Mike Frysinger [Wed, 6 Jan 2010 17:23:23 +0000 (17:23 +0000)]
NOMMU: Avoiding duplicate icache flushes of shared maps
When working with FDPIC, there are many shared mappings of read-only
code regions between applications (the C library, applet packages like
busybox, etc.), but the current do_mmap_pgoff() function will issue an
icache flush whenever a VMA is added to an MM instead of only doing it
when the map is initially created.
The flush can instead be done when a region is first mmapped PROT_EXEC.
Note that we may not rely on the first mapping of a region being
executable - it's possible for it to be PROT_READ only, so we have to
remember whether we've flushed the region or not, and then flush the
entire region when a bit of it is made executable.
However, this also affects the brk area. That will no longer be
executable. We can mprotect() it to PROT_EXEC on MPU-mode kernels, but
for NOMMU mode kernels, when it increases the brk allocation, making
sys_brk() flush the extra from the icache should suffice. The brk area
probably isn't used by NOMMU programs since the brk area can only use up
the leavings from the stack allocation, where the stack allocation is
larger than requested.
Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Mike Frysinger [Wed, 6 Jan 2010 17:23:17 +0000 (17:23 +0000)]
FDPIC: Respect PT_GNU_STACK exec protection markings when creating NOMMU stack
The current code will load the stack size and protection markings, but
then only use the markings in the MMU code path. The NOMMU code path
always passes PROT_EXEC to the mmap() call. While this doesn't matter
to most people whilst the code is running, it will cause a pointless
icache flush when starting every FDPIC application. Typically this
icache flush will be of a region on the order of 128KB in size, or may
be the entire icache, depending on the facilities available on the CPU.
In the case where the arch default behaviour seems to be desired
(EXSTACK_DEFAULT), we probe VM_STACK_FLAGS for VM_EXEC to determine
whether we should be setting PROT_EXEC or not.
For arches that support an MPU (Memory Protection Unit - an MMU without
the virtual mapping capability), setting PROT_EXEC or not will make an
important difference.
It should be noted that this change also affects the executability of
the brk region, since ELF-FDPIC has that share with the stack. However,
this is probably irrelevant as NOMMU programs aren't likely to use the
brk region, preferring instead allocation via mmap().
Signed-off-by: Mike Frysinger <vapier@gentoo.org> Signed-off-by: David Howells <dhowells@redhat.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Thu, 7 Jan 2010 02:10:15 +0000 (18:10 -0800)]
Merge branch 'for-2.6.33' of git://linux-nfs.org/~bfields/linux
* 'for-2.6.33' of git://linux-nfs.org/~bfields/linux:
sunrpc: fix peername failed on closed listener
nfsd: make sure data is on disk before calling ->fsync
nfsd: fix "insecure" export option
Xiaotian Feng [Thu, 31 Dec 2009 02:52:36 +0000 (10:52 +0800)]
sunrpc: fix peername failed on closed listener
There're some warnings of "nfsd: peername failed (err 107)!"
socket error -107 means Transport endpoint is not connected.
This warning message was outputed by svc_tcp_accept() [net/sunrpc/svcsock.c],
when kernel_getpeername returns -107. This means socket might be CLOSED.
And svc_tcp_accept was called by svc_recv() [net/sunrpc/svc_xprt.c]
if (test_bit(XPT_LISTENER, &xprt->xpt_flags)) {
<snip>
newxpt = xprt->xpt_ops->xpo_accept(xprt);
<snip>
So this might happen when xprt->xpt_flags has both XPT_LISTENER and XPT_CLOSE.
Let's take a look at commit b0401d72, this commit has moved the close
processing after do recvfrom method, but this commit also introduces this
warnings, if the xpt_flags has both XPT_LISTENER and XPT_CLOSED, we should
close it, not accpet then close.
Signed-off-by: Xiaotian Feng <dfeng@redhat.com> Cc: J. Bruce Fields <bfields@fieldses.org> Cc: Neil Brown <neilb@suse.de> Cc: Trond Myklebust <Trond.Myklebust@netapp.com> Cc: David S. Miller <davem@davemloft.net> Cc: stable@kernel.org Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
nfsd: make sure data is on disk before calling ->fsync
nfsd is not using vfs_fsync, so I missed it when changing the calling
convention during the 2.6.32 window. This patch fixes it to not only
start the data writeout, but also wait for it to complete before calling
into ->fsync.
Signed-off-by: Christoph Hellwig <hch@lst.de> Cc: stable@kernel.org Signed-off-by: J. Bruce Fields <bfields@citi.umich.edu>
Linus Torvalds [Wed, 6 Jan 2010 18:46:27 +0000 (10:46 -0800)]
Merge branch 'davinci-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-davinci
* 'davinci-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/khilman/linux-davinci:
DaVinci: DM365: Add the device_enable for the DaVinci Keyscan
davinci: enable ARCH_HAS_HOLES_MEMORYMODEL for DaVinci
davinci: da8xx/omap-l1: mark RTC as a wakeup source
davinci: cp_intc: provide set_wake function
Davinci VPFE Capture: Take i2c adapter id through platform data
* git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-rc-fixes-2.6:
[SCSI] lpfc 8.3.7: Update Driver version to 8.3.7
[SCSI] lpfc 8.3.7: Fix discovery failures.
[SCSI] lpfc 8.3.7: Fix SCSI protocol related errors.
[SCSI] lpfc 8.3.7: Fix hardware/SLI relates issues
[SCSI] lpfc 8.3.7: Fix NPIV operation errors
[SCSI] lpfc 8.3.7: Fix FC protocol errors
[SCSI] stex: fix scan of nonexistent lun
[SCSI] pmcraid: fix to avoid twice scsi_dma_unmap for a command
[SCSI] qla2xxx: Update version number to 8.03.01-k9.
[SCSI] qla2xxx: Added to EEH support.
[SCSI] qla2xxx: Extend base EEH support in qla2xxx.
[SCSI] qla2xxx: Fix for a multiqueue bug in CPU affinity mode
[SCSI] qla2xxx: Get the link data rate explicitly during device resync.
[SCSI] cxgb3i: Fix a login over vlan issue
Sekhar Nori [Thu, 19 Nov 2009 10:04:01 +0000 (15:34 +0530)]
davinci: enable ARCH_HAS_HOLES_MEMORYMODEL for DaVinci
All DaVinci platforms include a DSP or co-processor for
audio/video acceleration.
While creating memory for the DSP/co-processor, system
integrator can end up creating a hole in the memory map
of the sort:
<kernel memory> <hole (memory for DSP)> <kernel memory>
This sort of configuration needs ARCH_HAS_HOLES_MEMORYMODEL
enabled. See further details see this discussion on ARM
linux mailing list:
http://www.mail-archive.com/linux-omap@vger.kernel.org/msg15262.html
The patch is boot tested on OMAP-L138, DM6446 and DM355 EVMs
Signed-off-by: Sekhar Nori <nsekhar@ti.com> CC: Sriramakrishnan <srk@ti.com> CC: Khasim Syed Mohammed <khasim@ti.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Sekhar Nori [Mon, 16 Nov 2009 11:51:30 +0000 (17:21 +0530)]
davinci: cp_intc: provide set_wake function
There is nothing special to be done for interrupts
which can wakeup the device from sleep on CP-INTC,
but not having a set_wake implemented prevents use
of common drivers which expect this function to be
implemented for all wakeup interrupt sources.
This patch fixes the issue encountered when using the
omap-rtc driver on DA850. On DA850 the RTC alarm
interrupt is used to wake up the SoC from deep sleep
mode. Without this patch, the disable_irq_wake throws
an unbalanced wake disable warning while resuming
because the previous enable call fails for lack of
set_wake implementation.
Signed-off-by: Sekhar Nori <nsekhar@ti.com> Signed-off-by: Kevin Hilman <khilman@deeprootsystems.com>
Linus Torvalds [Wed, 6 Jan 2010 09:41:07 +0000 (01:41 -0800)]
Merge branch 'for-linus' of git://git.open-osd.org/linux-open-osd
* 'for-linus' of git://git.open-osd.org/linux-open-osd:
exofs: simple_write_end does not mark_inode_dirty
exofs: fix pnfs_osd re-definitions in pre-pnfs trees
Linus Torvalds [Wed, 6 Jan 2010 09:16:26 +0000 (01:16 -0800)]
Merge master.kernel.org:/home/rmk/linux-2.6-arm
* master.kernel.org:/home/rmk/linux-2.6-arm: (22 commits)
ARM: 5865/1: nuc900 ethernet driver needs mii
ARM: 5864/1: Implement arch_reset() in NUC900
ARM: 5863/1: fix bugs of clock source of NUC900
ARM: 5858/1: Remove unused vma_vm_flags macro from v7wbi_flush_user_tlb_range
imx/mx3: depend on USB_ULPI for otg_ulpi_create
ARM: MX3: make CPU revision number detection work on all boards
mx25: pdk: add platform code for FEC support
mx25: add support for FEC on i.MX25
mx25: s/NO_PAD_CTL/NO_PAD_CTRL/
mx31moboard: fix usbh device names
mx3: add support for the mt9v022 camera sensor to pcm037 platform
mx27: mxt_td60: Remove not used UART pins
[ARM] pxa/poodle: fix incorrect 'gpio_card_detect' of MMC
[ARM] pxa/zylonite: simplify reduntant gpio settings on mmc slot
[ARM] pxa/ttc_dkb: remove duplicate macro definition
[ARM] pxa/zeus: provide power-source information when APM is enabled
[ARM] pxa/zeus: relax memory timings on Zeus ethernet ports
[ARM] pxa/zeus: make internal zeus_get_pcb_info static
[ARM] pxa/littleton: select CPU_PXA300 and CPU_PXA310
[ARM] pxa/littleton: add UART3 GPIO config
...
Ben Hutchings [Sat, 19 Dec 2009 14:43:01 +0000 (14:43 +0000)]
modules: Skip empty sections when exporting section notes
Commit 35dead4 "modules: don't export section names of empty sections
via sysfs" changed the set of sections that have attributes, but did
not change the iteration over these attributes in add_notes_attrs().
This can lead to add_notes_attrs() creating attributes with the wrong
names or with null name pointers.
Introduce a sect_empty() function and use it in both add_sect_attrs()
and add_notes_attrs().
Reported-by: Martin Michlmayr <tbm@cyrius.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Tested-by: Martin Michlmayr <tbm@cyrius.com> Cc: stable@kernel.org Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc-2.6:
sparc64: Fix Niagara2 perf event handling.
sparc64: Fix NMI programming when perf events are active.
bbc_envctrl: Clean up properly if kthread_run() fails.
As Linus said in 982d007a6ee: "There was something really messy about
cmpxchg8b and clone CPU's, so if you enable it on other CPUs later, do it
carefully."
This breaks lguest for those configs, but we can fix that by emulating
if we have to.
Linus Torvalds [Tue, 5 Jan 2010 23:59:56 +0000 (15:59 -0800)]
Merge branch 'for-2.6.33' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound-2.6
* 'for-2.6.33' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/sound-2.6:
ASoC: fixup oops in generic AC97 codec glue
ASoC: fix params_rate() macro use in several codecs
ASoC: fsi-ak4642: Remove ak4642_add_i2c_device
Li Jie [Thu, 31 Dec 2009 14:57:53 +0000 (15:57 +0100)]
ARM: 5863/1: fix bugs of clock source of NUC900
This patch fix following bugs:
1. typo error, CLOCK_EVT_MODE_PERIODIC -> CLOCK_EVT_FEAT_PERIODIC
2. TCSR register of timer1 missed PRESCALE
3. timer1 should be enabled before register it to clock source.
Signed-off-by: lijie <eltshanli@gmail.com> Acked-by: Wan ZongShun <mcuos.com@gmail.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
David S. Miller [Tue, 5 Jan 2010 07:16:03 +0000 (23:16 -0800)]
sparc64: Fix Niagara2 perf event handling.
For chips like Niagara2 that have true overflow indications
in the %pcr (which we don't actually need and don't use)
the interrupt signal persists until the overflow bits are
cleared by an explicit %pcr write.
Signed-off-by: David S. Miller <davem@davemloft.net>
Boaz Harrosh [Sun, 27 Dec 2009 15:01:42 +0000 (17:01 +0200)]
exofs: simple_write_end does not mark_inode_dirty
exofs uses simple_write_end() for it's .write_end handler. But
it is not enough because simple_write_end() does not call
mark_inode_dirty() when it extends i_size. So even if we do
call mark_inode_dirty at beginning of write out, with a very
long IO and a saturated system we might get the .write_inode()
called while still extend-writing to file and miss out on the last
i_size updates.
So override .write_end, call simple_write_end(), and afterwords if
i_size was changed call mark_inode_dirty().
It stands to logic that since simple_write_end() was the one extending
i_size it should also call mark_inode_dirty(). But it looks like all
users of simple_write_end() are memory-bound pseudo filesystems, who
could careless about mark_inode_dirty(). I might submit a
warning-comment patch to simple_write_end() in future.
Boaz Harrosh [Mon, 21 Dec 2009 14:36:23 +0000 (16:36 +0200)]
exofs: fix pnfs_osd re-definitions in pre-pnfs trees
Some on disk exofs constants and types are defined in the pnfs_osd_xdr.h
file. Since we needed these types before the pnfs-objects code was
accepted to mainline we duplicated the minimal needed definitions into
an exofs local header. The definitions where conditionally included
depending on !CONFIG_PNFS defined. So if PNFS was present in the tree
definitions are taken from there and if not they are defined locally.
That was all good but, the CONFIG_PNFS is planed to be included upstream
before the pnfs-objects is also included. (The first pnfs batch might be
pnfs-files only)
So condition exofs local definitions on the absence of pnfs_osd_xdr.h
inclusion (__PNFS_OSD_XDR_H__ not defined). User code must make sure
that in future pnfs_osd_xdr.h will be included before fs/exofs/pnfs.h,
which happens to be so in current code.
Once pnfs-objects hits mainline, exofs's local header will be removed.
David S. Miller [Mon, 4 Jan 2010 23:31:10 +0000 (15:31 -0800)]
bbc_envctrl: Clean up properly if kthread_run() fails.
In bbc_envctrl_init() we have to unlink the fan and temp instances
from the lists because our caller is going to free up the 'bp' object
if we return an error.
We can't rely upon bbc_envctrl_cleanup() to do this work for us in
this case.
Reported-by: Patrick Finnegan <pat@computer-refuge.org> Signed-off-by: David S. Miller <davem@davemloft.net>
sysfs: Add lockdep annotations for the sysfs active reference
Holding locks over device_del -> kobject_del -> sysfs_deactivate can
cause deadlocks if those same locks are grabbed in sysfs show or store
methods.
The I model s_active count + completion as a sleeping read/write lock.
I describe to lockdep sysfs_get_active as a read_trylock,
sysfs_put_active as a read_unlock, and sysfs_deactivate as a
write_lock and write_unlock pair. This seems to capture the essence
for purposes of finding deadlocks, and in my testing gives finds real
issues and ignores non-issues.
This brings us back to holding locks over kobject_del is a problem
that ideally we should find a way of addressing, but at least lockdep
can tell us about the problems instead of requiring developers to debug
rare strange system deadlocks, that happen when sysfs files are removed
while being written to.
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com> Acked-by: Tejun Heo <tj@kernel.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Mon, 4 Jan 2010 20:32:42 +0000 (12:32 -0800)]
Merge branch 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6
* 'for-linus' of git://git390.marist.edu/pub/scm/linux-2.6:
[S390] Update default configuration.
[S390] Have param.h simply include <asm-generic/param.h>.
[S390] qdio: convert global statistics to per-device stats
Linus Torvalds [Mon, 4 Jan 2010 20:28:26 +0000 (12:28 -0800)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ryusuke/nilfs2:
nilfs2: update mailing list address
nilfs2: Storage class should be before const qualifier
nilfs2: trivial coding style fix
Manuel Lauss [Mon, 4 Jan 2010 15:29:49 +0000 (16:29 +0100)]
ASoC: fixup oops in generic AC97 codec glue
Initialize the glue by calling snd_soc_new_ac97_codec() as is done
in other ASoC AC97 codecs. Fixes an oops caused by dereferencing
uninitialized members in snd_soc_new_pcms().
Run-tested on Au1250.
Signed-off-by: Manuel Lauss <manuel.lauss@gmail.com> Acked-by: Liam Girdwood <lrg@slimlogic.co.uk> Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com>
James Smart [Mon, 21 Dec 2009 22:03:15 +0000 (17:03 -0500)]
[SCSI] lpfc 8.3.7: Fix discovery failures.
Fix discovery failures:
- Move all accesses to the fc_flag field inside the host lock.
- Restore link state after going through linkdown processing for FCF DEAD event.
Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Fix hardware/SLI relates issues:
- Fix CNA uses more than one EQ when in INTx interrupt mode.
- Fix driver tries to process failed read FCF record mailbox request.
- Fix allocating single receive buffer breaks FCoE receive queue.
- Support new read FCF record mailbox error case.
Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@suse.de>
James Smart [Mon, 21 Dec 2009 22:01:23 +0000 (17:01 -0500)]
[SCSI] lpfc 8.3.7: Fix FC protocol errors
Fix FC protocol errors:
- Fix multi-frame unsolicited sequences not queued properly
- Fix frames for unsolicited sequences not being associated with sequence.
- Fix unsolicited frame buffer sizes are not set properly
- Fix Sequence count for unsolicited frame headers not byte swapped.
- Fix Multi-frame sequence response frames go to wrong DID.
Signed-off-by: James Smart <james.smart@emulex.com> Signed-off-by: James Bottomley <James.Bottomley@suse.de>
Ed Lin [Sat, 19 Dec 2009 01:34:51 +0000 (17:34 -0800)]
[SCSI] stex: fix scan of nonexistent lun
During a manual scan, a user can send command to a nonexistent
lun, precisely at the point of max_lun. Normally it's possible
(but not required) that the firmware has the knowledge that it
is an invalid lun. In the particular case when max_lun is 256,
however, the nonexistent lun 256 will be confused with lun 0,
because the lun member in a request message is only u8, and 256
will become 0. So we need to fix the problem, at least, at the
driver level.
Signed-off-by: Ed Lin <ed.lin@promise.com> Signed-off-by: James Bottomley <James.Bottomley@suse.de>
otg_ulpi_create is defined in drivers/usb/otg/ulpi.c which depends on
CONFIG_USB_ULPI. So protect its usage by the same symbol. Moreover
mxc_ulpi_access_ops needs CONFIG_MXC_ULPI.
Wolfgang Denk [Mon, 14 Dec 2009 23:27:42 +0000 (00:27 +0100)]
ARM: MX3: make CPU revision number detection work on all boards
Commit 52939c03 (ARM: MX3: fix CPU revision number detection) started
using the CPU's SREV register for revision number detection. This
makes it mandatory to have a valid SPBA0 mapping. Add this to the
global map_io code instead of adding multiple copies for each board.
Signed-off-by: Wolfgang Denk <wd@denx.de> Cc: Daniel Mack <daniel@caiaq.de> Cc: Sascha Hauer <s.hauer@pengutronix.de>
Tested on Qong (EVB-Lite) Tested-by: Wolfgang Denk <wd@denx.de> Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
Jiri Slaby [Thu, 19 Nov 2009 16:16:37 +0000 (17:16 +0100)]
resource: add helpers for fetching rlimits
We want to be sure that compiler fetches the limit variable only
once, so add helpers for fetching current and maximal resource
limits which do that.
Add them to sched.h (instead of resource.h) due to circular dependency
sched.h->resource.h->task_struct
Alternative would be to create a separate res_access.h or similar.
Signed-off-by: Jiri Slaby <jslaby@suse.cz> Cc: James Morris <jmorris@namei.org> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Ingo Molnar <mingo@elte.hu>
Don't pass current RLIMIT_RTTIME to update_rlimit_cpu() in
selinux_bprm_committing_creds, since update_rlimit_cpu expects
RLIMIT_CPU limit.
Use proper rlim[RLIMIT_CPU].rlim_cur instead to fix that.
Signed-off-by: Jiri Slaby <jirislaby@gmail.com> Acked-by: James Morris <jmorris@namei.org> Cc: Stephen Smalley <sds@tycho.nsa.gov> Cc: Eric Paris <eparis@parisplace.org> Cc: David Howells <dhowells@redhat.com>
[S390] Have param.h simply include <asm-generic/param.h>.
Since the files have identical content, might as well simplify.
Signed-off-by: Robert P. J. Day <rpjday@crashcourse.ca> Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Jan Glauber [Mon, 4 Jan 2010 08:05:42 +0000 (09:05 +0100)]
[S390] qdio: convert global statistics to per-device stats
Revamp the qdio performance statistics and move them from procfs to
debugfs using the seq_file interface. Since the statistics are not
intended for the general user the removal of /proc/qdio_perf should
not surprise anyone.
The per device statistics are disabled by default, writing 1 to
/<debugfs mountpoint>/qdio/<device bus ID>/statistics enables the
statistics for the given device.
Signed-off-by: Jan Glauber <jang@linux.vnet.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Paul Mundt [Mon, 4 Jan 2010 02:16:33 +0000 (11:16 +0900)]
sh: Don't default enable PMB support.
This has the adverse effect of converting many 29bit configs to 32bit
mode, while this is a change that needs to be done manually for each
platform. Turn it off by default in order to cut down on spurious bug
reports.
Paul Mundt [Mon, 4 Jan 2010 02:13:54 +0000 (11:13 +0900)]
sh: Disable PMB for SH4AL-DSP CPUs.
While the PMB is available on SH-4A parts, SH4AL-DSP parts exclude it
altogether. As such, explicitly disable PMB support for these parts. If
this changes in the future for newer subtypes, this will have to be made
more fine-grained.
reiserfs: Fix unwanted recursive reiserfs lock in reiserfs_unlink()
reiserfs_unlink() may or may not be called under the reiserfs
lock.
But it also takes the reiserfs lock and can then acquire it
recursively which leads to do_journal_begin_r() that fails to
relax the reiserfs lock before grabbing the journal mutex,
creating an unexpected lock inversion.
We need to ensure reiserfs_unlink() won't get the reiserfs lock
recursively using reiserfs_write_lock_once().
This fixes the following warning that precedes a lock inversion
report (reiserfs lock <-> journal mutex).
reiserfs: Relax lock before open xattr dir in reiserfs_xattr_set_handle()
We call xattr_lookup() from reiserfs_xattr_get(). We then hold
the reiserfs lock when we grab the i_mutex. But later, we may
relax the reiserfs lock, creating dependency inversion between
both locks.
The lookups and creation jobs ar already protected by the
inode mutex, so we can safely relax the reiserfs lock, dropping
the unwanted reiserfs lock -> i_mutex dependency, as shown
in the following lockdep report:
=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.32-atom #173
-------------------------------------------------------
cp/3204 is trying to acquire lock:
(&REISERFS_SB(s)->lock){+.+.+.}, at: [<c11432b9>] reiserfs_write_lock_once+0x29/0x50
but task is already holding lock:
(&sb->s_type->i_mutex_key#4/3){+.+.+.}, at: [<c1141e18>] open_xa_dir+0xd8/0x1b0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
reiserfs: Relax reiserfs lock while freeing the journal
Keeping the reiserfs lock while freeing the journal on
umount path triggers a lock inversion between bdev->bd_mutex
and the reiserfs lock.
We don't need the reiserfs lock at this stage. The filesystem
is not usable anymore, and there are no more pending commits,
everything got flushed (even this operation was done in parallel
and didn't required the reiserfs lock from the current process).
This fixes the following lockdep report:
=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.32-atom #172
-------------------------------------------------------
umount/3904 is trying to acquire lock:
(&bdev->bd_mutex){+.+.+.}, at: [<c10de2c2>] __blkdev_put+0x22/0x160
but task is already holding lock:
(&REISERFS_SB(s)->lock){+.+.+.}, at: [<c1143279>] reiserfs_write_lock+0x29/0x40
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
reiserfs: Fix reiserfs lock <-> i_mutex dependency inversion on xattr
While deleting the xattrs of an inode, we hold the reiserfs lock
and grab the inode->i_mutex of the targeted inode and the root
private xattr directory.
Later on, we may relax the reiserfs lock for various reasons, this
creates inverted dependencies.
We can remove the reiserfs lock -> i_mutex dependency by relaxing
the former before calling open_xa_dir(). This is fine because the
lookup and creation of xattr private directories done in
open_xa_dir() are covered by the targeted inode mutexes. And deeper
operations in the tree are still done under the write lock.
This fixes the following lockdep report:
=======================================================
[ INFO: possible circular locking dependency detected ]
2.6.32-atom #173
-------------------------------------------------------
cp/3204 is trying to acquire lock:
(&REISERFS_SB(s)->lock){+.+.+.}, at: [<c11432b9>] reiserfs_write_lock_once+0x29/0x50
but task is already holding lock:
(&sb->s_type->i_mutex_key#4/3){+.+.+.}, at: [<c1141e18>] open_xa_dir+0xd8/0x1b0
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
When we relax the reiserfs lock to avoid creating unwanted
dependencies against others locks while grabbing these,
we want to ensure it has not been taken recursively, otherwise
the lock won't be really relaxed. Only its depth will be decreased.
The unwanted dependency would then actually happen.
To prevent from that, add a reiserfs_lock_check_recursive() call
in the places that need it.
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Cc: Alexander Beregalov <a.beregalov@gmail.com> Cc: Chris Mason <chris.mason@oracle.com> Cc: Ingo Molnar <mingo@elte.hu>
i_xattr_sem depends on the reiserfs lock. But after we grab
i_xattr_sem, we may relax/relock the reiserfs lock while waiting
on a freezed filesystem, creating a dependency inversion between
the two locks.
In order to avoid the i_xattr_sem -> reiserfs lock dependency, let's
create a reiserfs_down_read_safe() that acts like
reiserfs_mutex_lock_safe(): relax the reiserfs lock while grabbing
another lock to avoid undesired dependencies induced by the
heivyweight reiserfs lock.
Theodore Ts'o [Fri, 1 Jan 2010 07:41:30 +0000 (02:41 -0500)]
ext4: Calculate metadata requirements more accurately
In the past, ext4_calc_metadata_amount(), and its sub-functions
ext4_ext_calc_metadata_amount() and ext4_indirect_calc_metadata_amount()
badly over-estimated the number of metadata blocks that might be
required for delayed allocation blocks. This didn't matter as much
when functions which managed the reserved metadata blocks were more
aggressive about dropping reserved metadata blocks as delayed
allocation blocks were written, but unfortunately they were too
aggressive. This was fixed in commit 0637c6f, but as a result the
over-estimation by ext4_calc_metadata_amount() would lead to reserving
2-3 times the number of pending delayed allocation blocks as
potentially required metadata blocks. So if there are 1 megabytes of
blocks which have been not yet been allocation, up to 3 megabytes of
space would get reserved out of the user's quota and from the file
system free space pool until all of the inode's data blocks have been
allocated.
This commit addresses this problem by much more accurately estimating
the number of metadata blocks that will be required. It will still
somewhat over-estimate the number of blocks needed, since it must make
a worst case estimate not knowing which physical blocks will be
needed, but it is much more accurate than before.
Linus Torvalds [Fri, 1 Jan 2010 01:26:05 +0000 (17:26 -0800)]
twl4030-irq.c: fix compiler warning due to raw-spinlock conversion
Commit 239007b844 ("genirq: Convert irq_desc.lock to raw_spinlock")
seems to have missed this driver, leaving it to use the normal spin-lock
functions for the irq descriptor lock, which is now a raw lock.
Reported-and-compile-tested-by: Stephen Rothwell <sfr@canb.auug.org.au> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Linus Torvalds [Fri, 1 Jan 2010 00:44:43 +0000 (16:44 -0800)]
pci: avoid compiler warning in quirks.c
Introduced by commit 5b889bf23 ("PCI: Fix build if quirks are not
enabled"), which made the pci_dev_reset_methods[] array static and
'const', but didn't then change the code to match, and use a const
pointer when moving it to quirks.c.
Trivially fixed by just adding the required 'const' to the iterator
variable.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au> Cc: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* git://git.kernel.org/pub/scm/linux/kernel/git/sfrench/cifs-2.6:
[CIFS] Enable mmap on forcedirectio mounts
cifs: NULL out tcon, pSesInfo, and srvTcp pointers when chasing DFS referrals
Linus Torvalds [Thu, 31 Dec 2009 20:01:24 +0000 (12:01 -0800)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
firewire, ieee1394: update Kconfig help
firewire, ieee1394: update MAINTAINERS entries
firewire: ohci: always use packet-per-buffer mode for isochronous reception
firewire: cdev: fix another memory leak in an error path
firewire: fix use of multiple AV/C devices, allow multiple FCP listeners
Comments from Stefan:
Distributors who still ship the old stack (ieee1394, ohci1394,
raw1394, sbp2, eth1394 and more) should now switch to the new one
(firewire-core, firewire-ohci, firewire-sbp2, firewire-net). In the
first iteration, those distributors might want to ship the old stack
also (but blacklisted) as a fallback for their users if unforeseen
problems with the newer replacement drivers are encountered.
The older FireWire stack contains several known problems which are
not going to be fixed; instead, those issues are addressed by the new
stack. An incomplete list of these issues is kept in bugzilla:
http://bugzilla.kernel.org/show_bug.cgi?id=10046
We have a guide on migration from the older to the newer stack:
Linus Torvalds [Thu, 31 Dec 2009 19:54:13 +0000 (11:54 -0800)]
Merge branch 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
* 'x86-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
x86/agp: Fix agp_amd64_init() initialization with CONFIG_GART_IOMMU enabled
x86: SGI UV: Fix writes to led registers on remote uv hubs
x86, kmemcheck: Use KERN_WARNING for error reporting
x86: Use KERN_DEFAULT log-level in __show_regs()
x86, compress: Force i386 instructions for the decompressor
x86/amd-iommu: Fix initialization failure panic
dma-debug: Do not add notifier when dma debugging is disabled.
x86: Fix objdump version check in chkobjdump.awk for different formats.
Trivial conflicts in arch/x86/include/asm/uv/uv_hub.h due to me having
applied an earlier version of an SGI UV fix.
Linus Torvalds [Thu, 31 Dec 2009 19:52:24 +0000 (11:52 -0800)]
Merge branch 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
* 'perf-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
perf kmem: Fix statistics typo
kprobes: Fix distinct type warning
perf: Rename perf_event_hw_event in design document
perf tools: Add missing header files to LIB_H Makefile variable
perf record: We should fork only if a program was specified to run
perf diff: Fix usage array, it must end with a NULL entry
Linus Torvalds [Thu, 31 Dec 2009 19:52:01 +0000 (11:52 -0800)]
Merge branch 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip
* 'tracing-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/linux-2.6-tip:
tracing: Fix sign fields in ftrace_define_fields_##call()
tracing/syscalls: Fix typo in SYSCALL_DEFINE0
tracing/kprobe: Show sign of fields in trace_kprobe format files
ksym_tracer: Remove trace_stat
ksym_tracer: Fix race when incrementing count
ksym_tracer: Fix to allow writing newline to ksym_trace_filter
ksym_tracer: Fix to make the tracer work
tracing: Kconfig spelling fixes and cleanups
tracing: Fix setting tracer specific options
Documentation: Update ftrace-design.txt
Documentation: Update tracepoint-analysis.txt
Documentation: Update mmiotrace.txt
Linus Torvalds [Thu, 31 Dec 2009 19:12:35 +0000 (11:12 -0800)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
Input: iforce - wait for command completion when closing the device
Input: twl4030-pwrbutton - switch to using threaded IRQ
Input: twl4030_keypad - switch to using threaded IRQ
Input: lifebook - add CONFIG_DMI dependency
Input: wistron - fix test for CONFIG_PM
Input: psmouse - fix compile warning in hgpk module
Input: matrix-keypad - handle cases when GPIOs can't be wakeup sources
Input: iforce - fix oops on device disconnect
Input: ff-memless - add notion of direction to for rumble effects
Input: ff-memless - another fix for signed to unsigned overflow
Input: ff-memless - start playing FF effects immediately
Input: serio - do not mark kseriod freezable anymore
Input: speed up suspend/shutdown for PS/2 mice and keyboards