pandora-kernel.git
10 years agosock_diag: fix filter code sent to userspace
Nicolas Dichtel [Wed, 5 Jun 2013 13:30:55 +0000 (15:30 +0200)]
sock_diag: fix filter code sent to userspace

Filters need to be translated to real BPF code for userland, like SO_GETFILTER.

Signed-off-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotuntap: fix a possible race between queue selection and changing queues
Jason Wang [Wed, 5 Jun 2013 08:44:57 +0000 (16:44 +0800)]
tuntap: fix a possible race between queue selection and changing queues

Complier may generate codes that re-read the tun->numqueues during
tun_select_queue(). This may be a race if vlan->numqueues were changed in the
same time and can lead unexpected result (e.g. very huge value).

We need prevent the compiler from generating such codes by adding an
ACCESS_ONCE() to make sure tun->numqueues were only read once.

Bug were introduced by commit c8d68e6be1c3b242f1c598595830890b65cea64a
(tuntap: multiqueue support).

Reported-by: Michael S. Tsirkin <mst@redhat.com>
Cc: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agovhost_net: clear msg.control for non-zerocopy case during tx
Jason Wang [Wed, 5 Jun 2013 07:40:46 +0000 (15:40 +0800)]
vhost_net: clear msg.control for non-zerocopy case during tx

When we decide not use zero-copy, msg.control should be set to NULL otherwise
macvtap/tap may set zerocopy callbacks which may decrease the kref of ubufs
wrongly.

Bug were introduced by commit cedb9bdce099206290a2bdd02ce47a7b253b6a84
(vhost-net: skip head management if no outstanding).

This solves the following warnings:

WARNING: at include/linux/kref.h:47 handle_tx+0x477/0x4b0 [vhost_net]()
Modules linked in: vhost_net macvtap macvlan tun nfsd exportfs bridge stp llc openvswitch kvm_amd kvm bnx2 megaraid_sas [last unloaded: tun]
CPU: 5 PID: 8670 Comm: vhost-8668 Not tainted 3.10.0-rc2+ #1566
Hardware name: Dell Inc. PowerEdge R715/00XHKG, BIOS 1.5.2 04/19/2011
ffffffffa0198323 ffff88007c9ebd08 ffffffff81796b73 ffff88007c9ebd48
ffffffff8103d66b 000000007b773e20 ffff8800779f0000 ffff8800779f43f0
ffff8800779f8418 000000000000015c 0000000000000062 ffff88007c9ebd58
Call Trace:
[<ffffffff81796b73>] dump_stack+0x19/0x1e
[<ffffffff8103d66b>] warn_slowpath_common+0x6b/0xa0
[<ffffffff8103d6b5>] warn_slowpath_null+0x15/0x20
[<ffffffffa0197627>] handle_tx+0x477/0x4b0 [vhost_net]
[<ffffffffa0197690>] handle_tx_kick+0x10/0x20 [vhost_net]
[<ffffffffa019541e>] vhost_worker+0xfe/0x1a0 [vhost_net]
[<ffffffffa0195320>] ? vhost_attach_cgroups_work+0x30/0x30 [vhost_net]
[<ffffffffa0195320>] ? vhost_attach_cgroups_work+0x30/0x30 [vhost_net]
[<ffffffff81061f46>] kthread+0xc6/0xd0
[<ffffffff81061e80>] ? kthread_freezable_should_stop+0x70/0x70
[<ffffffff817a1aec>] ret_from_fork+0x7c/0xb0
[<ffffffff81061e80>] ? kthread_freezable_should_stop+0x70/0x70

Signed-off-by: Jason Wang <jasowang@redhat.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/pablo/nf
David S. Miller [Mon, 10 Jun 2013 20:30:33 +0000 (13:30 -0700)]
Merge branch 'master' of git://git./linux/kernel/git/pablo/nf

Pablo Neira Ayuso says:

====================
The following patchset contains four fixes for Netfilter and one fix
for IPVS, they are:

* Fix data leak to user-space via getsockopt IP_VS_SO_GET_DESTS, from
  Dan Carpenter.

* Fix xt_TCPMSS if no TCP MSS is specified in syn packets, to avoid the
  violation of RFC879, from Phil Oester.

* Fix incomplete dump of objects via nfnetlink_acct and nfnetlink_cttimeout,
  from myself.

* Fix missing HW protocol in packets passed to user-space via NFQUEUE,
  from myself.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipvs: info leak in __ip_vs_get_dest_entries()
Dan Carpenter [Mon, 3 Jun 2013 09:00:49 +0000 (12:00 +0300)]
ipvs: info leak in __ip_vs_get_dest_entries()

The entry struct has a 2 byte hole after ->port and another 4 byte
hole after ->stats.outpkts.  You must have CAP_NET_ADMIN in your
namespace to hit this information leak.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Signed-off-by: Simon Horman <horms@verge.net.au>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonet_sched: qdisc_get_rtab() must check data[] array
Eric Dumazet [Sun, 2 Jun 2013 11:15:55 +0000 (11:15 +0000)]
net_sched: qdisc_get_rtab() must check data[] array

qdisc_get_rtab() should check not only the keys in struct tc_ratespec,
but also the full data[] array.

"tc ... linklayer atm " only perturbs values in the 256 slots array.

Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonetfilter: nfnetlink_queue: fix missing HW protocol
Pablo Neira Ayuso [Fri, 7 Jun 2013 16:42:00 +0000 (18:42 +0200)]
netfilter: nfnetlink_queue: fix missing HW protocol

Locally generated IPv4 and IPv6 traffic gets skb->protocol unset,
thus passing zero.

ip6tables -I OUTPUT -j NFQUEUE
libmnl/examples/netfilter# ./nf-queue 0 &
ping6 ::1
packet received (id=1 hw=0x0000 hook=3)
                         ^^^^^^

Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Fri, 7 Jun 2013 01:09:05 +0000 (18:09 -0700)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fix from David Miller:
 "This is a quick one commit pull request to cure the regression
  introduced by the MSG_CMSG_COMPAT change."

(Background: commit 1be374a0518a completely broke 32-bit COMPAT handling
by not only disallowing MSG_CMSG_COMPAT from user APIs, but clearing it
in our own internal use too!)

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net:
  net: Unbreak compat_sys_{send,recv}msg

10 years agoMerge tag 'staging-3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Thu, 6 Jun 2013 23:34:11 +0000 (16:34 -0700)]
Merge tag 'staging-3.10-rc4' of git://git./linux/kernel/git/gregkh/staging

Pull staging driver fixes from Greg Kroah-Hartman:
 "Here are some staging and IIO driver fixes for the 3.10-rc5 release.

  All of them are tiny, and fix a number of reported issues (build and
  runtime)"

Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
* tag 'staging-3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/staging:
  iio:inkern: Fix typo/bug in convert raw to processed.
  iio: frequency: ad4350: Fix bug / typo in mask
  inkern: iio_device_put after incorrect return/goto
  staging: alarm-dev: information leak in alarm_compat_ioctl()
  iio:callback buffer: free the scan_mask
  staging: alarm-dev: information leak in alarm_ioctl()
  drivers: staging: zcache: fix compile error
  staging: dwc2: fix value of dma_mask

10 years agoMerge tag 'tty-3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty
Linus Torvalds [Thu, 6 Jun 2013 23:33:35 +0000 (16:33 -0700)]
Merge tag 'tty-3.10-rc4' of git://git./linux/kernel/git/gregkh/tty

Pull tty/serial driver fixes from Greg Kroah-Hartman:
 "Here are some small bugfixes, and one revert, of serial driver issues
  that have been reported"

* tag 'tty-3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/tty:
  Revert "serial: 8250: Make SERIAL_8250_RUNTIME_UARTS work correctly"
  serial: samsung: enable clock before clearing pending interrupts during init
  serial/imx: disable hardware flow control at startup

10 years agoMerge tag 'usb-3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Thu, 6 Jun 2013 23:29:17 +0000 (16:29 -0700)]
Merge tag 'usb-3.10-rc4' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg Kroah-Hartman:
 "Here are a number of USB bugfixes and new device ids for the 3.10-rc5
  tree.

  Nothing major here, a number of new device ids (and movement from the
  option to the zte_ev driver of a number of ids that we had previously
  gotten wrong, some xhci bugfixes, some usb-serial driver fixes that
  were recently found, some host controller fixes / reverts, and a
  variety of smaller other things"

* tag 'usb-3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb: (29 commits)
  USB: option,zte_ev: move most ZTE CDMA devices to zte_ev
  USB: option: blacklist network interface on Huawei E1820
  USB: whiteheat: fix broken port configuration
  USB: serial: fix TIOCMIWAIT return value
  USB: mos7720: fix hardware flow control
  USB: keyspan: remove unused endpoint-array access
  USB: keyspan: fix bogus array index
  USB: zte_ev: fix broken open
  USB: serial: Add Option GTM681W to qcserial device table.
  USB: Serial: cypress_M8: Enable FRWD Dongle hidcom device
  USB: EHCI: fix regression related to qh_refresh()
  usbfs: Increase arbitrary limit for USB 3 isopkt length
  USB: zte_ev: fix control-message timeouts
  USB: mos7720: fix message timeouts
  USB: iuu_phoenix: fix bulk-message timeout
  USB: ark3116: fix control-message timeout
  USB: mos7840: fix DMA to stack
  USB: mos7720: fix DMA to stack
  USB: visor: fix initialisation of Treo/Kyocera devices
  USB: serial: fix Treo/Kyocera interrrupt-in urb context
  ...

10 years agoMerge tag 'pci-v3.10-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaa...
Linus Torvalds [Thu, 6 Jun 2013 23:28:15 +0000 (16:28 -0700)]
Merge tag 'pci-v3.10-fixes-3' of git://git./linux/kernel/git/helgaas/pci

Pull PCI fixes from Bjorn Helgaas:
 "This fixes a crash when booting a 32-bit kernel via the EFI boot stub.

  PCI ROM from EFI
      x86/PCI: Map PCI setup data with ioremap() so it can be in highmem"

* tag 'pci-v3.10-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
  x86/PCI: Map PCI setup data with ioremap() so it can be in highmem

10 years agoMerge tag 'for-linus-v3.10-rc5' of git://oss.sgi.com/xfs/xfs
Linus Torvalds [Thu, 6 Jun 2013 23:15:25 +0000 (16:15 -0700)]
Merge tag 'for-linus-v3.10-rc5' of git://oss.sgi.com/xfs/xfs

Pull more xfs updates from Ben Myers:
 "Here are several fixes for filesystems with CRC support turned on:
  fixes for quota, remote attributes, and recovery.  There is also some
  feature work related to CRCs: the implementation of CRCs for the inode
  unlinked lists, disabling noattr2/attr2 options when appropriate, and
  bumping the maximum number of ACLs.

  I would have preferred to defer this last category of items to 3.11.
  This would require setting a feature bit for the on-disk changes, so
  there is some pressure to get these in 3.10.  I believe this
  represents the end of the CRC related queue.

   - Rework of dquot CRCs
   - Fix for remote attribute invalidation of a leaf
   - Fix ordering of transaction replay in recovery
   - Implement CRCs for inode unlinked list
   - Disable noattr2/attr2 mount options when CRCs are enabled
   - Bump the limitation of ACL entries for v5 superblocks"

* tag 'for-linus-v3.10-rc5' of git://oss.sgi.com/xfs/xfs:
  xfs: increase number of ACL entries for V5 superblocks
  xfs: disable noattr2/attr2 mount options for CRC enabled filesystems
  xfs: inode unlinked list needs to recalculate the inode CRC
  xfs: fix log recovery transaction item reordering
  xfs: fix remote attribute invalidation for a leaf
  xfs: rework dquot CRCs

10 years agonet: Unbreak compat_sys_{send,recv}msg
Andy Lutomirski [Wed, 5 Jun 2013 19:38:26 +0000 (19:38 +0000)]
net: Unbreak compat_sys_{send,recv}msg

I broke them in this commit:

    commit 1be374a0518a288147c6a7398792583200a67261
    Author: Andy Lutomirski <luto@amacapital.net>
    Date:   Wed May 22 14:07:44 2013 -0700

        net: Block MSG_CMSG_COMPAT in send(m)msg and recv(m)msg

This patch adds __sys_sendmsg and __sys_sendmsg as common helpers that accept
MSG_CMSG_COMPAT and blocks MSG_CMSG_COMPAT at the syscall entrypoints.  It
also reverts some unnecessary checks in sys_socketcall.

Apparently I was suffering from underscore blindness the first time around.

Signed-off-by: Andy Lutomirski <luto@amacapital.net>
Tested-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoUSB: option,zte_ev: move most ZTE CDMA devices to zte_ev
Dan Williams [Wed, 5 Jun 2013 20:26:27 +0000 (15:26 -0500)]
USB: option,zte_ev: move most ZTE CDMA devices to zte_ev

Per some ZTE Linux drivers I found for the AC2716, the following patch
moves most ZTE CDMA devices from option to zte_ev.  The blacklist stuff
that option does is not required with zte_ev, because it doesn't
implement any of the send_setup hooks which the blacklist suppressed.

I did not move the 2718 over because I could not find any ZTE Linux
drivers for that device, nor even any Windows drivers.

Signed-off-by: Dan Williams <dcbw@redhat.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: option: blacklist network interface on Huawei E1820
Bjørn Mork [Thu, 6 Jun 2013 10:57:24 +0000 (12:57 +0200)]
USB: option: blacklist network interface on Huawei E1820

The mode used by Windows for the Huawei E1820 will use the
same ff/ff/ff class codes for both serial and network
functions.

Reported-by: Graham Inggs <graham.inggs@uct.ac.za>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: whiteheat: fix broken port configuration
Johan Hovold [Thu, 6 Jun 2013 11:32:47 +0000 (13:32 +0200)]
USB: whiteheat: fix broken port configuration

When configuring the port (e.g. set_termios) the port minor number
rather than the port number was used in the request (and they only
coincide for minor number 0).

Cc: stable@vger.kernel.org
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoxfs: increase number of ACL entries for V5 superblocks
Dave Chinner [Wed, 5 Jun 2013 02:09:10 +0000 (12:09 +1000)]
xfs: increase number of ACL entries for V5 superblocks

The limit of 25 ACL entries is arbitrary, but baked into the on-disk
format.  For version 5 superblocks, increase it to the maximum nuber
of ACLs that can fit into a single xattr.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Mark Tinguely <tinuguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 5c87d4bc1a86bd6e6754ac3d6e111d776ddcfe57)

10 years agoxfs: disable noattr2/attr2 mount options for CRC enabled filesystems
Dave Chinner [Wed, 5 Jun 2013 02:09:09 +0000 (12:09 +1000)]
xfs: disable noattr2/attr2 mount options for CRC enabled filesystems

attr2 format is always enabled for v5 superblock filesystems, so the
mount options to enable or disable it need to be cause mount errors.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit d3eaace84e40bf946129e516dcbd617173c1cf14)

10 years agoxfs: inode unlinked list needs to recalculate the inode CRC
Dave Chinner [Wed, 5 Jun 2013 02:09:08 +0000 (12:09 +1000)]
xfs: inode unlinked list needs to recalculate the inode CRC

The inode unlinked list manipulations operate directly on the inode
buffer, and so bypass the inode CRC calculation mechanisms. Hence an
inode on the unlinked list has an invalid CRC. Fix this by
recalculating the CRC whenever we modify an unlinked list pointer in
an inode, ncluding during log recovery. This is trivial to do and
results in  unlinked list operations always leaving a consistent
inode in the buffer.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 0a32c26e720a8b38971d0685976f4a7d63f9e2ef)

10 years agoxfs: fix log recovery transaction item reordering
Dave Chinner [Wed, 5 Jun 2013 02:09:07 +0000 (12:09 +1000)]
xfs: fix log recovery transaction item reordering

There are several constraints that inode allocation and unlink
logging impose on log recovery. These all stem from the fact that
inode alloc/unlink are logged in buffers, but all other inode
changes are logged in inode items. Hence there are ordering
constraints that recovery must follow to ensure the correct result
occurs.

As it turns out, this ordering has been working mostly by chance
than good management. The existing code moves all buffers except
cancelled buffers to the head of the list, and everything else to
the tail of the list. The problem with this is that is interleaves
inode items with the buffer cancellation items, and hence whether
the inode item in an cancelled buffer gets replayed is essentially
left to chance.

Further, this ordering causes problems for log recovery when inode
CRCs are enabled. It typically replays the inode unlink buffer long before
it replays the inode core changes, and so the CRC recorded in an
unlink buffer is going to be invalid and hence any attempt to
validate the inode in the buffer is going to fail. Hence we really
need to enforce the ordering that the inode alloc/unlink code has
expected log recovery to have since inode chunk de-allocation was
introduced back in 2003...

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Mark Tinguely <tinguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit a775ad778073d55744ed6709ccede36310638911)

10 years agoxfs: fix remote attribute invalidation for a leaf
Dave Chinner [Mon, 3 Jun 2013 05:28:49 +0000 (15:28 +1000)]
xfs: fix remote attribute invalidation for a leaf

When invalidating an attribute leaf block block, there might be
remote attributes that it points to. With the recent rework of the
remote attribute format, we have to make sure we calculate the
length of the attribute correctly. We aren't doing that in
xfs_attr3_leaf_inactive(), so fix it.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Mark Tinguely <tinuguely@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 59913f14dfe8eb772ff93eb442947451b4416329)

10 years agoxfs: rework dquot CRCs
Dave Chinner [Mon, 3 Jun 2013 05:28:46 +0000 (15:28 +1000)]
xfs: rework dquot CRCs

Calculating dquot CRCs when the backing buffer is written back just
doesn't work reliably. There are several places which manipulate
dquots directly in the buffers, and they don't calculate CRCs
appropriately, nor do they always set the buffer up to calculate
CRCs appropriately.

Firstly, if we log a dquot buffer (e.g. during allocation) it gets
logged without valid CRC, and so on recovery we end up with a dquot
that is not valid.

Secondly, if we recover/repair a dquot, we don't have a verifier
attached to the buffer and hence CRCs are not calculated on the way
down to disk.

Thirdly, calculating the CRC after we've changed the contents means
that if we re-read the dquot from the buffer, we cannot verify the
contents of the dquot are valid, as the CRC is invalid.

So, to avoid all the dquot CRC errors that are being detected by the
read verifier, change to using the same model as for inodes. That
is, dquot CRCs are calculated and written to the backing buffer at
the time the dquot is flushed to the backing buffer. If we modify
the dquot directly in the backing buffer, calculate the CRC
immediately after the modification is complete. Hence the dquot in
the on-disk buffer should always have a valid CRC.

Signed-off-by: Dave Chinner <dchinner@redhat.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Ben Myers <bpm@sgi.com>
Signed-off-by: Ben Myers <bpm@sgi.com>
(cherry picked from commit 6fcdc59de28817d1fbf1bd58cc01f4f3fac858fb)

10 years agoarch, mm: Remove tlb_fast_mode()
Peter Zijlstra [Wed, 5 Jun 2013 10:26:50 +0000 (12:26 +0200)]
arch, mm: Remove tlb_fast_mode()

Since the introduction of preemptible mmu_gather TLB fast mode has been
broken. TLB fast mode relies on there being absolutely no concurrency;
it frees pages first and invalidates TLBs later.

However now we can get concurrency and stuff goes *bang*.

This patch removes all tlb_fast_mode() code; it was found the better
option vs trying to patch the hole by entangling tlb invalidation with
the scheduler.

Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Russell King <linux@arm.linux.org.uk>
Cc: Tony Luck <tony.luck@intel.com>
Reported-by: Max Filippov <jcmvbkbc@gmail.com>
Signed-off-by: Peter Zijlstra <peterz@infradead.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
10 years agoMerge branch 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild
Linus Torvalds [Thu, 6 Jun 2013 01:05:45 +0000 (10:05 +0900)]
Merge branch 'rc-fixes' of git://git./linux/kernel/git/mmarek/kbuild

Pull kbuild fixes from Michal Marek:
 "There is one fix for a kbuild regression, plus three kconfig fixes for
  bugs that have alway been there, but are simple enough to be fixed in
  an -rc"

* 'rc-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mmarek/kbuild:
  kconfig/menu.c: fix multiple references to expressions in menu_add_prop()
  mconf: handle keys in empty dialogs
  kbuild: Don't assume dts files live in arch/*/boot/dts
  scripts/config: fix assignment of parameters for short version of --*-after options

10 years agox86/PCI: Map PCI setup data with ioremap() so it can be in highmem
Matt Fleming [Wed, 5 Jun 2013 14:15:41 +0000 (15:15 +0100)]
x86/PCI: Map PCI setup data with ioremap() so it can be in highmem

f9a37be0f0 ("x86: Use PCI setup data") added support for using PCI ROM
images from setup_data.  This used phys_to_virt(), which is not valid for
highmem addresses, and can cause a crash when booting a 32-bit kernel via
the EFI boot stub.

pcibios_add_device() assumes that the physical addresses stored in
setup_data are accessible via the direct kernel mapping, and that calling
phys_to_virt() is valid.  This isn't guaranteed to be true on x86 where the
direct mapping range is much smaller than on x86-64.

Calling phys_to_virt() on a highmem address results in the following:

 BUG: unable to handle kernel paging request at 39a3c198
 IP: [<c262be0f>] pcibios_add_device+0x2f/0x90
 ...
 Call Trace:
  [<c2370c73>] pci_device_add+0xe3/0x130
  [<c274640b>] pci_scan_single_device+0x8b/0xb0
  [<c2370d08>] pci_scan_slot+0x48/0x100
  [<c2371904>] pci_scan_child_bus+0x24/0xc0
  [<c262a7b0>] pci_acpi_scan_root+0x2c0/0x490
  [<c23b7203>] acpi_pci_root_add+0x312/0x42f
  ...

The solution is to use ioremap() instead of phys_to_virt() to map the
setup data into the kernel address space.

[bhelgaas: changelog]
Tested-by: Jani Nikula <jani.nikula@intel.com>
Signed-off-by: Matt Fleming <matt.fleming@intel.com>
Signed-off-by: Bjorn Helgaas <bhelgaas@google.com>
Cc: Matthew Garrett <mjg59@srcf.ucam.org>
Cc: Seth Forshee <seth.forshee@canonical.com>
Cc: Jesse Barnes <jbarnes@virtuousgeek.org>
Cc: stable@vger.kernel.org # v3.8+
10 years agoUSB: serial: fix TIOCMIWAIT return value
Johan Hovold [Wed, 5 Jun 2013 10:21:11 +0000 (12:21 +0200)]
USB: serial: fix TIOCMIWAIT return value

Fix regression introduced by commit 143d9d9616 ("USB: serial: add
tiocmiwait subdriver operation") which made the ioctl operation return
ENODEV rather than ENOIOCTLCMD when a subdriver TIOCMIWAIT
implementation is missing.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agonetfilter: xt_TCPMSS: Fix violation of RFC879 in absence of MSS option
Phil Oester [Tue, 4 Jun 2013 05:09:27 +0000 (05:09 +0000)]
netfilter: xt_TCPMSS: Fix violation of RFC879 in absence of MSS option

The clamp-mss-to-pmtu option of the xt_TCPMSS target can cause issues
connecting to websites if there was no MSS option present in the
original SYN packet from the client. In these cases, it may add a
MSS higher than the default specified in RFC879. Fix this by never
setting a value > 536 if no MSS option was specified by the client.

This closes netfilter's bugzilla #662.

Signed-off-by: Phil Oester <kernel@linuxace.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nfnetlink_cttimeout: fix incomplete dumping of objects
Pablo Neira Ayuso [Sat, 1 Jun 2013 13:36:02 +0000 (15:36 +0200)]
netfilter: nfnetlink_cttimeout: fix incomplete dumping of objects

Fix broken incomplete object dumping if the list of objects does not
fit into one single netlink message.

Reported-by: Gabriel Lazar <Gabriel.Lazar@com.utcluj.ro>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agonetfilter: nfnetlink_acct: fix incomplete dumping of objects
Pablo Neira Ayuso [Sat, 1 Jun 2013 13:31:40 +0000 (15:31 +0200)]
netfilter: nfnetlink_acct: fix incomplete dumping of objects

Fix broken incomplete object dumping if the list of objects does not
fit into one single netlink message.

Reported-by: Gabriel Lazar <Gabriel.Lazar@com.utcluj.ro>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Wed, 5 Jun 2013 10:19:04 +0000 (19:19 +0900)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Fix timeouts with direct mode authentication in mac80211, from
    Stanislaw Gruszka.

 2) Aggregation sessions can deadlock in ath9k, from Felix Fietkau.

 3) Netfilter's xt_addrtype doesn't work with ipv6 due to route lookups
    creating undesirable cache entries, from Florian Westphal.

 4) Fix netfilter's ipt_ULOG from generating non-NULL terminated
    strings.

 5) Fix netdev transmit queue crashes in mac80211, from Johannes Berg.

 6) Fix copy and paste error in 802.11 stack that broke reporting of
    64-bit station tx statistics, from Felix Fietkau.

 7) When qlge_probe fails, it leaks the netdev.  Fix from Wei Yongjun.

 8) SKB control block (where we store the IP options information,
    amongst other things) must be cleared properly otherwise ICMP
    sending can crash for IP tunnels.  Fix from Eric Dumazet.

 9) Verification of Energy Efficient Ether support was coded wrongly,
    the test was inversed.  Fix from Giuseppe CAVALLARO.

10) TCP handles redirects improperly because the wrong flow key is used
    for the route lookup.  From Michal Kubecek.

11) Don't interpret MSG_CMSG_COMPAT from userspace, fix from Andy
    Lutomirski.

12) The new AF_VSOCK was missing from the lockdep string table, fix from
    Federico Vaga.

13) be2net doesn't handle checksumming of IP fragments properly, from
    Somnath Kotur.

14) Fix several bugs in the device address list code that lead to
    crashes and other misbehaviors.  From Jay Vosburgh.

15) Fix ipv6 segmentation handling of fragmented GRE tunnel traffic,
    from Pravin B Shalr.

16) Fix usage of stale policies in IPSEC layer, from Paul Moore.

17) Fix team driver dump of ports when there are a large number of them,
    from Jiri Pirko.

18) Fix softlockups in UDP ipv4 socket lookup causes by and error in the
    hlist_nulls_for_each_entry_rcu() macro.  From Eric Dumazet.

19) Fix several regressions added by the high rate accuracy changes to
    the htb packet scheduler.  From Eric Dumazet.

20) Fix DMA'ing onto the stack in esd_usb2 and peak_usb CAN drivers,
    from Olivier Sobrie and Marc Kleine-Budde.

21) Fix unremovable network devices due to missing route pointer
    installation in the per-device ipv6 address list entries.  From Gao
    feng.

22) Apply the tg3 5719 DMA workaround on 5720 chips as well, otherwise
    we get stalls.  From Nithin Sujir.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (68 commits)
  net_sched: htb: do not mix 1ns and 64ns time units
  net: fix sk_buff head without data area
  tg3: Add read dma workaround for 5720
  net: ethernet: xilinx_emaclite: set protocol selector bits when writing ANAR
  bnx2x: Fix bridged GSO for 57710/57711 chips
  net: fec: add fallback to random MAC address
  bnx2x: fix TCP offload for tunneling ipv4 over ipv6
  ipv6: assign rt6_info to inet6_ifaddr in init_loopback
  net/mlx4_core: Keep VF assigned MAC in the PF admin table
  net/mlx4_en: Handle unassigned VF MAC address correctly
  net/mlx4_core: Return -EPROBE_DEFER when a VF is probed before PF is sufficiently initialized
  net/mlx4_en: Fix adaptive moderation cq update
  net: can: peak_usb: Do not do dma on the stack
  net: can: esd_usb2: Do not do dma on the stack
  net: can: kvaser_usb: fix reception on "USBcan Pro" and "USBcan R" type hardware.
  net_sched: restore "overhead xxx" handling
  net: force a reload of first item in hlist_nulls_for_each_entry_rcu
  hyperv: Fix vlan_proto setting in netvsc_recv_callback()
  team: fix port list dump for big number of ports
  list: introduce list_first_entry_or_null
  ...

10 years agonet_sched: htb: do not mix 1ns and 64ns time units
Eric Dumazet [Tue, 4 Jun 2013 07:11:48 +0000 (07:11 +0000)]
net_sched: htb: do not mix 1ns and 64ns time units

commit 56b765b79 ("htb: improved accuracy at high rates") added another
regression for low rates, because it mixes 1ns and 64ns time units.

So the maximum delay (mbuffer) was not 60 second, but 937 ms.

Lets convert all time fields to 1ns as 64bit arches are becoming the
norm.

Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Tested-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: fix sk_buff head without data area
Pablo Neira [Mon, 3 Jun 2013 09:28:43 +0000 (09:28 +0000)]
net: fix sk_buff head without data area

Eric Dumazet spotted that we have to check skb->head instead
of skb->data as skb->head points to the beginning of the
data area of the skbuff. Similarly, we have to initialize the
skb->head pointer, not skb->data in __alloc_skb_head.

After this fix, netlink crashes in the release path of the
sk_buff, so let's fix that as well.

This bug was introduced in (0ebd0ac net: add function to
allocate sk_buff head without data area).

Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agotg3: Add read dma workaround for 5720
Nithin Sujir [Mon, 3 Jun 2013 09:19:34 +0000 (09:19 +0000)]
tg3: Add read dma workaround for 5720

Commit 091f0ea30074bc43f9250961b3247af713024bc6 "tg3: Add New 5719 Read
DMA workaround" added a workaround for TX DMA stall on the 5719. This
workaround needs to be applied to the 5720 as well.

Cc: stable@vger.kernel.org
Reported-by: Roland Dreier <roland@purestorage.com>
Tested-by: Roland Dreier <roland@purestorage.com>
Signed-off-by: Nithin Nayak Sujir <nsujir@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: ethernet: xilinx_emaclite: set protocol selector bits when writing ANAR
Jens Renner \(EFE\) [Mon, 3 Jun 2013 04:32:52 +0000 (04:32 +0000)]
net: ethernet: xilinx_emaclite: set protocol selector bits when writing ANAR

This patch sets the protocol selector bits (4:0) of the PHY's MII_ADVERTISE
register (ANAR) when writing ADVERTISE_ALL. The protocol selector bits are
indicating IEEE 803.3u support and are fixed / read-only on some PHYs. Not
setting them correctly on others (like TI DP83630) makes the PHY fall back
to 10M HDX mode which should be avoided.

Tested for TI DP83630 PHY on Microblaze platform.

Signed-off-by: Jens Renner <renner@efe-gmbh.de>
Tested-by: Michal Simek <monstr@monstr.eu>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobnx2x: Fix bridged GSO for 57710/57711 chips
Yuval Mintz [Mon, 3 Jun 2013 02:59:57 +0000 (02:59 +0000)]
bnx2x: Fix bridged GSO for 57710/57711 chips

It was recently found out that GSO on 57710/57711 was broken, due to packets
being sent without a valid IP checksum.

Commit 057cf65 "bnx2x: Fix GSO for 57710/57711 chips" partially fixed this
issue, but failed to set the correct IP checksum when receiving GSO packets
via bridges, as such packets enter bnx2x_tx_split() and the FW flags needed
to calculate IP checksum were erroneously set in the incorrect
buffer descriptor.

This patch re-enables GSO in said scenario for 57710/57711 chips.

Signed-off-by: Yuval Mintz <yuvalmin@broadcom.com>
Signed-off-by: Ariel Elior <ariele@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge tag 'mmc-fixes-for-3.10-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 5 Jun 2013 00:13:06 +0000 (09:13 +0900)]
Merge tag 'mmc-fixes-for-3.10-rc5' of git://git./linux/kernel/git/cjb/mmc

Pull MMC fixes from Chris Ball:
 - sdhci-acpi: Fix initial runtime PM status, add more ACPI IDs
 - atmel-mci, omap_hsmmc: DT handling fixes
 - esdhc-imx: Fix SDIO IRQs, fix multiblock reads (both h/w errata)

* tag 'mmc-fixes-for-3.10-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/cjb/mmc:
  mmc: omap_hsmmc: Skip platform_get_resource_byname() for dt case
  mmc: omap_hsmmc: convert to dma_request_slave_channel_compat
  mmc: omap_hsmmc: Fix the DT pbias workaround for MMC controllers 2 to 5
  mmc: sdhci-pci: add more device ids
  mmc: sdhci-acpi: add more device ids
  mmc: sdhci-acpi: fix initial runtime pm status
  mmc: atmel-mci: convert to dma_request_slave_channel_compat()
  mmc: sdhci-esdhc-imx: fix multiblock reads on i.MX53
  mmc: sdhci-esdhc-imx: Fix SDIO interrupts

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Wed, 5 Jun 2013 00:11:06 +0000 (09:11 +0900)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixes from Dmitry Torokhov:
 "Just a 2 small driver fixups here"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: wacom - fix a typo for Cintiq 22HDT
  Input: synaptics - fix sync lost after resume on some laptops

10 years agoMerge branch 'fixes' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Wed, 5 Jun 2013 00:09:35 +0000 (09:09 +0900)]
Merge branch 'fixes' of git://git./virt/kvm/kvm

Pull kvm bugfixes from Gleb Natapov:
 "The bulk of the fixes is in MIPS KVM kernel<->userspace ABI.  MIPS KVM
  is new for 3.10 and some problems were found with current ABI.  It is
  better to fix them now and do not have a kernel with broken one"

* 'fixes' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: Fix race in apic->pending_events processing
  KVM: fix sil/dil/bpl/spl in the mod/rm fields
  KVM: Emulate multibyte NOP
  ARM: KVM: be more thorough when invalidating TLBs
  ARM: KVM: prevent NULL pointer dereferences with KVM VCPU ioctl
  mips/kvm: Use ENOIOCTLCMD to indicate unimplemented ioctls.
  mips/kvm: Fix ABI by moving manipulation of CP0 registers to KVM_{G,S}ET_ONE_REG
  mips/kvm: Use ARRAY_SIZE() instead of hardcoded constants in kvm_arch_vcpu_ioctl_{s,g}et_regs
  mips/kvm: Fix name of gpr field in struct kvm_regs.
  mips/kvm: Fix ABI for use of 64-bit registers.
  mips/kvm: Fix ABI for use of FPU.

10 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-fixes
Linus Torvalds [Wed, 5 Jun 2013 00:06:28 +0000 (09:06 +0900)]
Merge git://git./linux/kernel/git/steve/gfs2-3.0-fixes

Pull gfs2 fixes from Steven Whitehouse:
 "There are four patches this time.

  The first fixes a problem where the wrong descriptor type was being
  written into the log for journaled data blocks.

  The second fixes a race relating to the deallocation of allocator
  data.

  The third provides a fallback if kmalloc is unable to satisfy a
  request to allocate a directory hash table.

  The fourth fixes the iopen glock caching so that inodes are deleted in
  a more timely manner after rmdir/unlink"

* git://git.kernel.org/pub/scm/linux/kernel/git/steve/gfs2-3.0-fixes:
  GFS2: Don't cache iopen glocks
  GFS2: Fall back to vmalloc if kmalloc fails for dir hash tables
  GFS2: Increase i_writecount during gfs2_setattr_size
  GFS2: Set log descriptor type for jdata blocks

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi...
Linus Torvalds [Wed, 5 Jun 2013 00:03:31 +0000 (09:03 +0900)]
Merge branch 'for-linus' of git://git./linux/kernel/git/mszeredi/fuse

Pull fuse fixes from Miklos Szeredi:
 "One patch fixes an Oops introduced in 3.9 with the readdirplus
  feature.  The rest are fixes for async-dio in 3.10"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mszeredi/fuse:
  fuse: fix alignment in short read optimization for async_dio
  fuse: return -EIOCBQUEUED from fuse_direct_IO() for all async requests
  fuse: fix readdirplus Oops in fuse_dentry_revalidate
  fuse: update inode size and invalidate attributes on fallocate
  fuse: truncate pagecache range on hole punch
  fuse: allocate for_background dio requests based on io->async state

10 years agoMerge branch 'next' of git://git.monstr.eu/linux-2.6-microblaze
Linus Torvalds [Wed, 5 Jun 2013 00:02:09 +0000 (09:02 +0900)]
Merge branch 'next' of git://git.monstr.eu/linux-2.6-microblaze

Pull microblaze fixes from Michal Simek:
 "One is fixing warning reported by sparse and the second warning was
  reported by Geert in his build regressions/improvements status update
  for -rc4."

* 'next' of git://git.monstr.eu/linux-2.6-microblaze:
  microblaze: Use static inline functions in cacheflush.h
  microblaze: Fix sparse warnings

10 years agonet: fec: add fallback to random MAC address
Lucas Stach [Mon, 3 Jun 2013 00:38:39 +0000 (00:38 +0000)]
net: fec: add fallback to random MAC address

If no valid MAC address could be obtained from the hardware,
fall back to a randomly generated one.

Signed-off-by: Pavel Machek <pavel@denx.de>
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agobnx2x: fix TCP offload for tunneling ipv4 over ipv6
Dmitry Kravkov [Sun, 2 Jun 2013 23:28:41 +0000 (23:28 +0000)]
bnx2x: fix TCP offload for tunneling ipv4 over ipv6

FW was initialized with data from wrong header, this caused TSO packets
have wrong IP csum.

Signed-off-by: Dmitry Kravkov <dmitry@broadcom.com>
Signed-off-by: Ariel Elior <ariele@broadcom.com>
Signed-off-by: Eilon Greenstein <eilong@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoipv6: assign rt6_info to inet6_ifaddr in init_loopback
Gao feng [Sun, 2 Jun 2013 22:16:21 +0000 (22:16 +0000)]
ipv6: assign rt6_info to inet6_ifaddr in init_loopback

Commit 25fb6ca4ed9cad72f14f61629b68dc03c0d9713f
"net IPv6 : Fix broken IPv6 routing table after loopback down-up"
forgot to assign rt6_info to the inet6_ifaddr.
When disable the net device, the rt6_info which allocated
in init_loopback will not be destroied in __ipv6_ifa_notify.

This will trigger the waring message below
[23527.916091] unregister_netdevice: waiting for tap0 to become free. Usage count = 1

Reported-by: Arkadiusz Miskiewicz <a.miskiewicz@gmail.com>
Signed-off-by: Gao feng <gaofeng@cn.fujitsu.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge branch 'fixes-for-3.10' of git://gitorious.org/linux-can/linux-can
David S. Miller [Tue, 4 Jun 2013 21:30:12 +0000 (14:30 -0700)]
Merge branch 'fixes-for-3.10' of git://gitorious.org/linux-can/linux-can

Marc Kleine-Budde says:

====================
here are there fixes for the v3.10 release cycle:

The first patch by Jonas Peterson and Olivier Sobrie fixes the reception of CAN
frames on Kvaser's "USBcan Pro" and "USBcan R" type hardware.

The last two patches by Olivier Sobrie (for esd_usb2) and me (for peak_usb)
change the memory handling for the USB messages from stack to kmalloc(), as
memory used for DMA should not be allocated on stack.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_core: Keep VF assigned MAC in the PF admin table
Or Gerlitz [Tue, 4 Jun 2013 05:13:29 +0000 (05:13 +0000)]
net/mlx4_core: Keep VF assigned MAC in the PF admin table

MAC addresses assigned by the PF to VFs were not kept in the PF driver
admin table. As a result, displaying the VF MACs from the PF interface
to user space showed zero address where in fact the VF got non-zero
address from the PF, fix that.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Handle unassigned VF MAC address correctly
Or Gerlitz [Tue, 4 Jun 2013 05:13:28 +0000 (05:13 +0000)]
net/mlx4_en: Handle unassigned VF MAC address correctly

When a VF sense they didn't get MAC address, use random one. This will
address the case of administrator not assigning MAC to the VF through
the PF OS APIs and keep udev happy.

Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_core: Return -EPROBE_DEFER when a VF is probed before PF is sufficiently...
Jack Morgenstein [Tue, 4 Jun 2013 05:13:27 +0000 (05:13 +0000)]
net/mlx4_core: Return -EPROBE_DEFER when a VF is probed before PF is sufficiently initialized

In the PF initialization, SRIOV is enabled before the PF is fully initialized.
This allows the kernel to probe the newly-exposed VFs before the PF is ready
to handle them (nested probes).

Have the probe method return the -EPROBE_DEFER value in this situation (instead
of the VF probe method retrying its initialization in a loop, and returning -EIO
on failure). When -EPROBE_DEFER is returned by the VF probe method, the kernel
itself will retry the probe after a suitable delay.

Based upon a suggestion by Ben Hutchings <bhutchings@solarflare.com>

Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet/mlx4_en: Fix adaptive moderation cq update
Sagi Grimberg [Tue, 4 Jun 2013 05:13:26 +0000 (05:13 +0000)]
net/mlx4_en: Fix adaptive moderation cq update

When turning on adaptive_rx under adaptive moderation, the CQ's moderation
count wasn't updated according to rx_frames which resulted in too many
interrupts and bandwidth drop.

Signed-off-by: Sagi Grimberg <sagig@mellanox.com>
Signed-off-by: Amir Vadai <amirv@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoMerge tag 'iio-fixes-for-3.10b' of git://git.kernel.org/pub/scm/linux/kernel/git...
Greg Kroah-Hartman [Tue, 4 Jun 2013 19:30:11 +0000 (12:30 -0700)]
Merge tag 'iio-fixes-for-3.10b' of git://git./linux/kernel/git/jic23/iio into staging-linus

Jonathan writes:

Second round of IIO fixes for the 3.10 cycle.

A couple of little bits and pieces, some delayed due to traveling.

1) A memory leak fix in the callback buffer.
2) Wrong exit path due to a return when it should have been a goto.
3) Bug in a mask value in ad4350
4) Reading the wrong value in raw to processed utility function.

10 years agoiio:inkern: Fix typo/bug in convert raw to processed.
Michael Hennerich [Mon, 3 Jun 2013 08:04:00 +0000 (09:04 +0100)]
iio:inkern: Fix typo/bug in convert raw to processed.

Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Signed-off-by: Jonathan Cameron <jic23@kernel.org>
10 years agoiio: frequency: ad4350: Fix bug / typo in mask
Michael Hennerich [Mon, 3 Jun 2013 13:30:00 +0000 (14:30 +0100)]
iio: frequency: ad4350: Fix bug / typo in mask

Signed-off-by: Michael Hennerich <michael.hennerich@analog.com>
Reviewed-by: Lars-Peter Clausen <lars@metafoo.de>
Signed-off-by: Jonathan Cameron <jic23@kernel.org>
10 years agoinkern: iio_device_put after incorrect return/goto
Joe Perches [Tue, 4 Jun 2013 14:44:00 +0000 (15:44 +0100)]
inkern: iio_device_put after incorrect return/goto

The code uses

    return foo;
    goto err_type;

when instead the form should have been

    ret = foo;
    goto err_type;

Here this causes a useful iio_device_put to be skipped.

Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Jonathan Cameron <jic23@kernel.org>
10 years agostaging: alarm-dev: information leak in alarm_compat_ioctl()
Dan Carpenter [Tue, 4 Jun 2013 13:13:25 +0000 (16:13 +0300)]
staging: alarm-dev: information leak in alarm_compat_ioctl()

If we pass an invalid clock type then "ts" is never set.  We need to
check for errors earlier, otherwise we end up passing uninitialized
stack data to userspace.

Reported-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoiio:callback buffer: free the scan_mask
Jonathan Cameron [Wed, 22 May 2013 21:41:00 +0000 (22:41 +0100)]
iio:callback buffer: free the scan_mask

Reported-by: Michał Mirosław <mirq-linux@rere.qmqm.pl>
Signed-off-by: Jonathan Cameron <jic23@kernel.org>
10 years agoUSB: mos7720: fix hardware flow control
Johan Hovold [Tue, 4 Jun 2013 16:50:31 +0000 (18:50 +0200)]
USB: mos7720: fix hardware flow control

The register access to enable hardware flow control depends on the
device port number and not the port minor number.

Cc: stable@vger.kernel.org
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: keyspan: remove unused endpoint-array access
Johan Hovold [Tue, 4 Jun 2013 16:50:30 +0000 (18:50 +0200)]
USB: keyspan: remove unused endpoint-array access

Remove the no longer used endpoint-array access completely.

Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: keyspan: fix bogus array index
Johan Hovold [Tue, 4 Jun 2013 16:50:29 +0000 (18:50 +0200)]
USB: keyspan: fix bogus array index

The outcont_endpoints array was indexed using the port minor number
(which can be greater than the array size) rather than the device port
number.

Cc: stable@vger.kernel.org
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: zte_ev: fix broken open
Johan Hovold [Tue, 4 Jun 2013 16:50:28 +0000 (18:50 +0200)]
USB: zte_ev: fix broken open

Remove bogus port-number check in open and close, which prevented this
driver from being used with a minor number different from zero.

Cc: stable@vger.kernel.org
Signed-off-by: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoInput: wacom - fix a typo for Cintiq 22HDT
Ping Cheng [Thu, 23 May 2013 17:22:33 +0000 (10:22 -0700)]
Input: wacom - fix a typo for Cintiq 22HDT

And make the lines easier to read.

Signed-off-by: Ping Cheng <pingc@wacom.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
10 years agoInput: synaptics - fix sync lost after resume on some laptops
Eric Miao [Tue, 4 Jun 2013 16:30:55 +0000 (09:30 -0700)]
Input: synaptics - fix sync lost after resume on some laptops

In summary, the symptom is intermittent key events lost after resume
on some machines with synaptics touchpad (seems this is synaptics _only_),
and key events loss is due to serio port reconnect after psmouse sync lost.
Removing psmouse and inserting it back during the suspend/resume process
is able to work around the issue, so the difference between psmouse_connect()
and psmouse_reconnect() is the key to the root cause of this problem.

After comparing the two different paths, synaptics driver has its own
implementation of synaptics_reconnect(), and the missing psmouse_probe()
seems significant, the patch below added psmouse_probe() to the reconnect
process, and has been verified many times that the issue could not be reliably
reproduced.

There are two PS/2 commands in psmouse_probe():

  1. PSMOUSE_CMD_GETID
  2. PSMOUSE_CMD_RESET_DIS

Only the PSMOUSE_CMD_GETID seems to be significant. The
PSMOUSE_CMD_RESET_DIS is irrelevant to this issue after trying
several times.  So we have only implemented this patch to issue
the PSMOUSE_CMD_GETID so far.

Tested-by: Daniel Manrique <daniel.manrique@canonical.com>
Signed-off-by: James M Leddy <james.leddy@canonical.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
10 years agoMerge tag 'regulator-v3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 3 Jun 2013 21:34:51 +0000 (06:34 +0900)]
Merge tag 'regulator-v3.10-rc4' of git://git./linux/kernel/git/broonie/regulator

Pull regulator fixes from Mark Brown:
 "A few small fixes for v3.10, documentation things in the core and a
  few driver bugs."

* tag 'regulator-v3.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regulator:
  regulator: palmas: Fix "enable_reg" to point to the correct reg for SMPS10
  regulator: palmas: Fix incorrect condition
  regulator: core: Correct spelling mistake in comment
  regulator: dbx500: Make local symbol static
  regulator: Fix kernel-doc generation warnings.

10 years agoMerge tag 'jfs-3.10-rc5' of git://github.com/kleikamp/linux-shaggy
Linus Torvalds [Mon, 3 Jun 2013 21:33:44 +0000 (06:33 +0900)]
Merge tag 'jfs-3.10-rc5' of git://github.com/kleikamp/linux-shaggy

Pull jfs bugfixes from David Kleikamp:
 "A couple jfs bug fixes for 3.10-rc5"

* tag 'jfs-3.10-rc5' of git://github.com/kleikamp/linux-shaggy:
  fs/jfs: Add check if journaling to disk has been disabled in lbmRead()
  jfs: Several bugs in jfs_freeze() and jfs_unfreeze()

10 years agostaging: alarm-dev: information leak in alarm_ioctl()
Dan Carpenter [Mon, 3 Jun 2013 09:02:31 +0000 (02:02 -0700)]
staging: alarm-dev: information leak in alarm_ioctl()

Smatch complains that if we pass an invalid clock type then "ts" is
never set.  We need to check for errors earlier, otherwise we end up
passing uninitialized stack data to userspace.

Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Acked-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: serial: Add Option GTM681W to qcserial device table.
Richard Weinberger [Fri, 24 May 2013 10:01:51 +0000 (12:01 +0200)]
USB: serial: Add Option GTM681W to qcserial device table.

The Option GTM681W uses a qualcomm chip and can be
served by the qcserial device driver.

Signed-off-by: Richard Weinberger <richard@nod.at>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoUSB: Serial: cypress_M8: Enable FRWD Dongle hidcom device
Robert Butora [Fri, 31 May 2013 15:09:51 +0000 (18:09 +0300)]
USB: Serial: cypress_M8: Enable FRWD Dongle hidcom device

The patch adds a new HIDCOM device and does not affect other devices
driven by the cypress_M8 module. Changes are:
- add VendorID ProductID to device tables
- skip unstable speed check because FRWD uses 115200bps
- skip reset at probe which is an issue workaround for this
particular device.

Signed-off-by: Robert Butora <robert.butora.fi@gmail.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoRevert "serial: 8250: Make SERIAL_8250_RUNTIME_UARTS work correctly"
Kyle McMartin [Mon, 3 Jun 2013 13:38:26 +0000 (09:38 -0400)]
Revert "serial: 8250: Make SERIAL_8250_RUNTIME_UARTS work correctly"

This reverts commit cfcec52e9781f08948c6eb98198d65c45be75a70.

This regresses a longstanding behaviour on X86 systems, which end up with
PCI serial ports moving between ttyS4 and ttyS0 when you bisect to opposite
sides of this commit, resulting in the need to constantly modify the console
setting in order to bisect across it.

Please revert, we can work on solving this for ARM platforms in a less
disruptive way.

Signed-off-by: Kyle McMartin <kyle@mcmartin.ca>
Cc: Karthik Manamcheri <karthik.manamcheri@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoserial: samsung: enable clock before clearing pending interrupts during init
Chander Kashyap [Tue, 28 May 2013 13:02:07 +0000 (18:32 +0530)]
serial: samsung: enable clock before clearing pending interrupts during init

Ensure that the uart controller clock is enabled prior to writing to the
interrupt mask and pending registers in the s3c24xx_serial_init_port
function.

Signed-off-by: Chander Kashyap <chander.kashyap@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoserial/imx: disable hardware flow control at startup
Lucas Stach [Thu, 30 May 2013 13:47:04 +0000 (15:47 +0200)]
serial/imx: disable hardware flow control at startup

We only want to enable hardware flow control if RTS/CTS pins
are connected.

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Markus Pargmann <mpa@pengutronix.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
10 years agoGFS2: Don't cache iopen glocks
Bob Peterson [Wed, 29 May 2013 15:51:52 +0000 (11:51 -0400)]
GFS2: Don't cache iopen glocks

This patch makes GFS2 immediately reclaim/delete all iopen glocks
as soon as they're dequeued. This allows deleters to get an
EXclusive lock on iopen so files are deleted properly instead of
being set as unlinked.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
10 years agoGFS2: Fall back to vmalloc if kmalloc fails for dir hash tables
Bob Peterson [Thu, 30 May 2013 13:48:56 +0000 (09:48 -0400)]
GFS2: Fall back to vmalloc if kmalloc fails for dir hash tables

This version has one more correction: the vmalloc calls are replaced
by __vmalloc calls to preserve the GFP_NOFS flag.

When GFS2's directory management code allocates buffers for a
directory hash table, if it can't get the memory it needs, it
currently gives a bad return code. Rather than giving an error,
this patch allows it to use virtual memory rather than kernel
memory for the hash table. This should make it possible for
directories to function properly, even when kernel memory becomes
very fragmented.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
10 years agoGFS2: Increase i_writecount during gfs2_setattr_size
Bob Peterson [Tue, 28 May 2013 14:04:44 +0000 (10:04 -0400)]
GFS2: Increase i_writecount during gfs2_setattr_size

This patch calls get_write_access in a few functions. This
merely increases inode->i_writecount for the duration of the function.
That will ensure that any file closes won't delete the inode's
multi-block reservation while the function is running.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
10 years agoGFS2: Set log descriptor type for jdata blocks
Bob Peterson [Fri, 24 May 2013 19:02:49 +0000 (15:02 -0400)]
GFS2: Set log descriptor type for jdata blocks

This patch sets the log descriptor type according to whether the
journal commit is for (journaled) data or metadata. This was
recently broken when the functions to process data and metadata
log ops were combined.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>
10 years agofuse: fix alignment in short read optimization for async_dio
Maxim Patlasov [Thu, 30 May 2013 12:41:34 +0000 (16:41 +0400)]
fuse: fix alignment in short read optimization for async_dio

The bug was introduced with async_dio feature: trying to optimize short reads,
we cut number-of-bytes-to-read to i_size boundary. Hence the following example:

truncate --size=300 /mnt/file
dd if=/mnt/file of=/dev/null iflag=direct

led to FUSE_READ request of 300 bytes size. This turned out to be problem
for userspace fuse implementations who rely on assumption that kernel fuse
does not change alignment of request from client FS.

The patch turns off the optimization if async_dio is disabled. And, if it's
enabled, the patch fixes adjustment of number-of-bytes-to-read to preserve
alignment.

Note, that we cannot throw out short read optimization entirely because
otherwise a direct read of a huge size issued on a tiny file would generate
a huge amount of fuse requests and most of them would be ACKed by userspace
with zero bytes read.

Signed-off-by: Maxim Patlasov <MPatlasov@parallels.com>
Reviewed-by: Brian Foster <bfoster@redhat.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
10 years agofuse: return -EIOCBQUEUED from fuse_direct_IO() for all async requests
Brian Foster [Thu, 30 May 2013 19:35:50 +0000 (15:35 -0400)]
fuse: return -EIOCBQUEUED from fuse_direct_IO() for all async requests

If request submission fails for an async request (i.e.,
get_user_pages() returns -ERESTARTSYS), we currently skip the
-EIOCBQUEUED return and drop into wait_for_sync_kiocb() forever.

Avoid this by always returning -EIOCBQUEUED for async requests. If
an error occurs, the error is passed into fuse_aio_complete(),
returned via aio_complete() and thus propagated to userspace via
io_getevents().

Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Maxim Patlasov <MPatlasov@parallels.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
10 years agofuse: fix readdirplus Oops in fuse_dentry_revalidate
Miklos Szeredi [Mon, 3 Jun 2013 12:40:22 +0000 (14:40 +0200)]
fuse: fix readdirplus Oops in fuse_dentry_revalidate

Fix bug introduced by commit 4582a4ab2a "FUSE: Adapt readdirplus to application
usage patterns".

We need to check for a positive dentry; negative dentries are not added by
readdirplus.  Secondly we need to advise the use of readdirplus on the *parent*,
otherwise the whole thing is useless.  Thirdly all this is only relevant if
"readdirplus_auto" mode is selected by the filesystem.

We advise the use of readdirplus only if the dentry was still valid.  If we had
to redo the lookup then there was no use in doing the -plus version.

Reported-by: Bernd Schubert <bernd.schubert@itwm.fraunhofer.de>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: Feng Shuo <steve.shuo.feng@gmail.com>
CC: stable@vger.kernel.org
10 years agonet: can: peak_usb: Do not do dma on the stack
Marc Kleine-Budde [Thu, 16 May 2013 09:36:40 +0000 (11:36 +0200)]
net: can: peak_usb: Do not do dma on the stack

smatch reports the following warnings:
drivers/net/can/usb/peak_usb/pcan_usb_pro.c:514 pcan_usb_pro_drv_loaded() error: doing dma on the stack (buffer)
drivers/net/can/usb/peak_usb/pcan_usb_pro.c:878 pcan_usb_pro_init() error: doing dma on the stack (&fi)
drivers/net/can/usb/peak_usb/pcan_usb_pro.c:889 pcan_usb_pro_init() error: doing dma on the stack (&bi)

See "Documentation/DMA-API-HOWTO.txt" section "What memory is DMA'able?"

Cc: Stephane Grosjean <s.grosjean@peak-system.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agonet: can: esd_usb2: Do not do dma on the stack
Olivier Sobrie [Fri, 18 Jan 2013 08:14:04 +0000 (09:14 +0100)]
net: can: esd_usb2: Do not do dma on the stack

smatch reports the following warnings:
drivers/net/can/usb/esd_usb2.c:640 esd_usb2_start() error: doing dma on the stack (&msg)
drivers/net/can/usb/esd_usb2.c:846 esd_usb2_close() error: doing dma on the stack (&msg)
drivers/net/can/usb/esd_usb2.c:855 esd_usb2_close() error: doing dma on the stack (&msg)
drivers/net/can/usb/esd_usb2.c:923 esd_usb2_set_bittiming() error: doing dma on the stack (&msg)
drivers/net/can/usb/esd_usb2.c:1047 esd_usb2_probe() error: doing dma on the stack (&msg)
drivers/net/can/usb/esd_usb2.c:1053 esd_usb2_probe() error: doing dma on the stack (&msg)

See "Documentation/DMA-API-HOWTO.txt" section "What memory is DMA'able?"

Signed-off-by: Olivier Sobrie <olivier@sobrie.be>
Cc: Matthias Fuchs <matthias.fuchs@esd.eu>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agonet: can: kvaser_usb: fix reception on "USBcan Pro" and "USBcan R" type hardware.
Jonas Peterson [Tue, 7 May 2013 20:05:23 +0000 (22:05 +0200)]
net: can: kvaser_usb: fix reception on "USBcan Pro" and "USBcan R" type hardware.

Unlike Kvaser Leaf light devices, some other Kvaser devices (like USBcan
Pro, USBcan R) receive CAN messages in CMD_LOG_MESSAGE frames. This
patch adds support for it.

Cc: linux-stable <stable@vger.kernel.org> # >= v3.8
Signed-off-by: Jonas Peterson <jonas.peterson@gmail.com>
Signed-off-by: Olivier Sobrie <olivier@sobrie.be>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
10 years agomicroblaze: Use static inline functions in cacheflush.h
Michal Simek [Mon, 3 Jun 2013 09:30:04 +0000 (11:30 +0200)]
microblaze: Use static inline functions in cacheflush.h

Using static inline functions ensure proper type checking
which also remove compilation warning for no MMU

Compilation warning:
arch/microblaze/include/asm/cacheflush.h: warning: 'addr'
 may be used uninitialized in this function [-Wmaybe-uninitialized]

Signed-off-by: Michal Simek <michal.simek@xilinx.com>
10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux...
Linus Torvalds [Mon, 3 Jun 2013 09:09:42 +0000 (18:09 +0900)]
Merge branch 'for-linus' of git://git./linux/kernel/git/geert/linux-m68k

Pull m68k fix from Geert Uytterhoeven:
 "A boot lock-up on Mac, also destined for stable"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/geert/linux-m68k:
  m68k/mac: Fix unexpected interrupt with CONFIG_EARLY_PRINTK

10 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Mon, 3 Jun 2013 09:04:07 +0000 (18:04 +0900)]
Merge branch 'for-linus' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Martin Schwidefsky:
 "Recent bug fixes, one of them touches a common code file.

  It adds two #ifndef/#endif pairs to asm-generic/io.h to be able to
  override xlate_dev_kmem_ptr and xlate_dev_mem_ptr."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/pgtable: Fix gmap notifier address
  s390/dasd: fix handling of gone paths
  s390/pgtable: Fix check for pgste/storage key handling
  arch: s390: appldata: using strncpy() and strnlen() instead of sprintf()
  s390/smp: lost IPIs on cpu hotplug
  kernel: Fix s390 absolute memory access for /dev/mem
  s390/dma: do not call debug_dma after free

10 years agoMerge branch 'for-3.10-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj...
Linus Torvalds [Mon, 3 Jun 2013 08:57:16 +0000 (17:57 +0900)]
Merge branch 'for-3.10-fixes' of git://git./linux/kernel/git/tj/cgroup

Pull cgroup fixes from Tejun Heo:

 - Fix for yet another xattr bug which may lead to NULL deref.

 - A subtle bug in for_each_descendant_pre().  This bug requires quite
   specific conditions to trigger and isn't too likely to actually
   happen in the wild, but maybe that just makes it that much more
   nastier.

 - A warning message added for silly cgroup re-mount (not -o remount,
   but unmount followed by mount) behavior.

* 'for-3.10-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/cgroup:
  cgroup: warn about mismatching options of a new mount of an existing hierarchy
  cgroup: fix a subtle bug in descendant pre-order walk
  cgroup: initialize xattr before calling d_instantiate()

10 years agoMerge branch 'for-3.10-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj...
Linus Torvalds [Mon, 3 Jun 2013 08:55:09 +0000 (17:55 +0900)]
Merge branch 'for-3.10-fixes' of git://git./linux/kernel/git/tj/libata

Pull libata changes from Tejun Heo:
 "Nothing too interesting.  PCI ID additions, some sata_rcar fixes and a
  fringe bug fix for DMADIR handling which shouldn't affect any device
  remotely modern."

* 'for-3.10-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/libata:
  sata_rcar: fix interrupt handling
  ahci: add an observed PCI ID for Marvell 88se9172 SATA controller
  sata_rcar: clear STOP bit in bmdma_start() method
  libata: make ata_exec_internal_sg honor DMADIR
  ata_piix: add PCI IDs for Intel BayTail
  libata: update "Maintained by:" tags

10 years agomicroblaze: Fix sparse warnings
Michal Simek [Thu, 30 May 2013 13:10:52 +0000 (15:10 +0200)]
microblaze: Fix sparse warnings

arch/microblaze/include/asm/uaccess.h:101:3:
 warning: cast removes address space of expression
arch/microblaze/include/asm/uaccess.h:107:2:
 warning: cast removes address space of expression

Signed-off-by: Michal Simek <michal.simek@xilinx.com>
10 years agoKVM: Fix race in apic->pending_events processing
Gleb Natapov [Mon, 3 Jun 2013 08:30:02 +0000 (11:30 +0300)]
KVM: Fix race in apic->pending_events processing

apic->pending_events processing has a race that may cause INIT and
SIPI
processing to be reordered:

vpu0:                            vcpu1:
set INIT
                               test_and_clear_bit(KVM_APIC_INIT)
                                  process INIT
set INIT
set SIPI
                               test_and_clear_bit(KVM_APIC_SIPI)
                                  process SIPI

At the end INIT is left pending in pending_events. The following patch
fixes this by latching pending event before processing them.

Signed-off-by: Gleb Natapov <gleb@redhat.com>
10 years agoKVM: fix sil/dil/bpl/spl in the mod/rm fields
Paolo Bonzini [Thu, 30 May 2013 14:35:55 +0000 (16:35 +0200)]
KVM: fix sil/dil/bpl/spl in the mod/rm fields

The x86-64 extended low-byte registers were fetched correctly from reg,
but not from mod/rm.

This fixes another bug in the boot of RHEL5.9 64-bit, but it is still
not enough.

Cc: <stable@vger.kernel.org> # 3.9
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
10 years agoKVM: Emulate multibyte NOP
Paolo Bonzini [Thu, 30 May 2013 11:22:39 +0000 (13:22 +0200)]
KVM: Emulate multibyte NOP

This is encountered when booting RHEL5.9 64-bit.  There is another bug
after this one that is not a simple emulation failure, but this one lets
the boot proceed a bit.

Cc: <stable@vger.kernel.org> # 3.9
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Gleb Natapov <gleb@redhat.com>
10 years agoARM: KVM: be more thorough when invalidating TLBs
Marc Zyngier [Tue, 14 May 2013 11:11:34 +0000 (12:11 +0100)]
ARM: KVM: be more thorough when invalidating TLBs

The KVM/ARM MMU code doesn't take care of invalidating TLBs before
freeing a {pte,pmd} table. This could cause problems if the page
is reallocated and then speculated into by another CPU.

Reported-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>
10 years agoARM: KVM: prevent NULL pointer dereferences with KVM VCPU ioctl
Andre Przywara [Wed, 8 May 2013 22:28:06 +0000 (00:28 +0200)]
ARM: KVM: prevent NULL pointer dereferences with KVM VCPU ioctl

Some ARM KVM VCPU ioctls require the vCPU to be properly initialized
with the KVM_ARM_VCPU_INIT ioctl before being used with further
requests. KVM_RUN checks whether this initialization has been
done, but other ioctls do not.
Namely KVM_GET_REG_LIST will dereference an array with index -1
without initialization and thus leads to a kernel oops.
Fix this by adding checks before executing the ioctl handlers.

 [ Removed superflous comment from static function - Christoffer ]

Changes from v1:
 * moved check into a static function with a meaningful name

Signed-off-by: Andre Przywara <andre.przywara@linaro.org>
Signed-off-by: Christoffer Dall <cdall@cs.columbia.edu>
10 years agomips/kvm: Use ENOIOCTLCMD to indicate unimplemented ioctls.
David Daney [Thu, 23 May 2013 16:49:10 +0000 (09:49 -0700)]
mips/kvm: Use ENOIOCTLCMD to indicate unimplemented ioctls.

The Linux Way is to return -ENOIOCTLCMD to the vfs when an
unimplemented ioctl is requested.  Do this in kvm_mips instead of a
random mixture of -ENOTSUPP and -EINVAL.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Sanjay Lal <sanjayl@kymasys.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
10 years agomips/kvm: Fix ABI by moving manipulation of CP0 registers to KVM_{G,S}ET_ONE_REG
David Daney [Thu, 23 May 2013 16:49:09 +0000 (09:49 -0700)]
mips/kvm: Fix ABI by moving manipulation of CP0 registers to KVM_{G,S}ET_ONE_REG

Because not all 256 CP0 registers are ever implemented, we need a
different method of manipulating them.  Use the
KVM_SET_ONE_REG/KVM_GET_ONE_REG mechanism.

Now unused code and definitions are removed.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Sanjay Lal <sanjayl@kymasys.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
10 years agomips/kvm: Use ARRAY_SIZE() instead of hardcoded constants in kvm_arch_vcpu_ioctl_...
David Daney [Thu, 23 May 2013 16:49:08 +0000 (09:49 -0700)]
mips/kvm: Use ARRAY_SIZE() instead of hardcoded constants in kvm_arch_vcpu_ioctl_{s,g}et_regs

Also we cannot set special zero register, so force it to zero.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Sanjay Lal <sanjayl@kymasys.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
10 years agomips/kvm: Fix name of gpr field in struct kvm_regs.
David Daney [Thu, 23 May 2013 16:49:07 +0000 (09:49 -0700)]
mips/kvm: Fix name of gpr field in struct kvm_regs.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Sanjay Lal <sanjayl@kymasys.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
10 years agomips/kvm: Fix ABI for use of 64-bit registers.
David Daney [Thu, 23 May 2013 16:49:06 +0000 (09:49 -0700)]
mips/kvm: Fix ABI for use of 64-bit registers.

All registers are 64-bits wide, 32-bit guests use the least
significant portion of the register storage fields.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Sanjay Lal <sanjayl@kymasys.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
10 years agomips/kvm: Fix ABI for use of FPU.
David Daney [Thu, 23 May 2013 16:49:05 +0000 (09:49 -0700)]
mips/kvm: Fix ABI for use of FPU.

Define a non-empty struct kvm_fpu.

Signed-off-by: David Daney <david.daney@cavium.com>
Acked-by: Sanjay Lal <sanjayl@kymasys.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
10 years agonet_sched: restore "overhead xxx" handling
Eric Dumazet [Sun, 2 Jun 2013 13:55:05 +0000 (13:55 +0000)]
net_sched: restore "overhead xxx" handling

commit 56b765b79 ("htb: improved accuracy at high rates")
broke the "overhead xxx" handling, as well as the "linklayer atm"
attribute.

tc class add ... htb rate X ceil Y linklayer atm overhead 10

This patch restores the "overhead xxx" handling, for htb, tbf
and act_police

The "linklayer atm" thing needs a separate fix.

Reported-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Vimalkumar <j.vimal@gmail.com>
Cc: Jiri Pirko <jpirko@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agonet: force a reload of first item in hlist_nulls_for_each_entry_rcu
Eric Dumazet [Wed, 29 May 2013 09:06:27 +0000 (09:06 +0000)]
net: force a reload of first item in hlist_nulls_for_each_entry_rcu

Roman Gushchin discovered that udp4_lib_lookup2() was not reloading
first item in the rcu protected list, in case the loop was restarted.

This produced soft lockups as in https://lkml.org/lkml/2013/4/16/37

rcu_dereference(X)/ACCESS_ONCE(X) seem to not work as intended if X is
ptr->field :

In some cases, gcc caches the value or ptr->field in a register.

Use a barrier() to disallow such caching, as documented in
Documentation/atomic_ops.txt line 114

Thanks a lot to Roman for providing analysis and numerous patches.

Diagnosed-by: Roman Gushchin <klamm@yandex-team.ru>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Boris Zhmurov <zhmurov@yandex-team.ru>
Signed-off-by: Roman Gushchin <klamm@yandex-team.ru>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
10 years agoLinux 3.10-rc4 v3.10-rc4
Linus Torvalds [Sun, 2 Jun 2013 08:11:17 +0000 (17:11 +0900)]
Linux 3.10-rc4