Paul Jackson 4247bdc600 [PATCH] cpuset semaphore depth check deadlock fix ... The cpusets-formalize-intermediate-gfp_kernel-containment patch has a deadlock problem. This patch was part of a set of four patches to make more extensive use of the cpuset 'mem_exclusive' attribute to manage kernel GFP_KERNEL memory allocations and to constrain the out-of-memory (oom) killer. A task that is changing cpusets in particular ways on a system when it is very short of free memory could double trip over the global cpuset_sem semaphore (get the lock and then deadlock trying to get it again). The second attempt to get cpuset_sem would be in the routine cpuset_zone_allowed(). This was discovered by code inspection. I can not reproduce the problem except with an artifically hacked kernel and a specialized stress test. In real life you cannot hit this unless you are manipulating cpusets, and are very unlikely to hit it unless you are rapidly modifying cpusets on a memory tight system. Even then it would be a rare occurence. If you did hit it, the task double tripping over cpuset_sem would deadlock in the kernel, and any other task also trying to manipulate cpusets would deadlock there too, on cpuset_sem. Your batch manager would be wedged solid (if it was cpuset savvy), but classic Unix shells and utilities would work well enough to reboot the system. The unusual condition that led to this bug is that unlike most semaphores, cpuset_sem _can_ be acquired while in the page allocation code, when __alloc_pages() calls cpuset_zone_allowed. So it easy to mistakenly perform the following sequence: 1) task makes system call to alter a cpuset 2) take cpuset_sem 3) try to allocate memory 4) memory allocator, via cpuset_zone_allowed, trys to take cpuset_sem 5) deadlock The reason that this is not a serious bug for most users is that almost all calls to allocate memory don't require taking cpuset_sem. Only some code paths off the beaten track require taking cpuset_sem -- which is good. Taking a global semaphore on the main code path for allocating memory would not scale well. This patch fixes this deadlock by wrapping the up() and down() calls on cpuset_sem in kernel/cpuset.c with code that tracks the nesting depth of the current task on that semaphore, and only does the real down() if the task doesn't hold the lock already, and only does the real up() if the nesting depth (number of unmatched downs) is exactly one. The previous required use of refresh_mems(), anytime that the cpuset_sem semaphore was acquired and the code executed while holding that semaphore might try to allocate memory, is no longer required. Two refresh_mems() calls were removed thanks to this. This is a good change, as failing to get all the necessary refresh_mems() calls placed was a primary source of bugs in this cpuset code. The only remaining call to refresh_mems() is made while doing a memory allocation, if certain task memory placement data needs to be updated from its cpuset, due to the cpuset having been changed behind the tasks back. Signed-off-by: Paul Jackson <pj@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>		2005-09-10 10:06:21 -07:00
..
byteorder
dvb
hdlc
isdn
lockd
mmc	[MMC] Ensure correct mmc_priv() behaviour	2005-09-08 22:46:00 +01:00
mtd
netfilter
netfilter_arp
netfilter_bridge
netfilter_ipv4	[NETFILTER]: kill __ip_ct_expect_unlink_destroy	2005-09-06 15:10:46 -07:00
netfilter_ipv6
nfsd
raid	[PATCH] md: add write-intent-bitmap support to raid5	2005-09-09 16:39:12 -07:00
sunrpc	[PATCH] sunrpc: cache_register can use wrong module reference	2005-09-07 16:57:25 -07:00
tc_act
tc_ematch
8250_pci.h
a.out.h
ac97_codec.h
acct.h
acpi.h
adb.h
adfs_fs_i.h
adfs_fs_sb.h
adfs_fs.h
affs_fs.h
affs_hardblocks.h
agp_backend.h
agpgart.h
aio_abi.h
aio.h
amifd.h
amifdreg.h
amigaffs.h
apm_bios.h
arcdevice.h
arcfb.h
ata.h
atalk.h
atm_eni.h
atm_he.h
atm_idt77105.h
atm_nicstar.h
atm_suni.h
atm_tcp.h
atm_zatm.h
atm.h
atmapi.h
atmarp.h
atmbr2684.h
atmclip.h
atmdev.h
atmioc.h
atmlec.h
atmmpc.h
atmppp.h
atmsap.h
atmsvc.h
attribute_container.h
audit.h
auto_fs4.h
auto_fs.h
auxvec.h	[PATCH] auxiliary vector cleanups	2005-09-07 16:57:21 -07:00
awe_voice.h
ax25.h
b1lli.h
b1pcmcia.h
backing-dev.h
backlight.h
baycom.h
bcd.h
bfs_fs.h	[PATCH] bfs: fix endianness, signedness; add trivial bugfix	2005-09-09 13:57:32 -07:00
binfmts.h
bio.h	Merge master.kernel.org:/pub/scm/linux/kernel/git/jejb/scsi-for-linus-2.6	2005-09-07 17:31:27 -07:00
bit_spinlock.h	[PATCH] spinlock consolidation	2005-09-10 10:06:21 -07:00
bitmap.h
bitops.h
blkdev.h
blkpg.h
blockgroup_lock.h
bootmem.h
bpqether.h
buffer_head.h
cache.h
capability.h	[PATCH] VM: add capabilites check to set_zone_reclaim	2005-09-05 00:05:44 -07:00
capi.h
cciss_ioctl.h
cd1400.h
cdev.h
cdk.h
cdrom.h
chio.h
circ_buf.h
cobalt-nvram.h
coda_cache.h
coda_fs_i.h
coda_linux.h
coda_proc.h
coda_psdev.h
coda.h
coff.h
com20020.h
compat_ioctl.h
compat.h	[PATCH] compat: be more consistent about [ug]id_t	2005-09-07 16:57:19 -07:00
compiler-gcc2.h
compiler-gcc3.h
compiler-gcc4.h
compiler-gcc.h
compiler-intel.h
compiler.h
completion.h
comstats.h
concap.h
config.h
console_struct.h
console.h
consolemap.h
cpu.h
cpufreq.h
cpumask.h
cpuset.h	[PATCH] cpusets: confine oom_killer to mem_exclusive cpuset	2005-09-07 16:57:40 -07:00
cramfs_fs_sb.h
cramfs_fs.h
crash_dump.h
crc16.h	[PATCH] lib/crc16: added crc16 algorithm.	2005-09-08 14:41:27 -07:00
crc32.h
crc32c.h
crc-ccitt.h
crypto.h
cryptohash.h
ctype.h
cuda.h
cyclades.h
cyclomx.h
cycx_cfm.h
cycx_drv.h
cycx_x25.h
dcache.h	[PATCH] struct dentry: place d_hash close to d_parent and d_name to speedup lookups	2005-09-07 16:57:41 -07:00
dccp.h	[DCCP] Introduce dccp_timestamp	2005-09-09 02:38:35 -03:00
dcookies.h
debugfs.h
delay.h
devfs_fs_kernel.h
devfs_fs.h
device-mapper.h
device.h
devpts_fs.h
dio.h
dirent.h
divert.h
dm9000.h
dm-ioctl.h
dma-mapping.h
dmapool.h
dmi.h	[PATCH] dmi: add onboard devices discovery	2005-09-07 16:57:44 -07:00
dn.h
dnotify.h
dqblk_v1.h
dqblk_v2.h
dqblk_xfs.h
ds1286.h
dtlk.h
edd.h
eeprom.h
efi.h	[PATCH] x86: fix EFI memory map parsing	2005-09-05 00:06:09 -07:00
efs_dir.h
efs_fs_i.h
efs_fs_sb.h
efs_fs.h
efs_vh.h
eisa.h
elevator.h
elf-fdpic.h
elf.h	[PATCH] auxiliary vector cleanups	2005-09-07 16:57:21 -07:00
elfcore.h
err.h
errno.h
errqueue.h
etherdevice.h
ethtool.h
eventpoll.h
ext2_fs_sb.h
ext2_fs.h	[PATCH] disk quotas fail when /etc/mtab is symlinked to /proc/mounts	2005-09-07 16:57:23 -07:00
ext3_fs_i.h
ext3_fs_sb.h
ext3_fs.h	[PATCH] disk quotas fail when /etc/mtab is symlinked to /proc/mounts	2005-09-07 16:57:23 -07:00
ext3_jbd.h
fadvise.h
fb.h	[PATCH] fbcon: Break up bit_putcs into its component functions	2005-09-09 14:03:41 -07:00
fcdevice.h
fcntl.h
fd1772.h
fd.h
fddidevice.h
fdreg.h
file.h	[PATCH] files: files struct with RCU	2005-09-09 13:57:55 -07:00
filter.h
firmware.h	[PATCH] modified firmware_class.c to support no hotplug	2005-09-07 16:57:26 -07:00
flat.h
font.h
fs_struct.h
fs.h	[PATCH] files: files struct with RCU	2005-09-09 13:57:55 -07:00
fsl_devices.h
fsnotify.h
ftape-header-segment.h
ftape-vendors.h
ftape.h
fuse.h	[PATCH] FUSE: add fsync operation for directories	2005-09-09 14:03:47 -07:00
futex.h	[PATCH] FUTEX_WAKE_OP: pthread_cond_signal() speedup	2005-09-07 16:57:17 -07:00
gameport.h
gen_stats.h
genalloc.h
generic_serial.h
genhd.h
gfp.h	[PATCH] cpusets: new __GFP_HARDWALL flag	2005-09-07 16:57:40 -07:00
hardirq.h
harrier_defs.h
hash.h
hayesesp.h
hdlc.h
hdlcdrv.h
hdpu_features.h
hdreg.h
hdsmart.h
hiddev.h
highmem.h
highuid.h
hippidevice.h
hpet.h
hpfs_fs.h
hugetlb.h	[PATCH] remove hugetlb_clean_stale_pgtable() and fix huge_pte_alloc()	2005-09-05 00:05:46 -07:00
hwmon-sysfs.h	[PATCH] I2C: W83792D driver 1/3	2005-09-05 09:14:13 -07:00
hwmon-vid.h	[PATCH] hwmon: hwmon vs i2c, second round (10/11)	2005-09-05 09:14:23 -07:00
hwmon.h	[PATCH] hwmon: move SENSORS_LIMIT to hwmon.h	2005-09-05 09:14:17 -07:00
hysdn_if.h
i2c-algo-bit.h
i2c-algo-ite.h
i2c-algo-pca.h
i2c-algo-pcf.h
i2c-algo-sgi.h
i2c-algo-sibyte.h
i2c-dev.h
i2c-id.h	[PATCH] I2C: Kill i2c_algorithm.id (7/7)	2005-09-05 09:14:33 -07:00
i2c-isa.h	[PATCH] I2C: Kill i2c_algorithm.id (5/7)	2005-09-05 09:14:31 -07:00
i2c-pxa.h
i2c.h	[PATCH] I2C: Drop the I2C_ACK_TEST ioctl	2005-09-05 09:26:56 -07:00
i2o-dev.h
i2o.h
i8k.h
ibmtr.h
icmp.h
icmpv6.h
ide.h
idr.h
if_arcnet.h
if_arp.h
if_bonding.h
if_bridge.h
if_cablemodem.h
if_ec.h
if_eql.h
if_ether.h
if_fc.h
if_fddi.h
if_frad.h
if_hippi.h
if_infiniband.h
if_ltalk.h
if_packet.h
if_plip.h
if_ppp.h
if_pppox.h
if_shaper.h
if_slip.h
if_strip.h
if_tr.h
if_tun.h
if_tunnel.h
if_vlan.h
if_wanpipe_common.h
if_wanpipe.h
if.h
igmp.h
in6.h	[IPV6]: Support IPV6_{RECV,}TCLASS socket options / ancillary data.	2005-09-08 10:19:03 +09:00
in_route.h
in.h
inet_diag.h
inet.h
inetdevice.h
init_task.h	[PATCH] files: files struct with RCU	2005-09-09 13:57:55 -07:00
init.h
initrd.h
inotify.h	[PATCH] inotify: fix event loss on hardlinked files	2005-09-07 16:57:39 -07:00
input.h	[PATCH] Input: Add a new switch event type	2005-09-07 16:57:54 -07:00
interrupt.h
ioc4.h
ioctl32.h	[PATCH] remove register_ioctl32_conversion and unregister_ioctl32_conversion	2005-09-07 16:57:20 -07:00
ioctl.h
ioport.h
ioprio.h
ip6_tunnel.h
ip_mp_alg.h
ip.h
ipc.h
ipmi_msgdefs.h
ipmi_smi.h
ipmi.h	[PATCH] ipmi: fix panic ipmb response	2005-09-07 16:57:48 -07:00
ipsec.h
ipv6_route.h
ipv6.h	[IPV6]: Support IPV6_{RECV,}TCLASS socket options / ancillary data.	2005-09-08 10:19:03 +09:00
ipx.h
irda.h
irq_cpustat.h
irq.h	[PATCH] CHECK_IRQ_PER_CPU() to avoid dead code in __do_IRQ()	2005-09-07 16:57:29 -07:00
isapnp.h
isdn_divertif.h
isdn_ppp.h
isdn.h	[PATCH] Remove non-arch consumers of asm/segment.h	2005-09-07 16:57:34 -07:00
isdnif.h
isicom.h
iso_fs.h
istallion.h
ite_gpio.h
ixjuser.h
jbd.h	[PATCH] spinlock consolidation	2005-09-10 10:06:21 -07:00
jffs2_fs_i.h
jffs2_fs_sb.h
jffs2.h
jffs.h
jhash.h
jiffies.h
journal-head.h
joystick.h
kallsyms.h
kbd_diacr.h
kbd_kern.h
kd.h
kdev_t.h
kernel_stat.h
kernel.h
kernelcapi.h
kexec.h
key-ui.h
key.h
keyboard.h
keyctl.h
kfifo.h
klist.h	[PATCH] fix klist semantics for lists which have elements removed on traversal	2005-09-07 18:26:54 -07:00
kmalloc_sizes.h
kmod.h
kobj_map.h
kobject_uevent.h
kobject.h
kprobes.h	[PATCH] Kprobes: prevent possible race conditions generic	2005-09-07 16:57:59 -07:00
kref.h
kthread.h
lapb.h
lcd.h
libata.h	/spare/repo/libata-dev branch 'iomap-try3'	2005-09-05 05:20:33 -04:00
libps2.h
limits.h
linkage.h	[PATCH] Kprobes: prevent possible race conditions generic	2005-09-07 16:57:59 -07:00
linux_logo.h
list.h
llc.h
loop.h
lp.h
major.h
matroxfb.h
mbcache.h
mc6821.h
mc146818rtc.h
mca-legacy.h
mca.h
mempolicy.h	[PATCH] PCI: Run PCI driver initialization on local node	2005-09-08 14:57:23 -07:00
mempool.h
meye.h
mii.h
minix_fs.h
miscdevice.h
mm_inline.h
mm.h
mman.h
mmtimer.h
mmzone.h
mod_devicetable.h
module.h
moduleloader.h
moduleparam.h
mount.h
mpage.h
mqueue.h
mroute.h
msdos_fs.h
msg.h	[PATCH] ipc: convert /proc/sysvipc/* to generic seq_file interface	2005-09-07 16:57:26 -07:00
mtio.h
mv643xx.h	[PATCH] ppc32: mv64x60 updates & enhancements	2005-09-05 00:06:00 -07:00
n_r3964.h
namei.h
namespace.h
nbd.h
ncp_fs_i.h
ncp_fs_sb.h
ncp_fs.h
ncp_mount.h
ncp_no.h
ncp.h
net.h
netdevice.h
netfilter_arp.h
netfilter_bridge.h
netfilter_decnet.h
netfilter_ipv4.h
netfilter_ipv6.h
netfilter_logging.h
netfilter.h
netlink.h
netpoll.h
netrom.h
nfs2.h
nfs3.h
nfs4_acl.h
nfs4_mount.h
nfs4.h
nfs_fs_i.h
nfs_fs_sb.h
nfs_fs.h
nfs_idmap.h
nfs_mount.h
nfs_page.h
nfs_xdr.h
nfs.h
nfsacl.h
nfsd_idmap.h
nls.h
nmi.h
node.h
nodemask.h
notifier.h
nubus.h
numa.h
nvram.h
openprom_fs.h
oprofile.h
page-flags.h	[PATCH] VM: add page_state info to per-node meminfo	2005-09-05 00:05:49 -07:00
pagemap.h
pagevec.h
param.h
parport_pc.h
parport.h
parser.h
patchkey.h
pci_ids.h	[PATCH] net: add driver for the NIC on Cell Blades	2005-09-06 22:17:49 -04:00
pci_regs.h	[PATCH] PCI: restore BAR values after D3hot->D0 for devices that need it	2005-09-08 14:57:24 -07:00
pci-acpi.h
pci.h	Remove "must_check" attributes in PCI-land	2005-09-09 15:43:46 -07:00
pcieport_if.h
percpu_counter.h
percpu.h
personality.h
pfkeyv2.h
pg.h
phonedev.h
phy.h
pid.h
pipe_fs_i.h	[PATCH] remove pipe definitions	2005-09-07 16:57:24 -07:00
pkt_cls.h
pkt_sched.h
pktcdvd.h
platform.h
pm.h	[PATCH] swsusp: switch pm_message_t to struct	2005-09-05 00:06:16 -07:00
pmu.h
pnp.h	[PATCH] PNP: make pnp_dbg conditional directly on CONFIG_PNP_DEBUG	2005-09-07 16:57:25 -07:00
pnpbios.h
poll.h
posix_acl_xattr.h
posix_acl.h
posix_types.h
posix-timers.h
ppdev.h
ppp_channel.h
ppp_defs.h
ppp-comp.h
prctl.h
preempt.h
prefetch.h
prio_tree.h
proc_fs.h
profile.h
ps2esdi.h
ptrace.h	[PATCH] remove duplicated code from proc and ptrace	2005-09-07 16:57:43 -07:00
qic117.h
qnx4_fs.h
qnxtypes.h
quota.h
quotaio_v1.h
quotaio_v2.h
quotaops.h
radeonfb.h
radix-tree.h
raid_class.h
ramfs.h
random.h
raw.h
rbtree.h
rcupdate.h	[PATCH] files: fix rcu initializers	2005-09-09 13:57:54 -07:00
rcuref.h	[PATCH] files: rcuref APIs	2005-09-09 13:57:54 -07:00
reboot_fixups.h
reboot.h
reiserfs_acl.h
reiserfs_fs_i.h
reiserfs_fs_sb.h
reiserfs_fs.h
reiserfs_xattr.h
relayfs_fs.h	[PATCH] relayfs	2005-09-07 16:57:18 -07:00
resource.h
rmap.h
romfs_fs.h
root_dev.h
rose.h
route.h
rslib.h
rtc.h
rtnetlink.h
rwsem-spinlock.h
rwsem.h
sc26198.h
scatterlist.h
scc.h
sched.h	[PATCH] cpuset semaphore depth check deadlock fix	2005-09-10 10:06:21 -07:00
sctp.h
scx200_gpio.h
scx200.h
sdla_asy.h
sdla_chdlc.h
sdla_fr.h
sdla_ppp.h
sdla_x25.h
sdla.h
sdladrv.h
sdlapci.h
sdlasfm.h
seccomp.h
securebits.h
security.h	[PATCH] remove the inode_post_link and inode_post_rename LSM hooks	2005-09-09 13:57:28 -07:00
selection.h
selinux_netlink.h
sem.h	[PATCH] ipc: convert /proc/sysvipc/* to generic seq_file interface	2005-09-07 16:57:26 -07:00
seq_file.h
seqlock.h
serial167.h
serial_8250.h	[SERIAL] Use an enum for serial8250 platform device IDs	2005-09-08 16:04:41 +01:00
serial_core.h	Fix up ARM serial driver compile failure	2005-09-08 14:47:12 -07:00
serial_reg.h
serial.h
serialP.h
serio.h
shm.h
shmem_fs.h
signal.h
skbuff.h	[NET]: Optimize pskb_trim_rcsum()	2005-09-08 12:32:03 -07:00
slab.h	[PATCH] introduce and use kzalloc	2005-09-07 16:57:45 -07:00
smb_fs_i.h
smb_fs_sb.h
smb_fs.h
smb_mount.h
smb.h
smbno.h
smp_lock.h
smp.h
snmp.h
socket.h
sockios.h
som.h
sonet.h
sonypi.h	[PATCH] sonypi SPIC initialisation fix	2005-09-07 16:57:24 -07:00
sort.h
sound.h
soundcard.h
spinlock_api_smp.h	[PATCH] spinlock consolidation	2005-09-10 10:06:21 -07:00
spinlock_api_up.h	[PATCH] spinlock consolidation	2005-09-10 10:06:21 -07:00
spinlock_types_up.h	[PATCH] spinlock consolidation	2005-09-10 10:06:21 -07:00
spinlock_types.h	[PATCH] spinlock consolidation	2005-09-10 10:06:21 -07:00
spinlock_up.h	[PATCH] spinlock consolidation	2005-09-10 10:06:21 -07:00
spinlock.h	[PATCH] spinlock consolidation	2005-09-10 10:06:21 -07:00
stallion.h
stat.h
statfs.h
stddef.h
stop_machine.h
string.h
stringify.h
superhyway.h
suspend.h
swap.h
swapops.h	[PATCH] comment typo fix	2005-09-05 00:05:45 -07:00
synclink.h
sys.h
syscalls.h
sysctl.h	[PATCH] ipmi poweroff: fix chassis control	2005-09-07 16:57:49 -07:00
sysdev.h
sysfs.h
sysrq.h
sysv_fs.h
tcp.h
telephony.h
termios.h
textsearch_fsm.h
textsearch.h
thread_info.h
threads.h
ticable.h
time.h	[PATCH] Delete unused do_nanosleep declaration	2005-09-07 16:57:30 -07:00
timer.h	[PATCH] timer initialization cleanup: DEFINE_TIMER	2005-09-09 14:03:48 -07:00
times.h
timex.h	[PATCH] NTP: ntp-helper functions	2005-09-07 16:57:34 -07:00
tiocl.h
topology.h	[PATCH] cpusets: Move the ia64 domain setup code to the generic code	2005-09-07 16:57:40 -07:00
toshiba.h
transport_class.h
trdevice.h
tty_driver.h
tty_flip.h
tty_ldisc.h
tty.h	[PATCH] vesafb: Add blanking support	2005-09-09 13:57:58 -07:00
types.h
udf_fs_i.h
udf_fs_sb.h
udf_fs.h
udp.h
ufs_fs_i.h
ufs_fs_sb.h
ufs_fs.h
uinput.h
uio.h
ultrasound.h
umem.h
un.h
unistd.h
usb_cdc.h
usb_ch9.h
usb_gadget.h
usb_gadgetfs.h
usb_input.h
usb_isp116x.h	[PATCH] USB: isp116x-hcd: remove clock() and reset()	2005-09-08 16:22:48 -07:00
usb_otg.h
usb_sl811.h
usb.h	[PATCH] USB: URB_ASYNC_UNLINK flag removed from the kernel	2005-09-08 16:23:04 -07:00
usbdevice_fs.h
user.h
utime.h
uts.h
utsname.h
vermagic.h
vfs.h
via.h
video_decoder.h
video_encoder.h
videodev2.h	[PATCH] v4l: common part Updates and tuner additions	2005-09-09 13:57:49 -07:00
videodev.h	[PATCH] v4l: common part Updates and tuner additions	2005-09-09 13:57:49 -07:00
videotext.h
vmalloc.h	[PATCH] arm: allow for arch-specific IOREMAP_MAX_ORDER	2005-09-05 00:05:46 -07:00
vt_buffer.h
vt_kern.h
vt.h
wait.h
wanpipe.h
wanrouter.h
watchdog.h
wavefront.h
wireless.h	[wireless] build fixes after merging WE-19	2005-09-07 00:27:54 -04:00
workqueue.h
writeback.h
x25.h
xattr.h
xfrm.h
yam.h
zconf.h
zftape.h
zlib.h
zorro_ids.h
zorro.h
zutil.h