It has been thought that the per-user file descriptors limit would also
limit the resources that a normal user can request via the epoll
interface. Vegard Nossum reported a very simple program (a modified
version attached) that can make a normal user to request a pretty large
amount of kernel memory, well within the its maximum number of fds. To
solve such problem, default limits are now imposed, and /proc based
configuration has been introduced. A new directory has been created,
named /proc/sys/fs/epoll/ and inside there, there are two configuration
points:
max_user_instances = Maximum number of devices - per user
max_user_watches = Maximum number of "watched" fds - per user
The current default for "max_user_watches" limits the memory used by epoll
to store "watches", to 1/32 of the amount of the low RAM. As example, a
256MB 32bit machine, will have "max_user_watches" set to roughly 90000.
That should be enough to not break existing heavy epoll users. The
default value for "max_user_instances" is set to 128, that should be
enough too.
This also changes the userspace, because a new error code can now come out
from EPOLL_CTL_ADD (-ENOSPC). The EMFILE from epoll_create() was already
listed, so that should be ok.
[akpm@linux-foundation.org: use get_current_user()]
Signed-off-by: Davide Libenzi <davidel@xmailserver.org>
Cc: Michael Kerrisk <mtk.manpages@gmail.com>
Cc: <stable@kernel.org>
Cc: Cyrill Gorcunov <gorcunov@gmail.com>
Reported-by: Vegard Nossum <vegardno@ifi.uio.no>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
According to the manual the "tdfOnExit" flag must be set on the last byte
we want to send. The PSC controller holds SS low until the flag is set.
However, the flag was set always on the last byte of the FIFO,
independently if it is the last byte of the transfer. This generates
spurious toggling of the SS signals that breaks the protocol of some
peripherals. Fix.
Signed-off-by: Stefano Babic <sbabic@denx.de>
Acked-by: Grant Likely <grant.likely@secretlab.ca>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Use __initdata for data, not __init.
Signed-off-by: Roel Kluin <roel.kluin@gmail.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
make use of the new dmi device loading support to automatically load the
applesmc driver based on the dmi_match table.
Signed-off-by: Henrik Rydberg <rydberg@euromail.se>
Cc: Nicolas Boichat <nicolas@boichat.ch>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The netmos_9xx5_combo type assumes that PCI SSID provides always the
correct value for the number of parallel and serial ports, but there are
indeed broken devices with wrong numbers, which may result in Oops.
This patch simply adds the check of the array range.
Reference: Novell bnc#447067
https://bugzilla.novell.com/show_bug.cgi?id=447067
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Fixes for memcg/memory hotplug.
While memory hotplug allocate/free memmap, page_cgroup doesn't free
page_cgroup at OFFLINE when page_cgroup is allocated via bootomem.
(Because freeing bootmem requires special care.)
Then, if page_cgroup is allocated by bootmem and memmap is freed/allocated
by memory hotplug, page_cgroup->page == page is no longer true.
But current MEM_ONLINE handler doesn't check it and update
page_cgroup->page if it's not necessary to allocate page_cgroup. (This
was not found because memmap is not freed if SPARSEMEM_VMEMMAP is y.)
And I noticed that MEM_ONLINE can be called against "part of section".
So, freeing page_cgroup at CANCEL_ONLINE will cause trouble. (freeing
used page_cgroup) Don't rollback at CANCEL.
One more, current memory hotplug notifier is stopped by slub because it
sets NOTIFY_STOP_MASK to return vaule. So, page_cgroup's callback never
be called. (low priority than slub now.)
I think this slub's behavior is not intentional(BUG). and fixes it.
Another way to be considered about page_cgroup allocation:
- free page_cgroup at OFFLINE even if it's from bootmem
and remove specieal handler. But it requires more changes.
Addresses http://bugzilla.kernel.org/show_bug.cgi?id=12041
Signed-off-by: KAMEZAWA Hiruyoki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: Balbir Singh <balbir@in.ibm.com>
Cc: Pavel Emelyanov <xemul@openvz.org>
Tested-by: Badari Pulavarty <pbadari@us.ibm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Jim Radford has reported that the vmap subsystem rewrite was sometimes
causing his VIVT ARM system to behave strangely (seemed like going into
infinite loops trying to fault in pages to userspace).
We determined that the problem was most likely due to a cache aliasing
issue. flush_cache_vunmap was only being called at the moment the page
tables were to be taken down, however with lazy unmapping, this can happen
after the page has subsequently been freed and allocated for something
else. The dangling alias may still have dirty data attached to it.
The fix for this problem is to do the cache flushing when the caller has
called vunmap -- it would be a bug for them to write anything else to the
mapping at that point.
That appeared to solve Jim's problems.
Reported-by: Jim Radford <radford@blackbean.org>
Signed-off-by: Nick Piggin <npiggin@suse.de>
Cc: Russell King <rmk@arm.linux.org.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mfasheh/ocfs2:
ocfs2: fix regression in ocfs2_read_blocks_sync()
ocfs2: fix return value set in init_dlmfs_fs()
ocfs2: Small documentation update
ocfs2: fix wake_up in unlock_ast
ocfs2: initialize stack_user lvbptr
ocfs2: comments typo fix
We're panicing in ocfs2_read_blocks_sync() if a jbd-managed buffer is seen.
At first glance, this seems ok but in reality it can happen. My test case
was to just run 'exorcist'. A struct inode is being pushed out of memory but
is then re-read at a later time, before the buffer has been checkpointed by
jbd. This causes a BUG to be hit in ocfs2_read_blocks_sync().
Reviewed-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
In init_dlmfs_fs(), if calling kmem_cache_create() failed, the code will use return value from
calling bdi_init(). The correct behavior should be set status as -ENOMEM before going to "bail:".
Signed-off-by: Coly Li <coyli@suse.de>
Acked-by: Sunil Mushran <sunil.mushran@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
In ocfs2_unlock_ast(), call wake_up() on lockres before releasing
the spin lock on it. As soon as the spin lock is released, the
lockres can be freed.
Signed-off-by: David Teigland <teigland@redhat.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
The locking_state dump, ocfs2_dlm_seq_show, reads the lvb on locks where it
has not yet been initialized by a lock call.
Signed-off-by: David Teigland <teigland@redhat.com>
Acked-by: Joel Becker <joel.becker@oracle.com>
Signed-off-by: Mark Fasheh <mfasheh@suse.com>
* 'upstream-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jgarzik/libata-dev:
libata: blacklist Seagate drives which time out FLUSH_CACHE when used with NCQ
[libata] pata_rb532_cf: fix signature of the xfer function
[libata] pata_rb532_cf: fix and rename register definitions
ata_piix: add borked Tecra M4 to broken suspend list
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband:
IB/mlx4: Fix MTT leakage in resize CQ
IB/ehca: Fix problem with generated flush work completions
IB/ehca: Change misleading error message on memory hotplug
mlx4_core: Save/restore default port IB capability mask
Some recent Seagate harddrives have firmware bug which causes FLUSH
CACHE to timeout under certain circumstances if NCQ is being used.
This can be worked around by disabling NCQ and fixed by updating the
firmware. Implement ATA_HORKAGE_FIRMWARE_UPDATE and blacklist these
devices.
The wiki page has been updated to contain information on this issue.
http://ata.wiki.kernel.org/index.php/Known_issues
Signed-off-by: Tejun Heo <tj@kernel.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Per definition, this function should return the number of bytes
consumed. As the original parameter "buflen" is being decremented inside
the read/write loop, save it in "retlen" at the beginning.
Signed-off-by: Phil Sutter <n0-1@freewrt.org>
Acked-by: Sergei Shtyltov <sshtylyov@ru.mvista.com>
Acked-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
Acked-by: Florian Fainelli <florian@openwrt.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
The original standalone driver uses a custom address for the error
register. Use it in pata_rb532_cf, too.
Rename two register definitions:
- The address offset 0x0800 in fact is the ATA base, not ATA command
address.
- The offset 0x0C00 is not a regular ATA data address, but a buffered one
allowing 4-byte IO.
Signed-off-by: Phil Sutter <n0-1@freewrt.org>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
Tecra M4 sometimes forget what it is and reports bogus data via DMI
which makes the machine evade broken suspend matching and thus fail
suspend/resume. This patch updates piix_broken_suspend() such that it
can match such case. As the borked DMI data is a bit generic,
matching many entries to make the match more specific is necessary.
As the usual DMI matching is limited to four entries, this patch uses
hard coded manual matching.
This is reported by Alexandru Romanescu.
Signed-off-by: Tejun Heo <tj@kernel.org>
Cc: Alexandru Romanescu <a_romanescu@yahoo.co.uk>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>
When resizing a CQ, MTTs associated with the old CQE buffer were not
freed. As a result, if any app used resize CQ repeatedly, all MTTs
were eventually exhausted, which led to all memory registration
operations failing until the driver is reloaded.
Once the RESIZE_CQ command returns successfully from FW, FW no longer
accesses the old CQ buffer, so it is safe to deallocate the MTT
entries used by the old CQ buffer.
Finally, if the RESIZE_CQ command fails, the MTTs allocated for the
new CQEs buffer also need to be de-allocated.
This fixes <https://bugs.openfabrics.org/show_bug.cgi?id=1416>.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
This fix enables ehca device driver to generate flush work completions
even if the application doesn't request completions for all work
requests. The current implementation of ehca will generate flush work
completions for the wrong work requests if an application uses non
signaled work completions.
Signed-off-by: Stefan Roscher <stefan.roscher@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
The error message printed when the eHCA driver prevents memory hotplug
is misleading -- the user might think that hot-removing the lhca,
hotplugging memory, then hot-adding the lhca again will work, but it
actually doesn't.
Signed-off-by: Joachim Fenkes <fenkes@de.ibm.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394-2.6:
ieee1394: sbp2: fix race condition in state change
ieee1394: fix list corruption (reported at module removal)
firewire: fw-sbp2: another iPod mini quirk entry
ieee1394: sbp2: another iPod mini quirk entry
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/hid:
HID: Apple ALU wireless keyboards are bluetooth devices
HID: remove setup mutex, fix possible deadlock
HID: add USB ID for another dual gameron adapter
HID: unignore mouse on unibody macbooks
HID: fix blacklist entries for greenasia/pantherlord
We alloc a fake tty in usb serial console setup function. we should
init the tty's kref otherwise we will face WARN_ON after following
invoke of tty_port_tty_set --> tty_kref_get.
Signed-off-by: Kevin Hao <kexin.hao@windriver.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
When no console is configured tty_open tries to call kref_get on a NULL
pointer, return ENODEV instead.
Signed-off-by: Will Newton <will.newton@gmail.com>
Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
* git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb-2.6:
USB: serial: add more Onda device ids to option driver
USB: usb-storage: unusual_devs entry for Nikon D2H
USB: storage: unusual_devs entry for Mio C520-GPS
USB: fsl_usb2_udc: Report disconnect before unbinding
USB: fsl_qe_udc: Report disconnect before unbinding
USB: fix SB600 USB subsystem hang bug
Revert "USB: improve ehci_watchdog's side effect in CPU power management"
The zone's rotation statistics must not be accessed without the
corresponding LRU lock held. Fix an unprotected write in
shrink_active_list().
Acked-by: Rik van Riel <riel@redhat.com>
Reviewed-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Signed-off-by: Johannes Weiner <hannes@saeurebad.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This reverts commit e669dae614, since it
is incomplete, and clashes with fuller patches and the sparc 32/64
unification effort.
Requested-by: David Miller <davem@davemloft.net>
Acked-by: Al Viro <viro@ZenIV.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This patch adds an unusual_devs entry for the Nikon D2H camera.
From: Tobias Kunze Briseño <t@fictive.com>,
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as1176) adds an unusual_devs entry for the Mio C520 GPS
unit. Other devices also based on the Mitac hardware use the same USB
interface firmware, so the Vendor and Product names are generalized.
This fixes Bugzilla #11583.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Tested-by: Tamas Kerecsen <kerecsen@bigfoot.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Gadgets disable endpoints in their disconnect callbacks, so
we must call disconnect before unbinding. This also fixes
muram memory leak, since we free muram in the qe_ep_disable().
But mainly the patch fixes following badness:
root@b1:~# insmod fsl_qe_udc.ko
fsl_qe_udc: Freescale QE/CPM USB Device Controller driver, 1.0
fsl_qe_udc e01006c0.usb: QE USB controller initialized as device
root@b1:~# insmod g_ether.ko
g_ether gadget: using random self ethernet address
g_ether gadget: using random host ethernet address
usb0: MAC be:2d:3c:fa:be:f0
usb0: HOST MAC 62:b8:6a:df:38:66
g_ether gadget: Ethernet Gadget, version: Memorial Day 2008
g_ether gadget: g_ether ready
fsl_qe_udc e01006c0.usb: fsl_qe_udc bind to driver g_ether
g_ether gadget: high speed config #1: CDC Ethernet (ECM)
root@b1:~# rmmod g_ether.ko
------------[ cut here ]------------
Badness at drivers/usb/gadget/composite.c:871
[...]
NIP [d10c1374] composite_unbind+0x24/0x15c [g_ether]
LR [d10a82f4] usb_gadget_unregister_driver+0x128/0x168 [fsl_qe_udc]
Call Trace:
[cfb93e80] [cfb1f3a0] 0xcfb1f3a0 (unreliable)
[cfb93eb0] [d10a82f4] usb_gadget_unregister_driver+0x128/0x168 [fsl_qe_udc]
[cfb93ed0] [d10c2a3c] usb_composite_unregister+0x3c/0x4c [g_ether]
[cfb93ee0] [c006bde0] sys_delete_module+0x130/0x19c
[cfb93f40] [c00142d8] ret_from_syscall+0x0/0x38
[...]
fsl_qe_udc e01006c0.usb: unregistered gadget driver 'g_ether'
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com>
Acked-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch is required for all AMD SB600 revisions to avoid USB subsystem hang
symptom. The USB subsystem hang symptom is observed when the system has
multiple USB devices connected to it. In some cases a USB hub may be required
to observe this symptom.
Reported in bugzilla as #11599, the similar patch for SB700 old revision is:
commit b09bc6cbae
Reported-by: raffaele <ralfconn@tele2.it>
Tested-by: Roman Mamedov <roman@rm.pp.ru>
Signed-off-by: Shane Huang <shane.huang@amd.com>
Cc: stable <stable@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This reverts commit f0d781d59c.
It was the wrong thing to do, and does not really do what it said
it did.
Cc: Yi Yang <yi.y.yang@intel.com>
Cc: David Brownell <dbrownell@users.sourceforge.net>
Cc: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
attr_smt_snooze_delay is only defined for CONFIG_PPC64, so protect the
attribute removal with the same condition. This fixes this build error
on 32-bit SMP configurations:
/data/home/miltonm/next.git/arch/powerpc/kernel/sysfs.c: In function ‘unregister_cpu_online’:
/data/home/miltonm/next.git/arch/powerpc/kernel/sysfs.c:722: error: ‘attr_smt_snooze_delay’ undeclared (first use in this function)
/data/home/miltonm/next.git/arch/powerpc/kernel/sysfs.c:722: error: (Each undeclared identifier is reported only once
/data/home/miltonm/next.git/arch/powerpc/kernel/sysfs.c:722: error: for each function it appears in.)
Signed-off-by: Paul Mackerras <paulus@samba.org>
* 'drm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/airlied/drm-2.6:
drm/i915: Fix copy'n'pasteo that broke VT switch if flushing was non-empty.
* 'merge' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc:
powerpc: Fix system calls on Cell entered with XER.SO=1
powerpc/cell: Fix GDB watchpoints, again
powerpc/mpic: Don't reset affinity for secondary MPIC on boot
powerpc/cell/axon-msi: Retry on missing interrupt
powerpc: Fix boot freeze on machine with empty memory node
powerpc: Fix IRQ assignment for some PCIe devices
powerpc/spufs: Fix spinning in spufs_ps_fault on signal
powerpc/mpc832x_rdb: fix swapped ethernet ids
powerpc: Use generic PHY driver for Marvell 88E1111 PHY on GE Fanuc SBC610
powerpc/85xx: L2 cache size wrong in 8572DS dts
powerpc/virtex: Update defconfigs
powerpc/52xx: update defconfigs
xsysace: Fix driver to use resource_size_t instead of unsigned long
powerpc/virtex: fix various format/casting printk mismatches
powerpc/mpc5200: fix bestcomm Kconfig dependencies
powerpc/44x: Fix 460EX/460GT machine check handling
powerpc/40x: Limit allocable DRAM during early mapping
* master.kernel.org:/home/rmk/linux-2.6-arm:
Allow architectures to override copy_user_highpage()
[ARM] pxa/palmtx: misc fixes to use generic GPIO API
ARM: OMAP: Fixes for suspend / resume GPIO wake-up handling
[ARM] pxa/corgi: update default config to exclude tosa from being built
[ARM] pxa/pcm990: use negative number for an invalid GPIO in camera data
ARM: OMAP: Typo fix for clock_allow_idle
ARM: OMAP: Remove broken LCD driver for SX1
[ARM] 5335/1: pxa25x_udc: Fix is_vbus_present to return 1 or 0
[ARM] pxa/MioA701: bluetooth resume fix
[ARM] pxa/MioA701: fix memory corruption.
It turns out that on Cell, on a kernel with CONFIG_VIRT_CPU_ACCOUNTING
= y, if a program sets the SO (summary overflow) bit in the XER and
then does a system call, the SO bit in CR0 will be set on return
regardless of whether the system call detected an error. Since CR0.SO
is used as the error indication from the system call, this means that
all system calls appear to fail.
The reason is that the workaround for the timebase bug on Cell uses a
compare instruction. With CONFIG_VIRT_CPU_ACCOUNTING = y, the
ACCOUNT_CPU_USER_ENTRY macro reads the timebase, so we end up doing a
compare instruction, which copies XER.SO to CR0.SO. Since we were
doing this in the system call entry patch after clearing CR0.SO but
before saving the CR, this meant that the saved CR image had CR0.SO
set if XER.SO was set on entry.
This fixes it by moving the clearing of CR0.SO to after the
ACCOUNT_CPU_USER_ENTRY call in the system call entry path.
Signed-off-by: Paul Mackerras <paulus@samba.org>
Acked-by: Arnd Bergmann <arnd@arndb.de>
Acked-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
An earlier patch from Jens Osterkamp attempted to fix GDB
watchpoints by enabling the DABRX register at boot time.
Unfortunately, this did not work on SMP setups, where
secondary CPUs were still using the power-on DABRX value.
This introduces the same change for secondary CPUs on cell
as well.
Reported-by: Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
Tested-by: Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
Kexec/kdump currently fails on the IBM QS2x blades when the kexec happens
on a CPU other than the initial boot CPU. It turns out that this is the
result of mpic_init trying to set affinity of each interrupt vector to the
current boot CPU.
As far as I can tell, the same problem is likely to exist on any
secondary MPIC, because they have to deliver interrupts to the first
output all the time. There are two potential solutions for this: either
not set up affinity at all for secondary MPICs, or assume that a single
CPU output is connected to the upstream interrupt controller and hardcode
affinity to that per architecture.
This patch implements the second approach, defaulting to the first output.
Currently, all known secondary MPICs are routed to their upstream port
using the first destination, so we hardcode that.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
The MSI capture logic for the axon bridge can sometimes
lose interrupts in case of high DMA and interrupt load,
when it signals an MSI interrupt to the MPIC interrupt
controller while we are already handling another MSI.
Each MSI vector gets written into a FIFO buffer in main
memory using DMA, and that DMA access is normally flushed
by the actual interrupt packet on the IOIF. An MMIO
register in the MSIC holds the position of the last
entry in the FIFO buffer that was written. However,
reading that position does not flush the DMA, so that
we can observe stale data in the buffer.
In a stress test, we have observed the DMA to arrive
up to 14 microseconds after reading the register.
This patch works around this problem by retrying the
access to the FIFO buffer.
We can reliably detect the conditioning by writing
an invalid MSI vector into the FIFO buffer after
reading from it, assuming that all MSIs we get
are valid. After detecting an invalid MSI vector,
we udelay(1) in the interrupt cascade for up to
100 times before giving up.
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Paul Mackerras <paulus@samba.org>
I got a bug report about a distro kernel not booting on a particular
machine. It would freeze during boot:
> ...
> Could not find start_pfn for node 1
> [boot]0015 Setup Done
> Built 2 zonelists in Node order, mobility grouping on. Total pages: 123783
> Policy zone: DMA
> Kernel command line:
> [boot]0020 XICS Init
> [boot]0021 XICS Done
> PID hash table entries: 4096 (order: 12, 32768 bytes)
> clocksource: timebase mult[7d0000] shift[22] registered
> Console: colour dummy device 80x25
> console handover: boot [udbg0] -> real [hvc0]
> Dentry cache hash table entries: 1048576 (order: 7, 8388608 bytes)
> Inode-cache hash table entries: 524288 (order: 6, 4194304 bytes)
> freeing bootmem node 0
I've reproduced this on 2.6.27.7. It is caused by commit
8f64e1f2d1 ("powerpc: Reserve in bootmem
lmb reserved regions that cross NUMA nodes").
The problem is that Jon took a loop which was (in pseudocode):
for_each_node(nid)
NODE_DATA(nid) = careful_alloc(nid);
setup_bootmem(nid);
reserve_node_bootmem(nid);
and broke it up into:
for_each_node(nid)
NODE_DATA(nid) = careful_alloc(nid);
setup_bootmem(nid);
for_each_node(nid)
reserve_node_bootmem(nid);
The issue comes in when the 'careful_alloc()' is called on a node with
no memory. It falls back to using bootmem from a previously-initialized
node. But, bootmem has not yet been reserved when Jon's patch is
applied. It gives back bogus memory (0xc000000000000000) and pukes
later in boot.
The following patch collapses the loop back together. It also breaks
the mark_reserved_regions_for_nid() code out into a function and adds
some comments. I think a huge part of introducing this bug is because
for loop was too long and hard to read.
The actual bug fix here is the:
+ if (end_pfn <= node->node_start_pfn ||
+ start_pfn >= node_end_pfn)
+ continue;
Signed-off-by: Dave Hansen <dave@linux.vnet.ibm.com>
Signed-off-by: Paul Mackerras <paulus@samba.org>
