Changeset 3869c4aa18835c8c61b44bd0f3ace36e9d3b5bd0
that went in after 2.6.30-rc1 was a seemingly small change to _set_memory_wc()
to make it complaint with SDM requirements. But, introduced a nasty bug, which
can result in crash and/or strange corruptions when set_memory_wc is used.
One such crash reported here
http://lkml.org/lkml/2009/7/30/94
Actually, that changeset introduced two bugs.
* change_page_attr_set() takes &addr as first argument and can the addr value
might have changed on return, even for single page change_page_attr_set()
call. That will make the second change_page_attr_set() in this routine
operate on unrelated addr, that can eventually cause strange corruptions
and bad page state crash.
* The second change_page_attr_set() call, before setting _PAGE_CACHE_WC, should
clear the earlier _PAGE_CACHE_UC_MINUS, as otherwise cache attribute will not
be WC (will be UC instead).
The patch below fixes both these problems. Sending a single patch to fix both
the problems, as the change is to the same line of code. The change to have a
addr_copy is not very clean. But, it is simpler than making more changes
through various routines in pageattr.c.
A huge thanks to Jerome for reporting this problem and providing a simple test
case that helped us root cause the problem.
Reported-by: Jerome Glisse <glisse@freedesktop.org> Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi@intel.com> Signed-off-by: Suresh Siddha <suresh.b.siddha@intel.com>
LKML-Reference: <20090730214319.GA1889@linux-os.sc.intel.com> Acked-by: Dave Airlie <airlied@redhat.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
11static inline unsigned long native_save_fl(void)
12{
13 unsigned long flags;
14
15 asm volatile("# __raw_save_flags\n\t"
16 "pushf ; pop %0"
17 : "=g" (flags)
18 : /* no input */
19 : "memory");
20
21 return flags;
22}
If gcc chooses to put flags on the stack, for instance because this is
inlined into a larger function with more register pressure, the offset
of the flags variable from the stack pointer will change when the
pushf is performed. gcc doesn't attempt to understand that fact, and
address used for pop will still be the same. It will write to
somewhere near flags on the stack but not actually into it and
overwrite some other value.
I saw this happen in the ide_device_add_all function when running in a
simulator I work on. I'm assuming that some quirk of how the simulated
hardware is set up caused the code path this is on to be executed when
it normally wouldn't.
A simple fix might be to change "=g" to "=r".
Reported-by: Gabe Black <spamforgabe@umich.edu> Signed-off-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The code was incorrectly reserving memtypes using the page
virtual address instead of the physical address. Furthermore,
the code was not ignoring highmem pages as it ought to.
( upstream does not pass in highmem pages yet - but upcoming
graphics code will do it and there's no reason to not handle
this properly in the CPA APIs.)
Narayanan reports "The regression is around 15%. There is no disk controller
as our setup is based on Samsung OneNAND used as a memory mapped device on a
OMAP2430 based board."
The page allocator tries to preserve contiguous PFN ordering when returning
pages such that repeated callers to the allocator have a strong chance of
getting physically contiguous pages, particularly when external fragmentation
is low. However, of the bulk of the allocations have __GFP_COLD set as they
are due to aio_read() for example, then the PFNs are in reverse PFN order.
This can cause performance degration when used with IO controllers that could
have merged the requests.
This patch attempts to preserve the contiguous ordering of PFNs for users of
__GFP_COLD.
This is because hugetlb_unreserve_pages() is unconditionally removing
blocks_per_huge_page(h) on each call rather than using the freed amount.
If there were 0 blocks, it goes negative, resulting in the above.
usb0 and usb1 mux settings in the sicrl register were swapped (twice!)
in mpc834x_usb_cfg(), leading to various strange issues with fsl-ehci
and full speed devices.
The USB port config on mpc834x is done using 2 muxes: Port 0 is always
used for MPH port 0, and port 1 can either be used for MPH port 1 or DR
(unless DR uses UTMI phy or OTG, then it uses both ports) - See 8349 RM
figure 1-4..
mpc8349_usb_cfg() had this inverted for the DR, and it also had the bit
positions of the usb0 / usb1 mux settings swapped. It would basically
work if you specified port1 instead of port0 for the MPH controller (and
happened to use ULPI phys), which is what all the 834x dts have done,
even though that configuration is physically invalid.
Instead fix mpc8349_usb_cfg() and adjust the dts files to match reality.
Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk> Signed-off-by: Kumar Gala <galak@kernel.crashing.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This fixes regression (battery "vanishing" on resume) introduced by
commit d0c71fe7ebc180f1b7bc7da1d39a07fc19eec768 ("ACPI Suspend: Enable
ACPI during resume if SCI_EN is not set") and also the issue with
the "screaming" IRQ 9.
Reported-and-tested-by: Alan Jenkins <alan-jenkins@tuffmail.co.uk> Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com> Acked-by: Len Brown <lenb@kernel.org> Signed-off-by: Rafael J. Wysocki <rjw@sisk.pl> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
These pointers can be NULL, the is_mesh() case isn't
ever hit in the current kernel, but cmp_ies() can be
hit under certain conditions.
Signed-off-by: Johannes Berg <johannes@sipsolutions.net> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
loff_t is a signed type. If userspace passes a negative ppos, the "count"
range check is weakened. "count"s bigger than HPEE_MAX_LENGTH will pass the check.
Also, if ppos is negative, the readb(eisa_eeprom_addr + *ppos) will poke in random
memory.
About a half events are missing when we splice_read
from trace_pipe. They are unexpectedly consumed because we ignore
the TRACE_TYPE_NO_CONSUME return value used by the function graph
tracer when it needs to consume the events by itself to walk on
the ring buffer.
The same problem appears with ftrace_dump()
Example of an output before this patch:
1) | ktime_get_real() {
1) 2.846 us | read_hpet();
1) 4.558 us | }
1) 6.195 us | }
After this patch:
0) | ktime_get_real() {
0) | getnstimeofday() {
0) 1.960 us | read_hpet();
0) 3.597 us | }
0) 5.196 us | }
The fix also applies on 2.6.30
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com> Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <4A6EEC52.90704@cn.fujitsu.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When print_graph_entry() computes a function call entry event, it needs
to also check the next entry to guess if it matches the return event of
the current function entry.
In order to look at this next event, it needs to consume the current
entry before going ahead in the ring buffer.
However, if the current event that gets consumed is the last one in the
ring buffer head page, the ring_buffer may reuse the page for writers.
The consumed entry will then become invalid because of possible
racy overwriting.
Me must then handle this entry by making a copy of it.
The fix also applies on 2.6.30
Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com> Cc: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <4A6EEAEC.3050508@cn.fujitsu.com> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Andrea Gelmini gave me a report that a kernel oops hit on a nilfs
filesystem with a 1KB block size when doing rsync.
This turned out to be caused by an inconsistency of dirty state
between a page and its buffers storing b-tree node blocks.
If the page had multiple buffers split over multiple logs, and if the
logs were written at a time, a dirty flag remained in the page even
every dirty flag in the buffers was cleared.
This will fix the failure by dropping the dirty flag properly for
pages with the discrete multiple b-tree nodes.
Reported-by: Andrea Gelmini <andrea.gelmini@gmail.com> Signed-off-by: Ryusuke Konishi <konishi.ryusuke@lab.ntt.co.jp> Tested-by: Andrea Gelmini <andrea.gelmini@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Currently, the ThinkPad-ACPI bay and dock drivers are completely
broken, and cause a NULL pointer derreference in kernel mode (and,
therefore, an OOPS) when they try to issue events (i.e. on dock,
undock, bay ejection, etc).
OTOH, the standard ACPI dock driver can handle the hotplug bays and
docks of the ThinkPads just fine (including batteries) as of 2.6.27.
In fact, it does a much better job of it than thinkpad-acpi ever did.
It is just not worth the hassle to find a way to fix this crap without
breaking the (deprecated) thinkpad-acpi dock/bay ABI. This is old,
deprecated code that sees little testing or use.
As a quick fix suitable for -stable backports, mark the thinkpad-acpi
bay and dock subdrivers as BROKEN in Kconfig. The dead code will be
removed by a later patch.
This fixes bugzilla #13669, and should be applied to 2.6.27 and later.
Signed-off-by: Henrique de Moraes Holschuh <hmh@hmh.eng.br> Reported-by: Joerg Platte <jplatte@naasa.net> Signed-off-by: Len Brown <len.brown@intel.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
If the referral is malformed or the hostname can't be resolved, then
the current code generates an oops. Fix it to handle these errors
gracefully.
Reported-by: Sandro Mathys <sm@sandro-mathys.ch> Acked-by: Igor Mammedov <niallain@gmail.com> Signed-off-by: Jeff Layton <jlayton@redhat.com> Signed-off-by: Steve French <sfrench@us.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There's a hotplug problem in the way libsas allocates ports: it loops over the
available ports first trying to add to an existing for a wide port and
otherwise allocating the next free port. This scheme only works if the port
array is packed from zero, which fails if a port gets hot unplugged and the
array becomes sparse. In that case, a new port is formed even if there's a
wide port it should be part of. Fix this by creating two loops over all the
ports: the first to see if the phy should be part of a wide port and the
second to form a new port in an empty port slot.
Signed-off-by: Tom Peng <tom_peng@usish.com> Signed-off-by: Jack Wang <jack_wang@usish.com> Signed-off-by: Lindar Liu <lindar_liu@usish.com> Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Requests to get max LUN, for certain USB storage devices, require a
longer timeout before a correct reply is returned. This happens for a
Realtek USB Card Reader (0bda:0152), which has a max LUN of 3 but is set
to 0, thus losing functionality, because of the timeout occurring too
quickly.
Raising the timeout value fixes the issue and might help other devices
to return a correct max LUN value as well.
Update directory hardlink count when moving kobjects to a new parent.
Fixes the following problem which occurs when several devices are
moved to the same parent and then unregistered:
Unitialized fence register could leads to corrupted display. Problem
encountered on MacBooks (revision 1 and 2), directly booting from EFI
or through BIOS emulation.
(bug #21710 at freedestop.org)
Signed-off-by: Grégoire Henry <henry@pps.jussieu.fr> Signed-off-by: Eric Anholt <eric@anholt.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
With the DRM-driven DPMS code, encoders are considered idle unless a
connector is hooked to them, so mode setting is skipped. This makes load
detection fail as none of the hardware is enabled.
Signed-off-by: Keith Packard <keithp@keithp.com> Signed-off-by: Eric Anholt <eric@anholt.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Fix a FIXME in the intel LVDS bring-up code, adding the appropriate
blacklist entry for the AOpen Mini PC, courtesy of a dmidecode
dump from Florian Demmer.
Signed-off-by: Jarod Wilson <jarod@redhat.com> CC: Florian Demmer <florian@demmer.org> Signed-off-by: Eric Anholt <eric@anholt.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
As transparent proxying looks up the socket early and assigns
it to the skb for later processing, we must drop any existing
socket ownership prior to that in order to distinguish between
the case where tproxy is active and where it is not.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
In order to get the tun driver to account packets, we need to be
able to receive packets with destructors set. To be on the safe
side, I added an skb_orphan call for all protocols by default since
some of them (IP in particular) cannot handle receiving packets
destructors properly.
Now it seems that at least one protocol (CAN) expects to be able
to pass skb->sk through the rx path without getting clobbered.
So this patch attempts to fix this properly by moving the skb_orphan
call to where it's actually needed. In particular, I've added it
to skb_set_owner_[rw] which is what most users of skb->destructor
call.
This is actually an improvement for tun too since it means that
we only give back the amount charged to the socket when the skb
is passed to another socket that will also be charged accordingly.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Tested-by: Oliver Hartkopp <olver@hartkopp.net> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
As it stands skb fraglists can get past the check in dev_queue_xmit
if the skb is marked as GSO. In particular, if the packet doesn't
have the proper checksums for GSO, but can otherwise be handled by
the underlying device, we will not perform the fraglist check on it
at all.
If the underlying device cannot handle fraglists, then this will
break.
The fix is as simple as moving the fraglist check from the device
check into skb_gso_ok.
This has caused crashes with Xen when used together with GRO which
can generate GSO packets with fraglists.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When NAPI is disabled while we're in net_rx_action, we end up
calling __napi_complete without flushing GRO packets. This is
a bug as it would cause the GRO packets to linger, of course it
also literally BUGs to catch error like this :)
This patch changes it to napi_complete, with the obligatory IRQ
reenabling. This should be safe because we've only just disabled
IRQs and it does not materially affect the test conditions in
between.
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Commit e912b1142be8f1e2c71c71001dc992c6e5eb2ec1
(net: sk_prot_alloc() should not blindly overwrite memory)
took care of not zeroing whole new socket at allocation time.
sock_copy() is another spot where we should be very careful.
We should not set refcnt to a non null value, until
we are sure other fields are correctly setup, or
a lockless reader could catch this socket by mistake,
while not fully (re)initialized.
This patch puts sk_node & sk_refcnt to the very beginning
of struct sock to ease sock_copy() & sk_prot_alloc() job.
We add appropriate smp_wmb() before sk_refcnt initializations
to match our RCU requirements (changes to sock keys should
be committed to memory before sk_refcnt setting)
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Some sockets use SLAB_DESTROY_BY_RCU, and our RCU code correctness
depends on sk->sk_nulls_node.next being always valid. A NULL
value is not allowed as it might fault a lockless reader.
Current sk_prot_alloc() implementation doesnt respect this hypothesis,
calling kmem_cache_alloc() with __GFP_ZERO. Just call memset() around
the forbidden field.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The host-side CDC subset driver is binding more specifically
than it should ... only to PXA 210/25x/26x Linux-USB gadgets.
Loosen that restriction to match the gadget driver driver.
This will various PXA 27x and PXA 3xx devices happier when
talking to Linux hosts, potentially others.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Tested-by: Aric D. Blumer <aric@sdgsystems.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Converting checksum field from le16 to CPU byte order fixes the issue.
Signed-off-by: Anton Vorontsov <avorontsov@ru.mvista.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Our CAST algorithm is called cast5, not cast128. Clearly nobody
has ever used it :)
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
E100 places it's RX packet descriptors inside skb->data and uses them
with bidirectional streaming DMA mapping. Unfortunately it fails to
transfer skb->data ownership to the device after it reads the
descriptor's status, breaking on non-coherent (e.g., ARM) platforms.
This have to be converted to use coherent memory for the descriptors.
Signed-off-by: Krzysztof Halasa <khc@pm.waw.pl> Acked-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
While testing the driver on PPC, we ran into a crash with LRO, Jumbo frames.
With CONFIG_PPC_64K_PAGES configured (a default in PPC), MAX_SKB_FRAGS drops to 3 and we were crossing the array limits on skb_shinfo(skb)->frags[].
Now we coalesce the frags from the same physical page into one slot in
skb_shinfo(skb)->frags[] and go to the next index when the frag is from
different physical page.
This patch is against the net-2.6 tree.
Signed-off-by: Ajit Khaparde <ajitk@serverengines.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This changes the power_level file to adhere to the "one value
per file" sysfs rule. The user will know which power level was
requested as it will be the number just written to this file. It
is thus not necessary to create a new sysfs file for this value.
In addition it fixes a problem where powertop's parsing expects
this value to be the first value in this file without any descriptions.
Signed-off-by: Reinette Chatre <reinette.chatre@intel.com> Signed-off-by: John W. Linville <linville@tuxdriver.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Tag 11 packets are stored in the metadata section of an eCryptfs file to
store the key signature(s) used to encrypt the file encryption key.
After extracting the packet length field to determine the key signature
length, a check is not performed to see if the length would exceed the
key signature buffer size that was passed into parse_tag_11_packet().
Thanks to Ramon de Carvalho Valle for finding this bug using fsfuzzer.
With the "security: use mmap_min_addr indepedently of security models"
change, mmap_min_addr is used in common areas, which susbsequently blows
up the nommu build. This stubs in the definition in the nommu case as
well.
Signed-off-by: Paul Mundt <lethal@linux-sh.org> Cc: Mike Frysinger <vapier.adi@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de> Signed-off-by: James Morris <jmorris@namei.org>
ata_eh_reset() was missing error return handling after follow-up SRST
allowing EH to continue the normal probing path after reset failure.
This was discovered while testing new WD 2TB drives which take longer
than 10 secs to spin up and cause the first follow-up SRST to time
out.
This patch adds DMI information to automatically load the correct
layout for the Maxdata Pro 7000X/DX notebook models. Such notebooks
are clones of Fujitsu Amilo V2000, the hook for the v2000 is being
used and I have tested that perfectly works.
The immediate result of integrating this patch is that the five
special buttons will work on these specific notebook models and that
the RF killswitch will not be activated after suspend. This patch
definitively obsoletes the fsam7400 module which I was still needing
to enable wifi and to fix the RF killswitch suspend problem; in the
current 2.6.30 kernel it is necessary to load the wistron_btns module
with options 'force=1 keymap=1557/MS2141', which was not anyway a
complete workaround.
alloc_etherdev() used to install a default implementation of this
operation, but it must now be explicitly installed in struct
net_device_ops.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
alloc_etherdev() used to install default implementations of these
operations, but they must now be explicitly installed in struct
net_device_ops.
Signed-off-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When a slab cache uses SLAB_DESTROY_BY_RCU, we must be careful when allocating
objects, since slab allocator could give a freed object still used by lockless
readers.
In particular, nf_conntrack RCU lookups rely on ct->tuplehash[xxx].hnnode.next
being always valid (ie containing a valid 'nulls' value, or a valid pointer to next
object in hash chain.)
kmem_cache_zalloc() setups object with NULL values, but a NULL value is not valid
for ct->tuplehash[xxx].hnnode.next.
Fix is to call kmem_cache_alloc() and do the zeroing ourself.
As spotted by Patrick, we also need to make sure lookup keys are committed to
memory before setting refcount to 1, or a lockless reader could get a reference
on the old version of the object. Its key re-check could then pass the barrier.
Signed-off-by: Eric Dumazet <eric.dumazet@gmail.com> Signed-off-by: Patrick McHardy <kaber@trash.net> Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
When NAT helpers change the TCP packet size, the highest seen sequence
number needs to be corrected. This is currently only done upwards, when
the packet size is reduced the sequence number is unchanged. This causes
TCP conntrack to falsely detect unacknowledged data and decrease the
timeout.
Fix by updating the highest seen sequence number in both directions after
packet mangling.
Tested-by: Krzysztof Piotr Oledzki <ole@ans.pl> Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Compiling the kernel with clang has shown this warning:
net/netfilter/xt_rateest.c:69:16: warning: self-comparison always results in a
constant value
ret &= pps2 == pps2;
^
Looking at the code:
if (info->flags & XT_RATEEST_MATCH_BPS)
ret &= bps1 == bps2;
if (info->flags & XT_RATEEST_MATCH_PPS)
ret &= pps2 == pps2;
Judging from the MATCH_BPS case it seems to be a typo, with the intention of
comparing pps1 with pps2.
http://bugzilla.kernel.org/show_bug.cgi?id=13535
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Commit v2.6.29-rc5-872-gacc738f ("xtables: avoid pointer to self")
forgot to copy the initial quota value supplied by iptables into the
private structure, thus counting from whatever was in the memory
kmalloc returned.
Signed-off-by: Jan Engelhardt <jengelh@medozas.de> Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The RCU protected conntrack hash lookup only checks whether the entry
has a refcount of zero to decide whether it is stale. This is not
sufficient, entries are explicitly removed while there is at least
one reference left, possibly more. Explicitly check whether the entry
has been marked as dying to fix this.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
New connection tracking entries are inserted into the hash before they
are fully set up, namely the CONFIRMED bit is not set and the timer not
started yet. This can theoretically lead to a race with timer, which
would set the timeout value to a relative value, most likely already in
the past.
Perform hash insertion as the final step to fix this.
Signed-off-by: Patrick McHardy <kaber@trash.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Commit 31207dab7d2e63795eb15823947bd2f7025b08e2
"Fix incorrect allocation of interrupt rev-map"
introduced a regression crashing on boot on machines using
a "DCR" based MPIC, such as the Cell blades.
The reason is that the irq host data structure is initialized
much later as a result of that patch, causing our calls to
mpic_map() do be done before we have a host setup.
Unfortunately, this breaks _mpic_map_dcr() which uses the
mpic->irqhost to get to the device node.
This fixes it by, instead, passing the device node explicitely
to mpic_map().
In testing a backport of the write_begin/write_end AOPs, a 10% re-read
regression was noticed when running iozone. This regression was
introduced because the old AOPs would always do a mark_page_accessed(page)
after the commit_write, but when the new AOPs where introduced, the only
place this was kept was in pagecache_write_end().
This patch does the same thing in the generic case as what is done in
pagecache_write_end(), which is just to mark the page accessed before we
do write_end().
Signed-off-by: Josef Bacik <jbacik@redhat.com> Acked-by: Nick Piggin <npiggin@suse.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
It's really not right to use 'access_ok()', since that is meant for the
normal "get_user()" and "copy_from/to_user()" accesses, which are done
through the TLB, rather than through the page tables.
Why? access_ok() does both too few, and too many checks. Too many,
because it is meant for regular kernel accesses that will not honor the
'user' bit in the page tables, and because it honors the USER_DS vs
KERNEL_DS distinction that we shouldn't care about in GUP. And too few,
because it doesn't do the 'canonical' check on the address on x86-64,
since the TLB will do that for us.
So instead of using a function that isn't meant for this, and does
something else and much more complicated, just do the real rules: we
don't want the range to overflow, and on x86-64, we want it to be a
canonical low address (on 32-bit, all addresses are canonical).
Acked-by: Ingo Molnar <mingo@elte.hu> Cc: H. Peter Anvin <hpa@zytor.com> Cc: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
vmscan: do not unconditionally treat zones that fail zone_reclaim() as full
On NUMA machines, the administrator can configure zone_reclaim_mode that
is a more targetted form of direct reclaim. On machines with large NUMA
distances for example, a zone_reclaim_mode defaults to 1 meaning that
clean unmapped pages will be reclaimed if the zone watermarks are not
being met. The problem is that zone_reclaim() failing at all means the
zone gets marked full.
This can cause situations where a zone is usable, but is being skipped
because it has been considered full. Take a situation where a large tmpfs
mount is occuping a large percentage of memory overall. The pages do not
get cleaned or reclaimed by zone_reclaim(), but the zone gets marked full
and the zonelist cache considers them not worth trying in the future.
This patch makes zone_reclaim() return more fine-grained information about
what occured when zone_reclaim() failued. The zone only gets marked full
if it really is unreclaimable. If it's a case that the scan did not occur
or if enough pages were not reclaimed with the limited reclaim_mode, then
the zone is simply skipped.
There is a side-effect to this patch. Currently, if zone_reclaim()
successfully reclaimed SWAP_CLUSTER_MAX, an allocation attempt would go
ahead. With this patch applied, zone watermarks are rechecked after
zone_reclaim() does some work.
This bug was introduced by commit 9276b1bc96a132f4068fdee00983c532f43d3a26
("memory page_alloc zonelist caching speedup") way back in 2.6.19 when the
zonelist_cache was introduced. It was not intended that zone_reclaim()
aggressively consider the zone to be full when it failed as full direct
reclaim can still be an option. Due to the age of the bug, it should be
considered a -stable candidate.
x86, setup (2.6.30-stable) fix 80x34 and 80x60 console modes
Note: this is not in upstream since upstream is not affected due to the
new "BIOS glovebox" subsystem.
As coded, most INT10 calls in video-vga.c allow the compiler to assume
EAX remains unchanged across them, which is not always the case. This
triggers an optimisation issue that causes vga_set_vertical_end() to be
called with an incorrect number of scanlines. Fix this by beefing up
the asm constraints on these calls.
Reported-by: Marc Aurele La France <tsi@xfree86.org> Signed-off-by: Marc Aurele La France <tsi@xfree86.org> Acked-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
There are two reasons to expose the memory *a in the asm:
1) To prevent the compiler from discarding a preceeding write to *a, and
2) to prevent it from caching *a in a register over the asm.
The change has had a few days testing with a SMP build of 2.6.22.19
running on a rp3440.
This patch is about the correctness of the __ldcw() macro itself.
The use of the macro should be confined to small inline functions
to try to limit the effect of clobbering memory on GCC's optimization
of loads and stores.
The TLB flushing functions on hppa, which causes PxTLB broadcasts on the system
bus, needs to be protected by irq-safe spinlocks to avoid irq handlers to deadlock
the kernel. The deadlocks only happened during I/O intensive loads and triggered
pretty seldom, which is why this bug went so long unnoticed.
Signed-off-by: Helge Deller <deller@gmx.de>
[edited to use spin_lock_irqsave on UP as well since we'd been locking there
all this time anyway, --kyle] Signed-off-by: Kyle McMartin <kyle@mcmartin.ca> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
AMI BIOS with low memory corruption was found on Intel DG45ID
board (Bug 13710). Add this board to the blacklist - in the
(somewhat optimistic) hope of future boards/BIOSes from Intel
not having this bug.
The movq instruction, generated by __put_user_asm() when used for
64-bit data, takes a sign-extended immediate ("e") not a zero-extended
immediate ("Z").
Signed-off-by: H. Peter Anvin <hpa@zytor.com> Cc: Uros Bizjak <ubizjak@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
arch/x86/include/asm/uaccess_64.h uses wrong asm operand constraint
("ir") for movq insn. Since movq sign-extends its immediate operand,
"er" constraint should be used instead.
Attached patch changes all uses of __put_user_asm in uaccess_64.h to use
"er" when "q" insn suffix is involved.
Patch was compile tested on x86_64 with defconfig.
Signed-off-by: Uros Bizjak <ubizjak@gmail.com> Signed-off-by: H. Peter Anvin <hpa@zytor.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Timer interrupts are excluded from being disabled during suspend. The
clock events code manages the disabling of clock events on its own
because the timer interrupt needs to be functional before the resume
code reenables the device interrupts.
The mfgpt timer request its interrupt without setting the IRQF_TIMER
flag so suspend_device_irqs() disables it as well which results in a
fatal resume failure.
Adding IRQF_TIMER to the interupt flags when requesting the mrgpt
timer interrupt solves the problem.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
LKML-Reference: <new-submission> Cc: Andres Salomon <dilinger@debian.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The recent commit 7513c2a761d69d2a93f17146b3563527d3618ba0 (dm raid1:
add is_remote_recovering hook for clusters) changed do_writes() to
update the ms->writes list but forgot to wake up kmirrord to process it.
The rule is that when anything is being added on ms->reads, ms->writes
or ms->failures and the list was empty before we must call
wakeup_mirrord (for immediate processing) or delayed_wake (for delayed
processing). Otherwise the bios could sit on the list indefinitely.
commit e3c8ca8336 (sched: do not count frozen tasks toward load) broke
the nr_uninterruptible accounting on freeze/thaw. On freeze the task
is excluded from accounting with a check for (task->flags &
PF_FROZEN), but that flag is cleared before the task is thawed. So
while we prevent that the task with state TASK_UNINTERRUPTIBLE
is accounted to nr_uninterruptible on freeze we decrement
nr_uninterruptible on thaw.
Use a separate flag which is handled by the freezing task itself. Set
it before calling the scheduler with TASK_UNINTERRUPTIBLE state and
clear it after we return from frozen state.
Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
x86: multi pci root bus with different io resource range, on
64-bit
It turns out that:
1. that AMD-based systems have two HT chains.
2. BIOS doesn't allocate resources for BAR 6 of devices under 8132 etc
3. that multi-peer-root patch will try to split root resources to peer
root resources according to PCI conf of NB
4. PCI core assigns unassigned resources, but they overlap with BARs
that are used by ioapic addr of io4 and 8132.
The reason: at that point ioapic address are not inserted yet. Solution
is to insert ioapic resources into the tree a bit earlier.
Reported-by: Stephen Frost <sfrost@snowman.net> Reported-and-Tested-by: dann frazier <dannf@hp.com> Signed-off-by: Yinghai Lu <yinghai@kernel.org> Signed-off-by: Jesse Barnes <jbarnes@virtuousgeek.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
We want to set _spin_*:traceon:10 to set_ftrace_filter, it complains
with "Invalid argument", but the operation is successful.
This is because ftrace_process_regex() returns the number of functions that
matched the pattern. If the number is not 0, this value is returned
by ftrace_regex_write() whereas we want to return the number of bytes
virtually written.
Also the file offset pointer is not updated in this case.
If the number of matched functions is lower than the number of bytes written
by the user, this results to a reprocessing of the string given by the user with
a lower size, leading to a malformed ftrace regex and then a -EINVAL returned.
So, this patch fixes it by returning 0 if no error occured.
The fix also applies on 2.6.30
Fixes an easily triggerable BUG() when setting process affinities.
Make sure to count the number of migratable tasks in the same place:
the root rt_rq. Otherwise the number doesn't make sense and we'll hit
the BUG in set_cpus_allowed_rt().
Also, make sure we only count tasks, not groups (this is probably
already taken care of by the fact that rt_se->nr_cpus_allowed will be 0
for groups, but be more explicit)
This fixes a bug that checkpoint count gets wrong on errors when
deleting a series of checkpoints.
The count error is persistent since the checkpoint count is stored on
disk. Some userland programs refer to the count via ioctl, and this
bugfix is needed to prevent malfunction of such programs.
Leandro Lucarella gave me a report that nilfs gets stuck after its
write function fails.
The problem turned out to be caused by bugs which leave writeback flag
on pages. This fixes the problem by ensuring to clear the writeback
flag in error path.
=
The semantics of SG_DXFER_TO_FROM_DEV were:
- copy user space buffer to kernel (LLD) buffer
- do SCSI command which is assumed to be of the DATA_IN
(data from device) variety. This would overwrite
some or all of the kernel buffer
- copy kernel (LLD) buffer back to the user space.
The idea was to detect short reads by filling the original
user space buffer with some marker bytes ("0xec" it would
seem in this report). The "resid" value is a better way
of detecting short reads but that was only added this century
and requires co-operation from the LLD.
=
This patch changes the block layer mapping API to support this
semantics. This simply adds another field to struct rq_map_data and
enables __bio_copy_iov() to copy data from user space even with READ
requests.
It's better to add the flags field and kills null_mapped and the new
from_user fields in struct rq_map_data but that approach makes it
difficult to send this patch to stable trees because st and osst
drivers use struct rq_map_data (they were converted to use the block
layer in 2.6.29 and 2.6.30). Well, I should clean up the block layer
mapping API.
zhou sf reported this regiression and tested this patch:
> Since the fixmap pages are assigned higher address to lower,
> text_poke() has to use it with inverted order (FIX_TEXT_POKE1
> to FIX_TEXT_POKE0).
I prefer to just invert the order of the fixmap declaration.
It's simpler and more straightforward.
Backward fixmaps seems to be used by both x86 32 and 64.
It's really rare but a nasty bug, because it only hurts when
instructions to patch are crossing a page boundary. If this
happens, the fixmap write accesses will spill on the following
fixmap, which may very well crash the system. And this does not
crash the system, it could leave illegal instructions in place.
Thanks Masami for finding this.
It seems to have crept into the 2.6.30-rc series, so this calls
for a -stable inclusion.
The merge of the 32- and 64-bit fixmap headers made a latent
bug on x86-64 a real one: with the right config settings
it is possible for FIX_OHCI1394_BASE to overlap the FIX_BTMAP_*
range.
The buffer for the consoles are unconditionally allocated at con_init()
time, which miss the creation of the vcs(a) devices.
Since 2.6.30 (commit 4995f8ef9d3aac72745e12419d7fbaa8d01b1d81, 'vcs:
hook sysfs devices into object lifetime instead of "binding"' to be
exact) these devices are no longer created at open() and removed on
close(), but controlled by the lifetime of the buffers.
git commit f67f129e "Driver core: implement uevent suppress in kobject"
contains this chunk for fs/partitions/check.c:
/* suppress uevent if the disk supresses it */
- if (!ddev->uevent_suppress)
+ if (!dev_get_uevent_suppress(pdev))
kobject_uevent(&pdev->kobj, KOBJ_ADD);
However that should have been
- if (!ddev->uevent_suppress)
+ if (!dev_get_uevent_suppress(ddev))
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Acked-by: Ming Lei <tom.leiming@gmail.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Standard data flow for MMC/SD/SDIO cards requires that the mvsdio
controller be set for big endian operation. This is causing problems
with buffers which length is not a multiple of 4 bytes as the last
partial word doesn't get shifted all the way and stored properly in
memory. Let's compensate for this.
Gateway T6834c laptops need EAPD always on while the default behavior
for the STAC9205 reference board is to turn it off upon every HP plug.
By using the special "eapd" model, which is first introduced for Gateway
T1616 laptops for this same reason, this peculiarity can be properly
handled.
The recent rewrite of the codec parser for STAC9872 caused a regression
for some Sony VAIO models that don't give proper pin default configs
by BIOS. Even using model=vaio doesn't work because the pin definitions
are set after the pin overrides.
This patch fixes the pin definitions in patch_stac9872() to be put
in the right place before the pin overrides. Also the patch adds the
new quirk entry for VAIO F/S to have the correct pin default configs.
The capture buffer size with 64kB seems broken with CA0106.
At least, either the update timing or the DMA position is wrong,
and this screws up pulseaudio badly.
This patch restricts the max buffer size less than that to make life
a bit easier.
This patch (as1259b) makes ehci-hcd return the total number of bytes
transferred in urb->actual_length for Isochronous transfers.
Until now, the actual_length value was unaccountably left at 0.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Acked-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
commit b7af0bb ("USB: allow malformed LANGID descriptors") broke support
for devices without string descriptor support.
Reporting string descriptors is optional to USB devices, and a device
lets us know it can't deal with strings by responding to the LANGID
request with a STALL token.
The kernel handled that correctly before b7af0bb came in, but failed
hard if the LANGID was reported but broken. More than that, if a device
was not able to provide string descriptors, the LANGID was retrieved
over and over again at each string read request.
This patch changes the behaviour so that
a) the LANGID is only queried once
b) devices which can't handle string requests are not asked again
c) devices with malformed LANGID values have a sane fallback to 0x0409
Signed-off-by: Daniel Mack <daniel@caiaq.de> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
The reworked Ethernet gadget has an RNDIS interop problem when used
with the CDC subset driver ... e.g. on PXA 2xx and 3xx hardware,
which currently has a hard time talking to MS-Windows hosts.
The issue is that Microsoft requires USB_CLASS_COMM. Fix by tweaking
the CDC subset driver to not switch to USB_CLASS_VENDOR_SPEC if RNDIS
is used in some other device configuration.
[ UPDATED: some "statements" were comma-terminated; fix that. ]
commit 1a1fab513734b3a4fca1bee8229e5ff7e1cb873c accidentally added the
device id to both tables in the driver, which causes problems as this is
only a single port device, not a multiple port device.
projects / karo-tx-linux.git / log