| Age | Commit message (Collapse) | Author |
|---|
|
This is a build fix required after "x86-64: seccomp: fix 32/64 syscall
hole" (commit 5b1017404aea6d2e552e991b3fd814d839e9cd67). MIPS doesn't
have the issue that was fixed for x86-64 by that patch.
This also doesn't solve the N32 issue which is that N32 seccomp processes
will be treated as non-compat processes thus only have access to N64
syscalls.
Signed-off-by: Ralf Baechle <ralf@linux-mips.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
|
|
Impact: improve out-of-range fixmap index debugging
Commit "1b42f51630c7eebce6fb780b480731eb81afd325"
defined the __this_fixmap_does_not_exist() function
with a WARN_ON(1) in it.
This causes the linker to not report an error when
__this_fixmap_does_not_exist() is called with a
non-constant parameter.
Ingo defined __this_fixmap_does_not_exist() because he
wanted to get virt addresses of fix memory of nest level
by non-constant index.
But we can fix this and still keep the link-time check:
We can get the four slot virt addresses on link time and
store them to array slot_virt[].
Then we can then refer the slot_virt with non-constant index,
in the ioremap-leak detection code.
Signed-off-by: Wang Chen <wangchen@cn.fujitsu.com>
LKML-Reference: <49B2075B.4070509@cn.fujitsu.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
Impact: cleanup and code size reduction on 64-bit
This code is only applied to Intel Pentium and AMD K7 32-bit cpus.
Move those checks to intel_init()/amd_init() for 32-bit
so 64-bit will not build this code.
Also change to use cpu_index check to see if we need to emit warning.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <49B377D2.8030108@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
Commit 0d3e0460f307e84904968aad6cff97bd688583d8
"MMC: CSD and CID timeout values" inadvertently broke
the timeout for the MMC command SEND_EXT_CSD.
This patch puts it back again.
Depending on the characteristics of the controller,
this bug may prevent the use of MMC cards.
Signed-off-by: Adrian Hunter <adrian.hunter@nokia.com>
Signed-off-by: Pierre Ossman <drzeus@drzeus.cx>
|
|
In uv_flush_tlb_others() (arch/x86/kernel/tlb_uv.c),
the "WARN_ON(!in_atomic())" fails if CONFIG_PREEMPT is not enabled.
And CONFIG_PREEMPT is not enabled by default in the distribution that
most UV owners will use.
We could #ifdef CONFIG_PREEMPT the warning, but that is not good form.
And there seems to be no suitable fix to in_atomic() when CONFIG_PREMPT
is not on.
As Ingo commented:
> and we have no proper primitive to test for atomicity. (mainly
> because we dont know about atomicity on a non-preempt kernel)
So we drop the WARN_ON.
Signed-off-by: Cliff Wickman <cpw@sgi.com>
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
Protocol 0x37 has been reserved for iNexio devices and Sahara
was supposed to get 0x38.
Reported-by: Claudio Nieder <private@claudio.ch>
Signed-off-by: Dmitry Torokhov <dtor@mail.ru>
|
|
Following a thread on busybox mailing list
depmod -r option is ignored by module-init-tools depmod
-r option break busybox depmod.
So the best solution look to remove -r from kernel Makefile
Signed-off-by: Gilles Espinasse <g.esp@free.fr>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
|
|
Running 'make rpm' fails when CONFIG_LOCALVERSION_AUTO=y and using a kernel source
tree under SCM. This is due to KERNELRELEASE being different when the initial make
is run and when make is run from rpmbuild.
mkspec creates kernel.spec using KERNELRELEASE:
<mkspec>
echo "%files"
echo '%defattr (-, root, root)'
echo "%dir /lib/modules"
echo "/lib/modules/$KERNELRELEASE"
echo "/lib/firmware"
echo "/boot/*"
echo ""
</mkspec>
When CONFIG_LOCALVERSION_AUTO=y scripts/setlocalversion is called and grabs any
additional version info from SCM. Next, the srctree is tarred up and SCM
information is excluded.
rpmbuild reruns make and in the process generates a new include/config/kernel.release
and thus a new KERNELRELEASE. However this time the SCM information is gone so
KERNELRELEASE no longer has the additional version information. When "make modules_install"
runs, it uses the new KERNELRELEASE value to determine where to install the modules.
This conflicts with where the spec file assumes they are going because of the
mis-matching KERNELRELEASE versions.
<snippet>
+ INSTALL_MOD_PATH=/var/tmp/kernel-2.6.29rc4tip01479g5d85422-root
+ make -j16 modules_install
INSTALL crypto/aead.ko
INSTALL crypto/cbc.ko
INSTALL crypto/chainiv.ko
INSTALL crypto/crc32c.ko
INSTALL crypto/crypto_algapi.ko
INSTALL crypto/crypto_blkcipher.ko
INSTALL crypto/crypto_hash.ko
INSTALL crypto/cryptomgr.ko
INSTALL crypto/ecb.ko
INSTALL crypto/eseqiv.ko
INSTALL crypto/krng.ko
INSTALL crypto/md5.ko
INSTALL crypto/pcbc.ko
INSTALL crypto/rng.ko
INSTALL drivers/block/cciss.ko
INSTALL drivers/hid/hid-dummy.ko
INSTALL drivers/scsi/iscsi_tcp.ko
INSTALL drivers/scsi/libiscsi.ko
INSTALL drivers/scsi/libiscsi_tcp.ko
INSTALL drivers/scsi/scsi_transport_iscsi.ko
INSTALL drivers/scsi/scsi_wait_scan.ko
INSTALL fs/lockd/lockd.ko
INSTALL fs/nfs/nfs.ko
INSTALL fs/nfsd/nfsd.ko
INSTALL lib/libcrc32c.ko
INSTALL net/sunrpc/sunrpc.ko
DEPMOD 2.6.29-rc4-tip
+ cp arch/x86/boot/bzImage
/var/tmp/kernel-2.6.29rc4tip01479g5d85422-root/boot/vmlinuz-2.6.29-rc4-tip-01479-g5d85422
+ cp System.map
/var/tmp/kernel-2.6.29rc4tip01479g5d85422-root/boot/System.map-2.6.29-rc4-tip-01479-g5d85422
+ cp .config
/var/tmp/kernel-2.6.29rc4tip01479g5d85422-root/boot/config-2.6.29-rc4-tip-01479-g5d85422
+ cp vmlinux vmlinux.orig
+ bzip2 -9 vmlinux
+ mv vmlinux.bz2
/var/tmp/kernel-2.6.29rc4tip01479g5d85422-root/boot/vmlinux-2.6.29-rc4-tip-01479-g5d85422.bz2
+ mv vmlinux.orig vmlinux
+ /usr/lib/rpm/brp-compress
Processing files: kernel-2.6.29rc4tip01479g5d85422-2
error: File not found:
/var/tmp/kernel-2.6.29rc4tip01479g5d85422-root/lib/modules/2.6.29-rc4-tip-01479-g5d85422
RPM build errors:
File not found:
/var/tmp/kernel-2.6.29rc4tip01479g5d85422-root/lib/modules/2.6.29-rc4-tip-01479-g5d85422
make[1]: *** [rpm] Error 1
make: *** [rpm] Error 2
</snippet>
I have tested this patch on git -tip, Linus' git tree, and the kernel.org tar files, both
with and without CONFIG_LOCALVERSION_AUTO=y.
Signed-off-by: Josh Hunt <josh@scalex86.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
----
|
|
The contents of the %clean section in mkspec is currently commented
out leaving RPM_BUILD_ROOT and its contents on the build machine.
This patch removes it once the rpm build process is complete.
Signed-off-by: Josh Hunt <josh@scalex86.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
|
|
This fixes an error when compiling the kernel.
CHK include/linux/version.h
HOSTCC scripts/unifdef
scripts/unifdef.c:209: error: conflicting types for 'getline'
/usr/include/stdio.h:651: note: previous declaration of 'getline' was here
make[1]: *** [scripts/unifdef] Error 1
make: *** [__headers] Error 2
Signed-off-by: Justin P. Mattock <justinmattock@gmail.com>
Cc: Frederic Weisbecker <fweisbec@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
|
|
Impact: fix deadlock and allow atomic free
Percpu allocation always uses GFP_KERNEL and whole alloc/free paths
were protected by single mutex. All percpu allocations have been from
GFP_KERNEL-safe context and the original allocator had this assumption
too. However, by protecting both alloc and free paths with the same
mutex, the new allocator creates free -> alloc -> GFP_KERNEL
dependency which the original allocator didn't have. This can lead to
deadlock if free is called from FS or IO paths. Also, in general,
allocators are expected to allow free to be called from atomic
context.
This patch implements finer grained locking to break the deadlock and
allow atomic free. For details, please read the "Synchronization
rules" comment.
While at it, also add CONTEXT: to function comments to describe which
context they expect to be called from and what they do to it.
This problem was reported by Thomas Gleixner and Peter Zijlstra.
http://thread.gmane.org/gmane.linux.kernel/802384
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Thomas Gleixner <tglx@linutronix.de>
Reported-by: Peter Zijlstra <peterz@infradead.org>
|
|
Currently we unconditionally issue a flush from xfs_free_buftarg, but
since 2.6.29-rc1 this gives a warning in the style of
end_request: I/O error, dev vdb, sector 0
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Felix Blyakher <felixb@sgi.com>
|
|
The inode can't be locked by anyone else as we just created it a few
lines above and it's not been added to any lookup data structure yet.
So use a trylock that must succeed to get around the lockdep warnings.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Alexander Beregalov <a.beregalov@gmail.com>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
Reviewed-by: Felix Blyakher <felixb@sgi.com>
Signed-off-by: Felix Blyakher <felixb@sgi.com>
|
|
Andras Korn reported an oops on log replay causes by a corrupted
xfs_inode_log_format_t passing a 0 size to kmem_zalloc. This patch handles
to small or too large numbers of log regions gracefully by rejecting the
log replay with a useful error message.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reported-by: Andras Korn <korn-sgi.com@chardonnay.math.bme.hu>
Reviewed-by: Eric Sandeen <sandeen@sandeen.net>
Signed-off-by: Felix Blyakher <felixb@sgi.com>
|
|
|
|
Made the comments more like the comments for struct scsi_host_template.
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
If we've just created an interface and the an rport is
logging in we may have a request on the wire (say PRLI).
If we destroy the interface, we'll go through each rport
on the disc->rports list and set each rport's state to NONE.
Then the lport will reset the EM. The EM reset will send a
CLOSED event to the prli_resp() handler which will notice
that the state != PRLI. In this case it frees the frame
pointer, decrements the refcount and unlocks the rport.
The problem is that there isn't a frame in this case. It's
just a pointer with an embedded error code. The free causes
an Oops.
This patch moves the error checking to be before the state
checking.
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
Just rename the variable as per our naming convention.
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
We only need to use this macro when assigning a value to
rport->dd_data. All other accesses should just use dd_data.
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
When a sequence cannot be delivered to the target, the local
port will schedule retries, While this process is in progress,
if we destroy the FCoE interface, the fcoe_sw_destroy routine is
entered, and the fc_exch_mgr_free(lp->emp) is called. Thus
if fc_exch_alloc() is called when retrying the sequence,
the mempool_alloc() will fail to allocate the exchange because
the mempool of the exchange manager has already been released.
This patch is to cancel any pending retry work of the local
port before we start to destroy the interface.
Also, when resetting the local port, we should also stop the
scheduled pending retries.
Signed-off-by: Steve Ma <steve.ma@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
The fc_fcp_complete_locked detected data underrun in this case and set
the FC_DATA_UNDRUN but that was ignored by fc_io_compl for all cases
including read underrun.
Added code to not to ignore FC_DATA_UNDRUN for read IO and instead
suggested scsi-ml to retry cmd to recover from lost data frame.
Not sure if it is okay to ignore FC_DATA_UNDRUN for other case, so let
code as is for other cases but removed or-ing with zero valued fsp->cdb_status
for those cases.
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
This allows any rport ELS to retry on LS_RJT.
The rport error handling would only retry on resource allocation failures
and exchange timeouts. I have a target that will occasionally reject PLOGI
when we do a quick LOGO/PLOGI. When a critical ELS was rejected, libfc would
fail silently leaving the rport in a dead state.
The retry count and delay are managed by fc_rport_error_retry. If the retry
count is exceeded fc_rport_error will be called. When retrying is not the
correct course of action, fc_rport_error can be called directly.
Signed-off-by: Chris Leech <christopher.leech@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
lport->link_status
The fcoe_xmit could call fc_pause in case the pending skb queue len is larger
than FCOE_MAX_QUEUE_DEPTH, the fc_pause was trying to grab lport->lp_muex to
change lport->link_status and that had these issues :-
1. The fcoe_xmit was getting called with bh disabled, thus causing
"BUG: scheduling while atomic" when grabbing lport->lp_muex with bh disabled.
2. fc_linkup and fc_linkdown function calls lport_enter function with
lport->lp_mutex held and these enter function in turn calls fcoe_xmit to send
lport related FC frame, e.g. fc_linkup => fc_lport_enter_flogi to send flogi
req. In this case grabbing the same lport->lp_mutex again in fc_puase from
fcoe_xmit would cause deadlock.
The lport->lp_mutex was used for setting FC_PAUSE in fcoe_xmit path but
FC_PAUSE bit was not used anywhere beside just setting and clear this
bit in lport->link_status, instead used a separate field qfull in fc_lport
to eliminate need for lport->lp_mutex to track pending queue full condition
and in turn avoid above described two locking issues.
Also added check for lp->qfull in fc_fcp_lport_queue_ready to trigger
SCSI_MLQUEUE_HOST_BUSY when lp->qfull is set to prevent more scsi-ml cmds
while lp->qfull is set.
This patch eliminated FC_LINK_UP and FC_PAUSE and instead used dedicated
fields in fc_lport for this, this simplified all related conditional
code.
Also removed fc_pause and fc_unpause functions and instead used newly added
lport->qfull directly in fcoe.
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
The fc_seq_start_next grabs ep->ex_lock but this lock was already held here,
so instead called fc_seq_start_next_locked to avoid soft lockup.
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
Cleanup exchange held due to RRQ when RRQ exch times out, in this case the
ABTS is already done causing RRQ req therefore proceeding with cleanup in
fc_exch_rrq_resp should be okay to restore exch resource.
Signed-off-by: Vasu Dev <vasu.dev@intel.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
When a rport goes away, libFC does a plogi which will reset exchanges
at the rport. Clean exchanges at our end, both in transport and libFC.
If transport hooks into exch_mgr_reset, it will call back into
fc_exch_mgr_reset() to clean up libFC exchanges.
Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
fc_exch_mgr structure is private to fc_exch.c. To export exch_mgr_reset to
transport, transport needs access to the exch manager. Change
exch_mgr_reset to use lport param which is the shared structure between
libFC and transport.
Alternatively, fc_exch_mgr definition can be moved to libfc.h so that lport
can be accessed from mp*.
Signed-off-by: Abhijeet Joglekar <abjoglek@cisco.com>
Signed-off-by: Robert Love <robert.w.love@intel.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
|
|
This patch fixes a number of race conditions in the driver.
Up until now, "entry" pointer was initialized before acquiring the right lock.
Signed-off-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
freq_diff is unsigned, so test before subtraction
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: John W. Linville <linville@tuxdriver.com>
|
|
We recently discovered a problem with passing of DMA attributes on SN
systems with the older PIC chips.
[akpm@linux-foundation.org: coding-style fixes]
Signed-off-by: Jeremy Higdon <jeremy@sgi.com>
Cc: <habeck@sgi.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Tony Luck <tony.luck@intel.com>
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6 into s3c-fixes
|
|
Includes missed irqs.h in devices.c and mx1ads.c.
Signed-off-by: Darius Augulis <augulis.darius@gmail.com>
Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
|
|
Stephen Rothwell reported:
|Today's linux-next build (x86_64 allmodconfig) produced this warning:
|
|In file included from drivers/char/epca.c:49:
|drivers/char/digiFep1.h:7:1: warning: "GLOBAL" redefined
|In file included from include/linux/linkage.h:5,
| from include/linux/kernel.h:11,
| from arch/x86/include/asm/system.h:10,
| from arch/x86/include/asm/processor.h:17,
| from include/linux/prefetch.h:14,
| from include/linux/list.h:6,
| from include/linux/module.h:9,
| from drivers/char/epca.c:29:
|arch/x86/include/asm/linkage.h:55:1: warning: this is the location of the previous definition
|
|Probably introduced by commit 95695547a7db44b88a7ee36cf5df188de267e99e
|("x86: asm linkage - introduce GLOBAL macro") from the x86 tree.
Any assembler specific snippets being placed in headers
are to be protected by __ASSEMBLY__. Fixed.
Also move __ALIGN definition under the same protection as well.
Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Cyrill Gorcunov <gorcunov@openvz.org>
LKML-Reference: <20090306160833.GB7420@localhost>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
Impact: code reorganization for later changes
Do fully free chunk reclamation using a work. This change is to
prepare for locking changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
Impact: code reorganization for later changes
Separate out chunk area map extension into a separate function -
pcpu_extend_area_map() - and call it directly from pcpu_alloc() such
that pcpu_alloc_area() is guaranteed to have enough area map slots on
invocation.
With this change, pcpu_alloc_area() does only area allocation and the
only failure mode is when the chunk doens't have enough room, so
there's no need to distinguish it from memory allocation failures.
Make it return -1 on such cases instead of hacky -ENOSPC.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
Impact: code reorganization for later changes
With static map handling moved to pcpu_split_block(), pcpu_realloc()
only clutters the code and it's also unsuitable for scheduled locking
changes. Implement and use pcpu_mem_alloc/free() instead.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
ds_write_config() can write the BTS as well as the PEBS part of
the DS config. ds_request_pebs() passes the wrong qualifier, which
results in the wrong configuration to be written.
Reported-by: Stephane Eranian <eranian@googlemail.com>
Signed-off-by: Markus Metzger <markus.t.metzger@intel.com>
LKML-Reference: <20090305085721.A22550@sedona.ch.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
In case a ptraced task is reaped (while the tracer is still attached),
ds_exit_thread() is called before ptrace_exit(). The latter will
release the bts_tracer and remove the thread's ds_ctx.
The former will WARN() if the context is not NULL.
Oleg Nesterov submitted patches that move ptrace_exit() before
exit_thread() and thus reverse the order of the above calls.
Remove the bad warning. I will add it again when Oleg's changes are in.
Signed-off-by: Markus Metzger <markus.t.metzger@intel.com>
LKML-Reference: <20090305084954.A22000@sedona.ch.intel.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
As a preparational step for unifying noexec handling on 32-bit and 64-bit,
rename the do_not_nx variable to disable_nx on 64-bit.
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
LKML-Reference: <1236265497.31324.11.camel@penberg-laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
Signed-off-by: Pekka Enberg <penberg@cs.helsinki.fi>
LKML-Reference: <1236265466.31324.9.camel@penberg-laptop>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
Impact: make boot command line "memtest" do one loop by default
So don't need to guess many patterns in one loop.
Signed-off-by: Yinghai Lu <yinghai@kernel.org>
LKML-Reference: <49B10532.3020105@kernel.org>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
|
|
The recent changes over the DAC detection mechanism in patch_sigmatel.c
breaks the HP detection on the machines with multiple HP jacks.
It's basically because of the workaround to support the multi-channel
output. Since the HP detection is more important feature, disable
the HP-swap workaroud temporarily.
Reference: Novell bnc#482052
https://bugzilla.novell.com/show_bug.cgi?id=482052
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
"Headphone Playback ..." appears twice in slave_vols[] and slave_sws[].
They should be "Headphone Playback2 ..."
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Commit 1e42807918d17e8c93bf14fbb74be84b141334c1 introduced a bug where we
don't get front/back segment sizes in the bio in blk_recount_segments().
Fix this by tracking the back bio as well as the front bio in
__blk_recalc_rq_segments(), this also cleans up the interface by getting
rid of the segment size pointer passing.
Tested-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
|
|
Impact: fix relocation overflow during module load
x86_64 uses 32bit relocations for symbol access and static percpu
symbols whether in core or modules must be inside 2GB of the percpu
segement base which the dynamic percpu allocator doesn't guarantee.
This patch makes x86_64 reserve PERCPU_MODULE_RESERVE bytes in the
first chunk so that module percpu areas are always allocated from the
first chunk which is always inside the relocatable range.
This problem exists for any percpu allocator but is easily triggered
when using the embedding allocator because the second chunk is located
beyond 2GB on it.
This patch also changes the meaning of PERCPU_DYNAMIC_RESERVE such
that it only indicates the size of the area to reserve for dynamic
allocation as static and dynamic areas can be separate. New
PERCPU_DYNAMIC_RESERVED is increased by 4k for both 32 and 64bits as
the reserved area separation eats away some allocatable space and
having slightly more headroom (currently between 4 and 8k after
minimal boot sans module area) makes sense for common case
performance.
x86_32 can address anywhere from anywhere and doesn't need reserving.
Mike Galbraith first reported the problem first and bisected it to the
embedding percpu allocator commit.
Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Mike Galbraith <efault@gmx.de>
Reported-by: Jaswinder Singh Rajput <jaswinder@kernel.org>
|
|
variables
Impact: add reserved allocation functionality and use it for module
percpu variables
This patch implements reserved allocation from the first chunk. When
setting up the first chunk, arch can ask to set aside certain number
of bytes right after the core static area which is available only
through a separate reserved allocator. This will be used primarily
for module static percpu variables on architectures with limited
relocation range to ensure that the module perpcu symbols are inside
the relocatable range.
If reserved area is requested, the first chunk becomes reserved and
isn't available for regular allocation. If the first chunk also
includes piggy-back dynamic allocation area, a separate chunk mapping
the same region is created to serve dynamic allocation. The first one
is called static first chunk and the second dynamic first chunk.
Although they share the page map, their different area map
initializations guarantee they serve disjoint areas according to their
purposes.
If arch doesn't setup reserved area, reserved allocation is handled
like any other allocation.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
Impact: allow sharing page map, no functional difference yet
Make chunk->page access indirect by adding a pointer and renaming the
actual array to page_ar. This will be used by future changes.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
Impact: reduce unnecessary memory usage on certain configurations
Embedding percpu allocator allocates unit_size *
smp_num_possible_cpus() bytes consecutively and use it for the first
chunk. However, if the static area is small, this can result in
excessive prellocated free space in the first chunk due to
PCPU_MIN_UNIT_SIZE restriction.
This patch makes embedding percpu allocator preallocate only what's
necessary as described by PERPCU_DYNAMIC_RESERVE and return the
leftover to the bootmem allocator.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
|
Impact: argument semantic cleanup
In pcpu_setup_first_chunk(), zero @unit_size and @dyn_size meant
auto-sizing. It's okay for @unit_size as 0 doesn't make sense but 0
dynamic reserve size is valid. Alos, if arch @dyn_size is calculated
from other parameters, it might end up passing in 0 @dyn_size and
malfunction when the size is automatically adjusted.
This patch makes both @unit_size and @dyn_size ssize_t and use -1 for
auto sizing.
Signed-off-by: Tejun Heo <tj@kernel.org>
|
