USB2 LPM is disabled when device begin to suspend and enabled after device
is resumed. That's because USB spec does not define the transition from
U1/U2 state to U3 state.
If usb_port_suspend() fails, usb_port_resume() is never called, and USB2 LPM
is disabled in this situation. Enable USB2 LPM if port suspend fails.
Sergio reported that when he recorded audio from a USB headset mic
plugged into the USB 3.0 port on his ASUS N53SV-DH72, the audio sounded
"robotic". When plugged into the USB 2.0 port under EHCI on the same
laptop, the audio sounded fine. The device is:
Bus 002 Device 004: ID 046d:0a0c Logitech, Inc. Clear Chat Comfort USB Headset
The problem was tracked down to the Fresco Logic xHCI host controller
not correctly reporting short transfers on isochronous IN endpoints.
The driver would submit a 96 byte transfer, the device would only send
88 or 90 bytes, and the xHCI host would report the transfer had a
"successful" completion code, with an untransferred buffer length of 8
or 6 bytes.
The successful completion code and non-zero untransferred length is a
contradiction. The xHCI host is supposed to only mark a transfer as
successful if all the bytes are transferred. Otherwise, the transfer
should be marked with a short packet completion code. Without the EHCI
bus trace, we wouldn't know whether the xHCI driver should trust the
completion code or the untransferred length. With it, we know to trust
the untransferred length.
Add a new xHCI quirk for the Fresco Logic host controller. If a
transfer is reported as successful, but the untransferred length is
non-zero, print a warning. For the Fresco Logic host, change the
completion code to COMP_SHORT_TX and process the transfer like a short
transfer.
This should be backported to stable kernels that contain the commit f5182b4155b9d686c5540a6822486400e34ddd98 "xhci: Disable MSI for some
Fresco Logic hosts." That commit was marked for stable kernels as old
as 2.6.36.
Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Reported-by: Sergio Correia <lists@uece.net> Tested-by: Sergio Correia <lists@uece.net> Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
worker_enter_idle() has WARN_ON_ONCE() which triggers if nr_running
isn't zero when every worker is idle. This can trigger spuriously
while a cpu is going down due to the way trustee sets %WORKER_ROGUE
and zaps nr_running.
It first sets %WORKER_ROGUE on all workers without updating
nr_running, releases gcwq->lock, schedules, regrabs gcwq->lock and
then zaps nr_running. If the last running worker enters idle
inbetween, it would see stale nr_running which hasn't been zapped yet
and trigger the WARN_ON_ONCE().
Fix it by performing the sanity check iff the trustee is idle.
Signed-off-by: Tejun Heo <tj@kernel.org> Reported-by: "Paul E. McKenney" <paulmck@linux.vnet.ibm.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
We cannot dereference a removed USB interface for
dev_printk. Use pr_debug instead where necessary.
Flush errors are expected if device is unplugged and are
therefore best ingored at this point.
Move the kill_urbs() call in wdm_release with dev_dbg()
for the non disconnect, as we know it has already been
called if WDM_DISCONNECTING is set. This does not
actually fix anything, but keeps the code more consistent.
Cc: Oliver Neukum <oliver@neukum.org> Signed-off-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Device state cleanup is done in either wdm_disconnect or
wdm_release depending on the order they are called. Adding
a couple of debug messages to document the program flow.
Signed-off-by: Bjørn Mork <bjorn@mork.no> Acked-by: Oliver Neukum <oneukum@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The Kontron M2M development board, also known as the Fish River Island II,
has an optional daughter card providing access to the PCH_UART (EG20T) via
a ti_usb_3410_5052 uart to usb chip.
Signed-off-by: Darren Hart <dvhart@linux.intel.com> CC: Al Borchers <alborchers@steinerpoint.com> CC: Peter Berger <pberger@brimson.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
On big-endian systems (e.g., Apple PowerBook), trying to use a
logitech wireless mouse with the Logitech Unifying Receiver does not
work with v3.2 and later kernels. The device doesn't show up in
/dev/input. Older kernels work fine.
That is because the new hid-logitech-dj driver claims the device. The
device arrival notification appears:
20 00 41 02 00 00 00 00 00 00 00 00 00 00 00
and we read the report_types bitfield (02 00 00 00) to find out what
kind of device it is. Unfortunately the driver only reads the first 8
bits and treats that value as a 32-bit little-endian number, so on a
powerpc the report type seems to be 0x02000000 and is not recognized.
Even on little-endian machines, connecting a media center remote
control (report type 00 01 00 00) with this driver loaded would
presumably fail for the same reason.
Fix both problems by using get_unaligned_le32() to read all four
bytes, which is a little clearer anyway. After this change, the
wireless mouse works on Hugo's PowerBook again.
Based on a patch by Nestor Lopez Casado.
Addresses http://bugs.debian.org/671292
Reported-by: Hugo Osvaldo Barrera <hugo@osvaldobarrera.com.ar> Inspired-by: Nestor Lopez Casado <nlopezcasad@logitech.com> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Nestor Lopez Casado <nlopezcasad@logitech.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
A possible race condition appears because we are not initializing
the ohci->regs before calling usb_hcd_request_irqs().
We move the call to ohci_init() in hcd->driver->reset() instead of
hcd->driver->start() to fix this.
This was experienced when we share the same IRQ line between OHCI and EHCI
controllers.
Signed-off-by: Nicolas Ferre <nicolas.ferre@atmel.com> Tested-by: Christian Eggers <christian.eggers@kathrein.de> Acked-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
We incorrectly parse incoming IR data. The extra byte contains the upper
bits and not the lower bits of the x/y coordinates. User-space expects
absolute position data from us so this patch does not break existing
applications. On the contrary, it extends the virtual view and fixes
garbage reports for margin areas of the virtual screen.
Reported-by: Peter Bukovsky <bukovsky.peter@gmail.com> Signed-off-by: David Herrmann <dh.herrmann@googlemail.com> Signed-off-by: Jiri Kosina <jkosina@suse.cz>
[bwh: Backported to 3.2: adjust filename] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The out functions should only handle actual available data instead of the complete buffer.
Otherwise for example the ep0_consume function will report ghost events since it tries to decode
the complete buffer - which may contain partly invalid data.
Signed-off-by: Matthias Fend <matthias.fend@wolfvision.net> Acked-by: Michal Nazarewicz <mina86@mina86.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
dTD's next dtd pointer need to be updated once CPU writes it, or this
request may not be handled by controller, then host will get NAK from
device forever.
This problem occurs when there is a request is handling, we need to add
a new request to dTD list, if this new request is added before the current
one is finished, the new request is intended to added as next dtd pointer
at current dTD, but without wmb(), the dTD's next dtd pointer may not be
updated when the controller reads it. In that case, the controller will
still get Terminate Bit is 1 at dTD's next dtd pointer, that means there is
no next request, then this new request is missed by controller.
Signed-off-by: Peter Chen <peter.chen@freescale.com> Acked-by: Li Yang <leoli@freescale.com> Signed-off-by: Felipe Balbi <balbi@ti.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The upcoming Intel Lynx Point chipset includes an xHCI host controller
that can have ports switched from the EHCI host controller, just like
the Intel Panther Point xHCI host. This time, ports from both EHCI
hosts can be switched to the xHCI host controller. The PCI config
registers to do the port switching are in the exact same place in the
xHCI PCI configuration registers, with the same semantics.
Hooray for shipping patches for next-gen hardware before the current gen
hardware is even available for purchase!
This patch should be backported to stable kernels as old as 3.0,
that contain commit 69e848c2090aebba5698a1620604c7dccb448684
"Intel xhci: Support EHCI/xHCI port switching."
Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
If the user chooses to say "no" to CONFIG_USB_XHCI_HCD on a system
with an Intel Panther Point chipset, the PCI quirks code or the EHCI
driver will switch the ports over to the xHCI host, but the xHCI driver
will never load. The ports will be powered off and seem "dead" to the
user.
Fix this by only switching the ports over if CONFIG_USB_XHCI_HCD is
either compiled in, or compiled as a module.
This patch should be backported to stable kernels as old as 3.0,
that contain commit 69e848c2090aebba5698a1620604c7dccb448684
"Intel xhci: Support EHCI/xHCI port switching."
Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Reported-by: Eric Anholt <eric.anholt@intel.com> Reported-by: David Bein <d.bein@f5.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
While testing unplugging an UVC HD webcam with usb-redirection (so through
usbdevfs), my userspace usb-redir code was getting a value of -1 in
iso_frame_desc[n].status, which according to Documentation/usb/error-codes.txt
is not a valid value.
The source of this -1 is the default case in xhci-ring.c:process_isoc_td()
adding a kprintf there showed the value of trb_comp_code to be COMP_TX_ERR
in this case, so this patch adds handling for that completion code to
process_isoc_td().
This was observed and tested with the following xhci controller:
1033:0194 NEC Corporation uPD720200 USB 3.0 Host Controller (rev 04)
Note: I also wonder if setting frame->status to -1 (-EPERM) is the best we can
do, but since I cannot come up with anything better I've left that as is.
This patch should be backported to kernels as old as 2.6.36, which contain the
commit 04e51901dd44f40a5a385ced897f6bca87d5f40a "USB: xHCI: Isochronous
transfer implementation".
Signed-off-by: Hans de Goede <hdegoede@redhat.com> Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
We noticed that we were loosing data at speed less than 2400 baud.
It turned out our (TI16750 compatible) uart with 64 byte outgoing fifo
was truncated to 16 byte (bit 5 sets fifo len) when modifying the fcr
reg.
The input code still fills the buffer with 64 bytes if I remember
correctly and thus data is lost.
Our fix was to remove whiping of the fcr content and just add the
TRIGGER_1 which we want for latency.
I can't see why this would not work on less than 2400 always, for all
uarts ...
Otherwise one would have to make sure the filling of the fifo re-checks
the current state of available fifo size (urrk).
Signed-off-by: Christian Melki <christian.melki@ericsson.se> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2: adjust filename; replace *port with up->port] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The rules used to make 8250_pci "ignore" the PCH uarts are lacking pci subids
entries, preventing it to match and thus is breaking serial port support for
theses systems.
This has been tested on a nanoETXexpress-TT, which has a specifici uart clock.
cleanup() is not called if the last close() comes after
disconnect(). That leads to a memory leak. Rectified
by checking for an earlier disconnect() in release()
Signed-off-by: Oliver Neukum <oneukum@suse.de> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
[bwh: Backported to 3.2: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
We had for some reason overlooked the AIO interface, and it didn't use
the proper rw_verify_area() helper function that checks (for example)
mandatory locking on the file, and that the size of the access doesn't
cause us to overflow the provided offset limits etc.
Instead, AIO did just the security_file_permission() thing (that
rw_verify_area() also does) directly.
This fixes it to do all the proper helper functions, which not only
means that now mandatory file locking works with AIO too, we can
actually remove lines of code.
During early boot, when the scheduler hasn't really been fully set up,
we really can't do blocking allocations because with certain (dubious)
configurations the "might_resched()" calls can actually result in
scheduling events.
We could just make such users always use GFP_ATOMIC, but quite often the
code that does the allocation isn't really aware of the fact that the
scheduler isn't up yet, and forcing that kind of random knowledge on the
initialization code is just annoying and not good for anybody.
And we actually have a the 'gfp_allowed_mask' exactly for this reason:
it's just that the kernel init sequence happens to set it to allow
blocking allocations much too early.
So move the 'gfp_allowed_mask' initialization from 'start_kernel()'
(which is some of the earliest init code, and runs with preemption
disabled for good reasons) into 'kernel_init()'. kernel_init() is run
in the newly created thread that will become the 'init' process, as
opposed to the early startup code that runs within the context of what
will be the first idle thread.
So by the time we reach 'kernel_init()', we know that the scheduler must
be at least limping along, because we've already scheduled from the idle
thread into the init thread.
Reported-by: Steven Rostedt <rostedt@goodmis.org> Cc: David Rientjes <rientjes@google.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
OEM parameters [1] are parsed from the platform option-rom / efi
driver. By default the driver was validating the parameters for the
dual-controller case, but in single-controller case only the first set
of parameters may be valid.
Limit the validation to the number of actual controllers detected
otherwise the driver may fail to parse the valid parameters leading to
driver-load or runtime failures.
[1] the platform specific set of phy address, configuration,and analog
tuning values
[stable v3.0+] Reported-by: Dave Jiang <dave.jiang@intel.com> Tested-by: Dave Jiang <dave.jiang@intel.com> Signed-off-by: Dan Williams <dan.j.williams@intel.com>
[bwh: Backported to 3.2: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When setting the current task state to TASK_UNINTERRUPTIBLE this can
race with a different cpu. The other cpu could set the task state after
it inspected it (while it was still TASK_RUNNING) to TASK_RUNNING which
would change the state from TASK_UNINTERRUPTIBLE to TASK_RUNNING again.
This race was always present in the pfault interrupt code but didn't
cause anything harmful before commit f2db2e6c "[S390] pfault: cpu hotplug
vs missing completion interrupts" which relied on the fact that after
setting the task state to TASK_UNINTERRUPTIBLE the task would really
sleep.
Since this is not necessarily the case the result may be a list corruption
of the pfault_list or, as observed, a use-after-free bug while trying to
access the task_struct of a task which terminated itself already.
To fix this, we need to get a reference of the affected task when receiving
the initial pfault interrupt and add special handling if we receive yet
another initial pfault interrupt when the task is already enqueued in the
pfault list.
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com> Reviewed-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Log a warning and drop the abort message. Otherwise we will do a
bogus wake_up() and crash.
Signed-off-by: Steve Wise <swise@opengridcomputing.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This fixes a race where an ingress abort fails to wake up the thread
blocked in rdma_init() causing the app to hang.
Signed-off-by: Steve Wise <swise@opengridcomputing.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
If we fail while registering a regulator make sure we release the supply
for the regulator if there is one.
Signed-off-by: Mark Brown <broonie@opensource.wolfsonmicro.com> Acked-by: Liam Girdwood <lrg@ti.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Commit bc3e53f682d9 ("mm: distinguish between mlocked and pinned
pages") introduced a separate counter for pinned pages and used it in
the IB stack. However, in ib_umem_get() the pinned counter is
incremented, but ib_umem_release() wrongly decrements the locked
counter. Fix this.
Signed-off-by: Yishai Hadas <yishaih@mellanox.com> Reviewed-by: Christoph Lameter <cl@linux.com> Signed-off-by: Roland Dreier <roland@purestorage.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Use the 32-bit compat keyctl() syscall wrapper on Sparc64 for Sparc32 binary
compatibility.
Without this, keyctl(KEYCTL_INSTANTIATE_IOV) is liable to malfunction as it
uses an iovec array read from userspace - though the kernel should survive this
as it checks pointers and sizes anyway.
I think all the other keyctl() function should just work, provided (a) the top
32-bits of each 64-bit argument register are cleared prior to invoking the
syscall routine, and the 32-bit address space is right at the 0-end of the
64-bit address space. Most of the arguments are 32-bit anyway, and so for
those clearing is not required.
Signed-off-by: David Howells <dhowells@redhat.com
cc: "David S. Miller" <davem@davemloft.net>
cc: sparclinux@vger.kernel.org Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
An out-of-place "OK" response to the "AT+GMR" (get firmware version)
command turns out to be, more often than not, a delayed response to
a previous command rather than an actual error, so continue waiting
for the version number in that case.
Signed-off-by: Tilman Schmidt <tilman@imap.cc> Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.2: adjust indentation] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
If DISCONNECT_B3_IND was synthesized because of a DISCONNECT_REQ
with existing logical connections, the connection state wasn't
updated accordingly. Also the emitted DISCONNECT_B3_IND message
wasn't included in the debug log as requested.
This patch fixes both of these issues.
Signed-off-by: Tilman Schmidt <tilman@imap.cc> Signed-off-by: David S. Miller <davem@davemloft.net>
[bwh: Backported to 3.2: adjust context] Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Introduce a global ratelimit for CAPI message dumps to protect
against possible log flood.
Drop the ratelimit for ignored messages which is now covered by the
global one.
Signed-off-by: Tilman Schmidt <tilman@imap.cc> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The ioc->pfacts member in the IOC structure is getting set to zero
following a call to _base_get_ioc_facts due to the memset in that routine.
So if the ioc->pfacts was read after a host reset, there would be a NULL
pointer dereference. The routine _base_get_ioc_facts is called from context
of host reset. The problem in _base_get_ioc_facts is the size of
Mpi2IOCFactsReply is 64, whereas the sizeof "struct mpt2sas_facts" is 60,
so there is a four byte overflow resulting from the memset.
Also, there is memset in _base_get_port_facts using the incorrect structure,
it should be "struct mpt2sas_port_facts" instead of Mpi2PortFactsReply.
Signed-off-by: Nagalakshmi Nandigama <nagalakshmi.nandigama@lsi.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
It has happened twice now where elaborate troubleshooting has
undergone on systems where CONFIG_CFG80211_INTERNAL_REGDB [0]
has been set but yet net/wireless/db.txt was not updated.
Despite the documentation on this it seems system integrators could
use some more help with this, so throw out a kernel warning at boot time
when their database is empty.
This does mean that the error-prone system integrator won't likely
realize the issue until they boot the machine but -- it does not seem
to make sense to enable a build bug breaking random build testing.
We reset the bool names and values array to NULL, but do not reset the
number of entries in these arrays to 0. If we error out and then get back
into this function we will walk these NULL pointers based on the belief
that they are non-zero length.
Signed-off-by: Eric Paris <eparis@redhat.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Upgraded firmware on Smart Array P7xx (and some others) made them show up as
SCSI revision 5 devices and this caused the driver to fail to map MSA2xxx
logical drives to the correct bus/target/lun. A symptom of this would be that
the target ID of the logical drives as presented by the external storage array
is ignored, and all such logical drives are assigned to target zero,
differentiated only by LUN. Some multipath software reportedly does not deal
well with this behavior, failing to recognize different paths to the same
device as such.
Signed-off-by: Stephen M. Cameron <scameron@beardog.cce.hp.com> Signed-off-by: Scott Teel <scott.teel@hp.com> Signed-off-by: James Bottomley <JBottomley@Parallels.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Some discussion with the glibc mailing lists revealed that this was
necessary for 64-bit platforms with MIPS-like sign-extension rules
for 32-bit values. The original symptom was that passing (uid_t)-1 to
setreuid() was failing in programs linked -pthread because of the "setxid"
mechanism for passing setxid-type function arguments to the syscall code.
SYSCALL_WRAPPERS handles ensuring that all syscall arguments end up with
proper sign-extension and is thus the appropriate fix for this problem.
On other platforms (s390, powerpc, sparc64, and mips) this was fixed
in 2.6.28.6. The general issue is tracked as CVE-2009-0029.
Signed-off-by: Chris Metcalf <cmetcalf@tilera.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Due to an errata, the PA7300LC generates a TLB miss interruption even on the
prefetch instruction. This means that prefetch(NULL), which is supposed to be
a nop on linux actually generates a NULL deref fault. Fix this by testing the
address of prefetch against NULL before doing the prefetch.
Signed-off-by: James Bottomley <JBottomley@Parallels.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
As pointed out by serveral people, PA1.1 only has a type 26 instruction
meaning that the space register must be explicitly encoded. Not giving an
explicit space means that the compiler uses the type 24 version which is PA2.0
only resulting in an illegal instruction crash.
Reported-by: Helge Deller <deller@gmx.de> Signed-off-by: John David Anglin <dave.anglin@bell.net> Signed-off-by: James Bottomley <JBottomley@Parallels.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
because a PA2.0 instruction was accidentally introduced into the PA1.1 TLB
insertion interruption path when it was consolidated with the do_alias macro.
Fix the do_alias macro only to use PA2.0 instructions if compiled for 64 bit. Signed-off-by: James Bottomley <JBottomley@Parallels.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
6d1d8050b4bc8 "block, partition: add partition_meta_info to hd_struct"
added part_unpack_uuid() which assumes that the passed in buffer has
enough space for sprintfing "%pU" - 37 characters including '\0'.
Unfortunately, b5af921ec0233 "init: add support for root devices
specified by partition UUID" supplied 33 bytes buffer to the function
leading to the following panic with stackprotector enabled.
The number of bio_get_nr_vecs() is passed down via bio_alloc() to
bvec_alloc_bs(), which fails the bio allocation if
nr_iovecs > BIO_MAX_PAGES. For the underlying caller this causes an
unexpected bio allocation failure.
Limiting to queue_max_segments() is not sufficient, as max_segments
also might be very large.
There were two places bio_get_nr_vecs() could overflow:
First, it did a left shift to convert from sectors to bytes immediately
before dividing by PAGE_SIZE. If PAGE_SIZE ever was less than 512 a great
many things would break, so dividing by PAGE_SIZE >> 9 is safe and will
generate smaller code too.
The nastier overflow was in the DIV_ROUND_UP() (that's what the code was
effectively doing, anyways). If n + d overflowed, the whole thing would
return 0 which breaks things rather effectively.
bio_get_nr_vecs() doesn't claim to give an exact value anyways, so the
DIV_ROUND_UP() is silly; we could do a straight divide except if a
device's queue_max_sectors was less than PAGE_SIZE we'd return 0. So we
just add 1; this should always be safe - things will break badly if
bio_get_nr_vecs() returns > BIO_MAX_PAGES (bio_alloc() will suddenly start
failing) but it's queue_max_segments that must guard against this, if
queue_max_sectors is preventing this from happen things are going to
explode on architectures with different PAGE_SIZE.
Signed-off-by: Kent Overstreet <koverstreet@google.com> Cc: Tejun Heo <tj@kernel.org> Acked-by: Valdis Kletnieks <Valdis.Kletnieks@vt.edu> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Jens Axboe <axboe@kernel.dk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit 7eef87dc99e419b1cc051e4417c37e4744d7b661 (KVM: s390: fix
register setting) added a load of the floating point control register
to the KVM_SET_FPU path. Lets make sure that the fpc is valid.
Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
In handle_stop() handle the stop bit before doing the store status as
described for "Stop and Store Status" in the Principles of Operation.
We have to give up the local_int.lock before calling kvm store status
since it calls gmap_fault() which might sleep. Since local_int.lock
only protects local_int.* and not guest memory we can give up the lock.
Signed-off-by: Jens Freimann <jfrei@linux.vnet.ibm.com> Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com> Signed-off-by: Avi Kivity <avi@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The code which checks whether to inject a pagefault to L1 or L2 (in
nested VMX) was wrong, incorrect in how it checked the PF_VECTOR bit.
Thanks to Dan Carpenter for spotting this.
Signed-off-by: Nadav Har'El <nyh@il.ibm.com> Reported-by: Dan Carpenter <dan.carpenter@oracle.com> Signed-off-by: Avi Kivity <avi@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
As pointed out by Jason Baron, when assigning a device to a guest
we first set the iommu domain pointer, which enables mapping
and unmapping of memory slots to the iommu. This leaves a window
where this path is enabled, but we haven't synchronized the iommu
mappings to the existing memory slots. Thus a slot being removed
at that point could send us down unexpected code paths removing
non-existent pinnings and iommu mappings. Take the slots_lock
around creating the iommu domain and initial mappings as well as
around iommu teardown to avoid this race.
Signed-off-by: Alex Williamson <alex.williamson@redhat.com> Signed-off-by: Marcelo Tosatti <mtosatti@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
If some vcpus are created before KVM_CREATE_IRQCHIP, then
irqchip_in_kernel() and vcpu->arch.apic will be inconsistent, leading
to potential NULL pointer dereferences.
Fix by:
- ensuring that no vcpus are installed when KVM_CREATE_IRQCHIP is called
- ensuring that a vcpu has an apic if it is installed after KVM_CREATE_IRQCHIP
This is somewhat long winded because vcpu->arch.apic is created without
kvm->lock held.
Based on earlier patch by Michael Ellerman.
Signed-off-by: Michael Ellerman <michael@ellerman.id.au> Signed-off-by: Avi Kivity <avi@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
commit c57b5468406 (pktgen: fix crash at module unload) did a very poor
job with list primitives.
1) list_splice() arguments were in the wrong order
2) list_splice(list, head) has undefined behavior if head is not
initialized.
3) We should use the list_splice_init() variant to clear pktgen_threads
list.
Signed-off-by: Eric Dumazet <edumazet@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Hold the pktgen_thread_lock while splicing pktgen_threads, and test
pktgen_exiting in pktgen_device_event() to make unload faster.
Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Eric W. Biederman <ebiederm@xmission.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
CC drivers/net/ethernet/stmicro/stmmac/mmc_core.o
drivers/net/ethernet/stmicro/stmmac/mmc_core.c: In function 'dwmac_mmc_ctrl':
drivers/net/ethernet/stmicro/stmmac/mmc_core.c:143:2: error: implicit
declaration of function 'pr_debug' [-Werror=implicit-function-declaration]
Signed-off-by: Stefan Roese <sr@denx.de> Cc: Giuseppe Cavallaro <peppe.cavallaro@st.com> Acked-by: Giuseppe Cavallaro <peppe.cavallaro@st.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This patch fixes the following build failure:
In file included from include/linux/mtd/qinfo.h:4:0,
from include/linux/mtd/pfow.h:7,
from drivers/mtd/lpddr/lpddr_cmds.c:27:
include/linux/mtd/map.h: In function 'inline_map_read':
include/linux/mtd/map.h:409:3: error: implicit declaration of function 'BUILD_BUG_ON' [-Werror=implicit-function-declaration]
Signed-off-by: Artem Bityutskiy <artem.bityutskiy@linux.intel.com> Signed-off-by: David Woodhouse <David.Woodhouse@intel.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Killing reset task while adapter is resetting causes deadlock.
Only kill reset task if adapter is not resetting.
Ref bug #43132 on bugzilla.kernel.org
Signed-off-by: Tushar Dave <tushar.n.dave@intel.com> Tested-by: Aaron Brown <aaron.f.brown@intel.com> Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Since recent changes on TCP splicing (starting with commits 2f533844
"tcp: allow splice() to build full TSO packets" and 35f9c09f "tcp:
tcp_sendpages() should call tcp_push() once"), I started seeing
massive stalls when forwarding traffic between two sockets using
splice() when pipe buffers were larger than socket buffers.
Latest changes (net: netdev_alloc_skb() use build_skb()) made the
problem even more apparent.
The reason seems to be that if do_tcp_sendpages() fails on out of memory
condition without being able to send at least one byte, tcp_push() is not
called and the buffers cannot be flushed.
After applying the attached patch, I cannot reproduce the stalls at all
and the data rate it perfectly stable and steady under any condition
which previously caused the problem to be permanent.
The issue seems to have been there since before the kernel migrated to
git, which makes me think that the stalls I occasionally experienced
with tux during stress-tests years ago were probably related to the
same issue.
This issue was first encountered on 3.0.31 and 3.2.17, so please backport
to -stable.
Signed-off-by: Willy Tarreau <w@1wt.eu> Acked-by: Eric Dumazet <edumazet@google.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This patch fixes a bug in the handling of FILEIO w/ underlying block_device
resize operations where the original fd_dev->fd_dev_size was incorrectly being
used in fd_get_blocks() for READ_CAPACITY response payloads.
This patch avoids using fd_dev->fd_dev_size for FILEIO devices with
an underlying block_device, and instead changes fd_get_blocks() to
get the sector count directly from i_size_read() as recommended by hch.
Reported-by: Christoph Hellwig <hch@lst.de> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Use del_timer_sync to remove timer before mddev_suspend finishes.
We don't want a timer going off after an mddev_suspend is called. This is
especially true with device-mapper, since it can call the destructor function
immediately following a suspend. This results in the removal (kfree) of the
structures upon which the timer depends - resulting in a very ugly panic.
Therefore, we add a del_timer_sync to mddev_suspend to prevent this.
Signed-off-by: NeilBrown <neilb@suse.de> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Reported-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Needs to be tagged with FLAG_WWAN, which since it has generic
descriptors, won't happen if we don't override the generic
driver info.
Cc: Oliver Neukum <oliver@neukum.org> Signed-off-by: Dan Williams <dcbw@redhat.com> Acked-by: Oliver Neukum <oliver@neukum.org> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Commit 4231d47e6fe69f061f96c98c30eaf9fb4c14b96d(net/usbnet: avoid
recursive locking in usbnet_stop()) fixes the recursive locking
problem by releasing the skb queue lock before unlink, but may
cause skb traversing races:
- after URB is unlinked and the queue lock is released,
the refered skb and skb->next may be moved to done queue,
even be released
- in skb_queue_walk_safe, the next skb is still obtained
by next pointer of the last skb
- so maybe trigger oops or other problems
This patch extends the usage of entry->state to describe 'start_unlink'
state, so always holding the queue(rx/tx) lock to change the state if
the referd skb is in rx or tx queue because we need to know if the
refered urb has been started unlinking in unlink_urbs.
The other part of this patch is based on Huajun's patch:
always traverse from head of the tx/rx queue to get skb which is
to be unlinked but not been started unlinking.
Signed-off-by: Huajun Li <huajun.li.lee@gmail.com> Signed-off-by: Ming Lei <tom.leiming@gmail.com> Cc: Oliver Neukum <oneukum@suse.de> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
BIOS on some HP laptops don't set the speaker-pins as fixed but expose
as jacks, and this confuses the driver as if these pins are
jack-detectable. As a result, the machine doesn't get sounds from
speakers because the driver prepares the power-map update via jack
unsol events which never come up in reality. The bug was introduced
in some time in 3.2 for enabling the power-mapping feature.
This patch fixes the problem by replacing the check of the persistent
power-map bits with a proper is_jack_detectable() call.
Without CRYPTO_HASH being selected, mv_cesa has a lot of hooks
into undefined exports.
----
MODPOST 81 modules
Kernel: arch/arm/boot/Image is ready
AS arch/arm/boot/compressed/head.o
GZIP arch/arm/boot/compressed/piggy.gzip
CC arch/arm/boot/compressed/misc.o
CC arch/arm/boot/compressed/decompress.o
ERROR: "crypto_ahash_type" [drivers/crypto/mv_cesa.ko] undefined!
ERROR: "crypto_shash_final" [drivers/crypto/mv_cesa.ko] undefined!
ERROR: "crypto_register_ahash" [drivers/crypto/mv_cesa.ko] undefined!
ERROR: "crypto_unregister_ahash" [drivers/crypto/mv_cesa.ko] undefined!
ERROR: "crypto_shash_update" [drivers/crypto/mv_cesa.ko] undefined!
ERROR: "crypto_shash_digest" [drivers/crypto/mv_cesa.ko] undefined!
ERROR: "crypto_shash_setkey" [drivers/crypto/mv_cesa.ko] undefined!
ERROR: "crypto_alloc_shash" [drivers/crypto/mv_cesa.ko] undefined!
make[1]: *** [__modpost] Error 1
make: *** [modules] Error 2
make: *** Waiting for unfinished jobs....
----
Signed-off-by: Alexander Clouter <alex@digriz.org.uk> Signed-off-by: Jason Cooper <jason@lakedaemon.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This patch addresses a bug in a special case for target core SPC-2 RELEASE
logic where the same physical client (eg: iSCSI InitiatorName) with
differing iSCSI session identifiers (ISID) is allowed to incorrectly release
the same client's SPC-2 reservation from the non reservation holding path.
Note this bug is specific to iscsi-target w/ SPC-2 reservations, and
with the default enforce_pr_isids=1 device attr setting in target-core
controls if a InitiatorName + different ISID reservations are handled
the same as a single iSCSI client entity.
Signed-off-by: Bernhard Kohl <bernhard.kohl@gmx.net> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The vfp_enable function enables access to the VFP co-processor register
space (cp10 and cp11) on the current CPU and must be called with
preemption disabled. Unfortunately, the vfp_init late initcall does not
disable preemption and can lead to an oops during boot if thread
migration occurs at the wrong time and we end up attempting to access
the FPSID on a CPU with VFP access disabled.
This patch fixes the initcall to call vfp_enable from a non-preemptible
context on each CPU and adds a BUG_ON(preemptible) to ensure that any
similar problems are easily spotted in the future.
Reported-by: Hyungwoo Yang <hwoo.yang@gmail.com> Signed-off-by: Hyungwoo Yang <hyungwooy@nvidia.com> Signed-off-by: Will Deacon <will.deacon@arm.com> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The A-MPDU code checked against a retry limit, but it was using
the wrong variable to do so. This patch fixes this to assure
proper retry mechanism.
This problem had a side-effect causing the mac80211 flush callback
to remain waiting forever as well. That side effect has been fixed
by commit by Stanislaw Gruszka:
While debugging udev > 170 failure on Debian Wheezy
(http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=648325), it appears
that the issue was in fact due to missing accept4() in ia64.
This patch simply adds accept4() to ia64.
Signed-off-by: Émeric Maschino <emeric.maschino@gmail.com> Signed-off-by: Tony Luck <tony.luck@intel.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Processes hang forever on a sync-mounted ext2 file system that
is mounted with the ext4 module (default in Fedora 16).
I can reproduce this reliably by mounting an ext2 partition with
"-o sync" and opening a new file an that partition with vim. vim
will hang in "D" state forever. The same happens on ext4 without
a journal.
I am attaching a small patch here that solves this issue for me.
In the sync mounted case without a journal,
ext4_handle_dirty_metadata() may call sync_dirty_buffer(), which
can't be called with buffer lock held.
Also move mb_cache_entry_release inside lock to avoid race
fixed previously by 8a2bfdcb ext[34]: EA block reference count racing fix
Note too that ext2 fixed this same problem in 2006 with b2f49033 [PATCH] fix deadlock in ext2
Signed-off-by: Martin.Wilck@ts.fujitsu.com
[sandeen@redhat.com: move mb_cache_entry_release before unlock, edit commit msg] Signed-off-by: Eric Sandeen <sandeen@redhat.com> Signed-off-by: "Theodore Ts'o" <tytso@mit.edu> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Currently, when spi-topcliff-pch receives transmit request over 4KByte,
this driver can't process correctly. This driver needs to divide the data
into 4Kbyte unit.
This patch fixes the issue.
Signed-off-by: Tomoya MORINAGA <tomoya.rohm@gmail.com> Signed-off-by: Grant Likely <grant.likely@secretlab.ca> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This bug was introduced by commit 54be5663
"gpio-ml-ioh: Support interrupt function" which adds a spinlock to struct
ioh_gpio but never init the spinlock.
Signed-off-by: Axel Lin <axel.lin@gmail.com> Signed-off-by: Grant Likely <grant.likely@secretlab.ca> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
%g2 is meant to hold the CPUID number throughout this routine, since
at the very beginning, and at the very end, we use %g2 to calculate
indexes into per-cpu arrays.
However we erroneously clobber it in order to hold the %cwp register
value mid-stream.
Fix this code to use %g3 for the %cwp read and related calulcations
instead.
Reported-by: Meelis Roos <mroos@linux.ee> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
If the requested scsi_dh module is already loaded then skip
request_module().
Multipath table loads can hang in an unnecessary __request_module.
Reported-by: Ben Marzinski <bmarzins@redhat.com> Signed-off-by: Mike Snitzer <snitzer@redhat.com> Signed-off-by: Alasdair G Kergon <agk@redhat.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When setting TRY crop on the sub-device the mutex was erroneously acquired
rather than released on exit path. This bug is present in kernels starting
from v3.2.
Signed-off-by: Sylwester Nawrocki <s.nawrocki@samsung.com> Signed-off-by: Kyungmin Park <kyungmin.park@samsung.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
The locking policy is such that the erase_complete_block spinlock is
nested within the alloc_sem mutex. This fixes a case in which the
acquisition order was erroneously reversed. This issue was caught by
the following lockdep splat:
=======================================================
[ INFO: possible circular locking dependency detected ]
3.0.5 #1
-------------------------------------------------------
jffs2_gcd_mtd6/299 is trying to acquire lock:
(&c->alloc_sem){+.+.+.}, at: [<c01f7714>] jffs2_garbage_collect_pass+0x314/0x890
but task is already holding lock:
(&(&c->erase_completion_lock)->rlock){+.+...}, at: [<c01f7708>] jffs2_garbage_collect_pass+0x308/0x890
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
Some RNDIS devices include a bogus CDC Union descriptor pointing
to non-existing interfaces. The RNDIS code is already prepared
to handle devices without a CDC Union descriptor by hardwiring
the driver to use interfaces 0 and 1, which is correct for the
devices with the bogus descriptor as well. So we can reuse the
existing workaround.
Cc: Markus Kolb <linux-201011@tower-net.de> Cc: Iker Salmón San Millán <shaola@esdebian.org> Cc: Jonathan Nieder <jrnieder@gmail.com> Cc: Oliver Neukum <oliver@neukum.org> Cc: 655387@bugs.debian.org Signed-off-by: Bjørn Mork <bjorn@mork.no> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
An early registration of an ISR was causing a crash to several users (for
example, with the ite-cir driver: http://bugs.launchpad.net/bugs/972723).
The reason was that IRQs were being triggered before a driver
initialisation was completed.
This patch fixes this by moving the invocation to request_irq() and to
request_region() to a later stage on the driver probe function.
Signed-off-by: Luis Henriques <luis.henriques@canonical.com> Acked-by: Jarod Wilson <jarod@redhat.com> Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
One of the OLPC changes lost a little in its translation to mainline,
leading to build errors on the ARM architecture. Remove the offending
line, and all will be well.
Reported-by: Mathieu Poirier <mathieu.poirier@linaro.org> Signed-off-by: Jonathan Corbet <corbet@lwn.net> Signed-off-by: Mauro Carvalho Chehab <mchehab@redhat.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This patch removes some potentially problematic legacy code within
core_clear_initiator_node_from_tpg() that was originally intended to
release left over se_lun_acl setup during dynamic NodeACL+MappedLUN
generate when running with TPG demo-mode operation.
Since we now only ever expect to allocate and release se_lun_acl from
within target_core_fabric_configfs.c:target_fabric_make_mappedlun() and
target_fabric_drop_mappedlun() context respectively, this code for
demo-mode release is incorrect and needs to be removed.
Cc: Christoph Hellwig <hch@lst.de> Cc: Andy Grover <agrover@redhat.com> Signed-off-by: Nicholas Bellinger <nab@linux-iscsi.org> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
NFSv4: Save the owner/group name string when doing open
...so that we can do the uid/gid mapping outside the asynchronous RPC
context.
This fixes a bug in the current NFSv4 atomic open code where the client
isn't able to determine what the true uid/gid fields of the file are,
(because the asynchronous nature of the OPEN call denies it the ability
to do an upcall) and so fills them with default values, marking the
inode as needing revalidation.
Unfortunately, in some cases, the VFS will do some additional sanity
checks on the file, and may override the server's decision to allow
the open because it sees the wrong owner/group fields.
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
Without this patch, logging into two different machines with home
directories mounted over NFS4 and then running "vim" and typing ":q"
in each reliably produces the following error on the second machine:
E137: Viminfo file is not writable: /users/system/rtheys/.viminfo
This regression was introduced by 80e52aced138 ("NFSv4: Don't do
idmapper upcalls for asynchronous RPC calls", merged during the 2.6.32
cycle) --- after the OPEN call, .viminfo has the default values for
st_uid and st_gid (0xfffffffe) cached because we do not want to let
rpciod wait for an idmapper upcall to fill them in.
The fix used in mainline is to save the owner and group as strings and
perform the upcall in _nfs4_proc_open outside the rpciod context,
which takes about 600 lines. For stable, we can do something similar
with a one-liner: make open check for the stale fields and make a
(synchronous) GETATTR call to fill them when needed.
Trond dictated the patch, I typed it in, and Rik tested it.
Addresses http://bugs.debian.org/659111 and
https://bugzilla.redhat.com/789298
Reported-by: Rik Theys <Rik.Theys@esat.kuleuven.be> Explained-by: David Flyn <davidf@rd.bbc.co.uk> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Tested-by: Rik Theys <Rik.Theys@esat.kuleuven.be> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
This patch fixes Auto Power Saving configuration in ip101a_config_init
which was broken as there is no phy register write followed after
setting IP101A_APS_ON flag.
This patch also fixes the return value of ip101a_config_init.
Without this patch ip101a_config_init returns 2 which is not an error
accroding to IS_ERR and the mac driver will continue accessing 2 as
valid pointer to phy_dev resulting in memory fault.
Signed-off-by: Srinivas Kandagatla <srinivas.kandagatla@st.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
Jonathan Nieder [Sat, 12 May 2012 00:35:33 +0000 (19:35 -0500)]
brcm80211: smac: pass missing argument to 'brcms_b_mute'
[Not needed upstream --- this bug is specific to 3.2.y.]
Commit c6c44893c864, which removes the flag argument from brcms_b_mute,
is not part of 3.2.y, and we forgot to adjust a new call accordingly
when applying commit badc4f07622f ("brcm80211: smac: resume transmit
fifo upon receiving frames").
drivers/net/wireless/brcm80211/brcmsmac/main.c: In function 'brcms_c_recvctl':
drivers/net/wireless/brcm80211/brcmsmac/main.c:7882:4: error: too few arguments to function 'brcms_b_mute'
drivers/net/wireless/brcm80211/brcmsmac/main.c:2538:13: note: declared here
Earlier build tests missed this because they didn't include this driver
due to 'depends on BCMA=n'.
Reported-by: Ben Hutchings <ben@decadent.org.uk> Signed-off-by: Jonathan Nieder <jrnieder@gmail.com> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
tcp_adv_win_scale default value is 2, meaning we expect a good citizen
skb to have skb->len / skb->truesize ratio of 75% (3/4)
In 2.6 kernels we (mis)accounted for typical MSS=1460 frame :
1536 + 64 + 256 = 1856 'estimated truesize', and 1856 * 3/4 = 1392.
So these skbs were considered as not bloated.
With recent truesize fixes, a typical MSS=1460 frame truesize is now the
more precise :
2048 + 256 = 2304. But 2304 * 3/4 = 1728.
So these skb are not good citizen anymore, because 1460 < 1728
(GRO can escape this problem because it build skbs with a too low
truesize.)
This also means tcp advertises a too optimistic window for a given
allocated rcvspace : When receiving frames, sk_rmem_alloc can hit
sk_rcvbuf limit and we call tcp_prune_queue()/tcp_collapse() too often,
especially when application is slow to drain its receive queue or in
case of losses (netperf is fast, scp is slow). This is a major latency
source.
We should adjust the len/truesize ratio to 50% instead of 75%
This patch :
1) changes tcp_adv_win_scale default to 1 instead of 2
2) increase tcp_rmem[2] limit from 4MB to 6MB to take into account
better truesize tracking and to allow autotuning tcp receive window to
reach same value than before. Note that same amount of kernel memory is
consumed compared to 2.6 kernels.
Signed-off-by: Eric Dumazet <edumazet@google.com> Cc: Neal Cardwell <ncardwell@google.com> Cc: Tom Herbert <therbert@google.com> Cc: Yuchung Cheng <ycheng@google.com> Acked-by: Neal Cardwell <ncardwell@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
When the cwnd reduction is done, ssthresh may be infinite
if TCP enters CWR via ECN or F-RTO. If cwnd is not undone, i.e.,
undo_marker is set, tcp_complete_cwr() falsely set cwnd to the
infinite ssthresh value. The correct operation is to keep cwnd
intact because it has been updated in ECN or F-RTO.
Signed-off-by: Yuchung Cheng <ycheng@google.com> Acked-by: Neal Cardwell <ncardwell@google.com> Signed-off-by: David S. Miller <davem@davemloft.net> Signed-off-by: Ben Hutchings <ben@decadent.org.uk>
projects / karo-tx-linux.git / log