tested this patch on an HP EVA with both ext4 and xfs, and it recovers
the lost performance. Comments and other testers, as always, are
appreciated.
Cheers,
Jeff
Signed-off-by: Jeff Moyer <jmoyer@redhat.com>
Acked-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jens Axboe <jaxboe@fusionio.com>
The clock gating functions are only assigned under KMS, so don't try
to call them under UMS.
Signed-off-by: Keith Packard <keithp@keithp.com>
Tested-by: Justin P. Mattock <justinmattock@gmail.com>
Disable this feature when KMS is not running by setting the
driver->get_vblank_timestamp function pointer to NULL.
Signed-off-by: Keith Packard <keithp@keithp.com>
Tested-by: Justin P. Mattock <justinmattock@gmail.com>
Signed-off-by: Matthew Garrett <mjg@redhat.com>
Cc: Richard Purdie <rpurdie@rpsys.net>
Cc: Chris Wilson <chris@chris-wilson.co.uk>
Cc: David Airlie <airlied@linux.ie>
Cc: Alex Deucher <alexdeucher@gmail.com>
Cc: Ben Skeggs <bskeggs@redhat.com>
Cc: Zhang Rui <rui.zhang@intel.com>
Cc: Len Brown <lenb@kernel.org>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Tested-by: Sedat Dilek <sedat.dilek@googlemail.com>
Tested-by: Michel Alexandre Salim <salimma@fedoraproject.org>
Tested-by: Kamal Mostafa <kamal@canonical.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Keith Packard <keithp@keithp.com>
* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus:
lguest: allow booting guest with CONFIG_RELOCATABLE=y
virtio: Add text copy of spec to Documentation/virtual.
* 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
firewire: ohci: fix DMA unmapping in an error path
firewire: cdev: fix 32 bit userland on 64 bit kernel compat corner cases
* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
drm/radeon/kms: don't try to be smart in the hpd handler
drm/radeon: re-POST the asic on Apple hardware when booted via EFI
drm/radeon: Allow panel preferred EDID to override BIOS native mode
drm/radeon/kms: make some watermark messages debug only
drm/radeon/kms: fix regression is handling >2 heads on cedar/caicos
drm/radeon/kms: don't enable connectors that are off in the hotplug handler
* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
cifs: Do not set cifs/ntfs acl using a file handle (try #4)
[CIFS] Cleanup use of CONFIG_CIFS_STATS2 ifdef to make transport routines more readable
Attempting to try and turn off disconnected display hw in the
hotput handler lead to more problems than it helped. For
now just register an event and only attempt the do something
interesting with DP. Other connectors are just too problematic:
- Some systems have an HPD pin assigned to LVDS, but it's rarely
if ever connected properly and we don't really care about hpd
events on LVDS anyway since it's always connected.
- The HPD pin is wired up correctly for eDP, but we don't really
have to do anything since the events since it's always connected.
- Some HPD pins fire more than once when you connect/disconnect
- etc.
Fixes:
https://bugs.freedesktop.org/show_bug.cgi?id=39882
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
BugLink: https://bugs.launchpad.net/bugs/826081
The original reporter needs 'Headphone Jack Sense' enabled to have
audible audio, so add his PCI SSID to the whitelist.
Reported-and-tested-by: Muhammad Khurram Khan
Cc: <stable@kernel.org>
Signed-off-by: Daniel T Chen <crimsun@ubuntu.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
The CONFIG_RELOCATABLE code tries to align the unpack destination to
the value of 'kernel_alignment' in the setup_hdr. If that's 0, it
tries to unpack to address 0, which in fact causes the gunzip code
to call 'error("Out of memory while allocating output buffer")'.
The bootloader (ie. the lguest Launcher in this case) should be doing
setting this field; the normal bzImage is 16M, we can use the same.
Reported-by: Stefanos Geraggelos <sgerag@cslab.ece.ntua.gr>
Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Cc: stable@kernel.org
Commit db64fe0225 ("mm: rewrite vmap layer") introduced code that does
address calculations under the assumption that VMAP_BLOCK_SIZE is a
power of two. However, this might not be true if CONFIG_NR_CPUS is not
set to a power of two.
Wrong vmap_block index/offset values could lead to memory corruption.
However, this has never been observed in practice (or never been
diagnosed correctly); what caught this was the BUG_ON in vb_alloc() that
checks for inconsistent vmap_block indices.
To fix this, ensure that VMAP_BLOCK_SIZE always is a power of two.
BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=31572
Reported-by: Pavel Kysilka <goldenfish@linuxsoft.cz>
Reported-by: Matias A. Fonzo <selk@dragora.org>
Signed-off-by: Clemens Ladisch <clemens@ladisch.de>
Signed-off-by: Stefan Richter <stefanr@s5r6.in-berlin.de>
Cc: Nick Piggin <npiggin@suse.de>
Cc: Jeremy Fitzhardinge <jeremy@goop.org>
Cc: Krzysztof Helt <krzysztof.h1@poczta.fm>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: 2.6.28+ <stable@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The snd_usb_caiaq driver currently assumes that output urbs are serviced
in time and doesn't track when and whether they are given back by the
USB core. That usually works fine, but due to temporary limitations of
the XHCI stack, we faced that urbs were submitted more than once with
this approach.
As it's no good practice to fire and forget urbs anyway, this patch
introduces a proper bit mask to track which requests have been submitted
and given back.
That alone however doesn't make the driver work in case the host
controller is broken and doesn't give back urbs at all, and the output
stream will stop once all pre-allocated output urbs are consumed. But
it does prevent crashes of the controller stack in such cases.
See http://bugzilla.kernel.org/show_bug.cgi?id=40702 for more details.
Signed-off-by: Daniel Mack <zonque@gmail.com>
Reported-and-tested-by: Matej Laitl <matej@laitl.cz>
Cc: Sarah Sharp <sarah.a.sharp@linux.intel.com>
Cc: stable@kernel.org
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Function genpd_queue_power_off_work() is not defined for
CONFIG_PM_RUNTIME, so pm_genpd_poweroff_unused() causes a build
error to happen in that case. Fix the problem by making
pm_genpd_poweroff_unused() depend on CONFIG_PM_RUNTIME too.
Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl>
Remove the name of Sergey Kostyliov as maintainer of befs.
In the MAINTAINERS file, befs is orphaned.
Signed-off-by: Marcos Souza <marcos.mage@gmail.com>
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
General cleanups to kernel-parameters.txt:
- add missing $ARCH that are being used/referenced
- alphabetize the parameter restrictions list
- spell "IA-64" as listed in arch/ia64/Kconfig instead of "IA64"
- remove trailing whitespace
- use hyphen in 32-bit etc.
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Usually kernel parameters are documented in kernel-parameters.txt
but user_debug is only documented in the Kconfig. Document the
option and point to the Kconfig help text for more info.
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Cc: Russell King <linux@arm.linux.org.uk>
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Dropping LSD (Linux Source Driver) since it hasn't been available
for a long time.
Signed-off-by: Luis de Bethencourt <luis@debethencourt.com>
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Host driver can't get a hint of DDR mode through ios->ddr flag anymore.
ios->timing is currently used to inform DDR mode as a substitute.
And capability of MMC_CAP_MMC_HIGHSPEED is added for DDR support.
Signed-off-by: Seungwon Jeon <tgih.jun@samsung.com>
Acked-by: Will Newton <will.newton@imgtec.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
When timeout_clk is calculated the host->clock could be zero.
So, instead of host->clock the calculation now uses mmc->f_max.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
This moves the calculation below the assignment of mmc->f_max, which
we need for calculating timeout_clk in the next patch in this series.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
This reverts commit 4b01681c77, which introduced a new potential
divide by zero in the process of fixing one. The subsequent commits
attempt to fix the issue properly.
Signed-off-by: Chris Ball <cjb@laptop.org>
Fix below compile warning:
CC drivers/mmc/host/tmio_mmc.o
drivers/mmc/host/tmio_mmc.c: In function 'tmio_mmc_suspend':
drivers/mmc/host/tmio_mmc.c:30: warning: unused variable 'mmc'
drivers/mmc/host/tmio_mmc.c: In function 'tmio_mmc_resume':
drivers/mmc/host/tmio_mmc.c:45: warning: unused variable 'mmc'
Signed-off-by: Axel Lin <axel.lin@gmail.com>
Acked-by: Guennadi Liakhovetski <g.liakhovetski@gmx.de>
Signed-off-by: Chris Ball <cjb@laptop.org>
Apply a workaround for the imx eSDHC controller to avoid missing
card interrupts. This makes SDIO work.
Signed-off-by: Tony Lin <tony.lin@freescale.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
A header change has removed an implicit inclusion of module.h, breaking
the build due to the use of THIS_MODULE. Fix that.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
The mask used inside this macro was assuming Buffer_Size1's [BS1's]
width to be 14 bits, it is actually 13 bits. Modify masks used in
IDMAC_SET_BUFFER1_SIZE such that they use only 13 bits instead of
current 14.
Signed-off-by: Shashidhar Hiremath <shashidharh@vayavyalabs.com>
Acked-by: Will Newton <will.newton@imgtec.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
During card removal and inserting cycle the test file in the debugfs could be
stalled until the host driver removes it. Let's keep the file in the linked
list and destroy it when card is removed.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Per Forlin <per.forlin@linaro.org>
Signed-off-by: Chris Ball <cjb@laptop.org>
Samsung SoCs need to set BROKEN_ADMA_ZEROLEN_DESC.
(If ADMA operation is more than 65535, maybe set by zero.)
Signed-off-by: Jaehoon Chung <jh80.chung@samsung.com>
Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
Enable the quirk.
(Best used in conjunction with patch downgrading ADMA to SDMA when
transfer is not aligned.)
Signed-off-by: Philip Rakity <prakity@marvell.com>
Acked-by: Zhangfei Gao <zhangfei.gao@marvell.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
Currently, the retuning timer for retuning mode 1 will be deleted in
function sdhci_tasklet_finish after a mmc request done, which will make
retuning timing never trigger again. This patch fixed this problem.
Signed-off-by: Aaron Lu <Aaron.Lu@amd.com>
Reviewed-by: Philip Rakity <prakity@marvell.com>
Signed-off-by: Chris Ball <cjb@laptop.org>
Bug discovered by Jan Kara:
Finally, commit 1449032be1 returned back
the old IO submission code but apparently it forgot to return the old
handling of uninitialized buffers so we unconditionnaly call
block_write_full_page() without specifying end_io function. So AFAICS
we never convert unwritten extents to written in some cases. For
example when I mount the fs as: mount -t ext4 -o
nomblk_io_submit,dioread_nolock /dev/ubdb /mnt and do
int fd = open(argv[1], O_RDWR | O_CREAT | O_TRUNC, 0600);
char buf[1024];
memset(buf, 'a', sizeof(buf));
fallocate(fd, 0, 0, 16384);
write(fd, buf, sizeof(buf));
I get a file full of zeros (after remounting the filesystem so that
pagecache is dropped) instead of seeing the first KB contain 'a's.
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
EXT4_IO_END_UNWRITTEN flag set and the increase of i_aiodio_unwritten
should be done simultaneously since ext4_end_io_nolock always clear
the flag and decrease the counter in the same time.
We don't increase i_aiodio_unwritten when setting
EXT4_IO_END_UNWRITTEN so it will go nagative and causes some process
to wait forever.
Part of the patch came from Eric in his e-mail, but it doesn't fix the
problem met by Michael actually.
http://marc.info/?l=linux-ext4&m=131316851417460&w=2
Reported-and-Tested-by: Michael Tokarev<mjt@tls.msk.ru>
Signed-off-by: Eric Sandeen <sandeen@redhat.com>
Signed-off-by: Tao Ma <boyu.mt@taobao.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
Flush inode's i_completed_io_list before calling ext4_io_wait to
prevent the following deadlock scenario: A page fault happens while
some process is writing inode A. During page fault,
shrink_icache_memory is called that in turn evicts another inode
B. Inode B has some pending io_end work so it calls ext4_ioend_wait()
that waits for inode B's i_ioend_count to become zero. However, inode
B's ioend work was queued behind some of inode A's ioend work on the
same cpu's ext4-dio-unwritten workqueue. As the ext4-dio-unwritten
thread on that cpu is processing inode A's ioend work, it tries to
grab inode A's i_mutex lock. Since the i_mutex lock of inode A is
still hold before the page fault happened, we enter a deadlock.
Also moves ext4_flush_completed_IO and ext4_ioend_wait from
ext4_destroy_inode() to ext4_evict_inode(). During inode deleteion,
ext4_evict_inode() is called before ext4_destroy_inode() and in
ext4_evict_inode(), we may call ext4_truncate() without holding
i_mutex lock. As a result, there is a race between flush_completed_IO
that is called from ext4_ext_truncate() and ext4_end_io_work, which
may cause corruption on an io_end structure. This change moves
ext4_flush_completed_IO and ext4_ioend_wait from ext4_destroy_inode()
to ext4_evict_inode() to resolve the race between ext4_truncate() and
ext4_end_io_work during inode deletion.
Signed-off-by: Jiaying Zhang <jiayingz@google.com>
Signed-off-by: "Theodore Ts'o" <tytso@mit.edu>
Cc: stable@kernel.org
