firefly-linux-kernel-4.4.55.git
9 years agoSunrpc: Supports hexadecimal number for sysctl files of sunrpc debug
Kinglong Mee [Sat, 12 Sep 2015 01:37:18 +0000 (09:37 +0800)]
Sunrpc: Supports hexadecimal number for sysctl files of sunrpc debug

The sunrpc debug sysctl files only accept decimal number right now.
But all the XXXDBUG_XXX macros are defined as hexadecimal.
It is not easy to set or check an separate flag.

This patch let those files support accepting hexadecimal number,
(decimal number is also supported). Also, display it as hexadecimal.

v2,
Remove duplicate parsing of '0x...', just using simple_strtol(tmpbuf, &s, 0)
Fix a bug of isspace() checking after parsing

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfs: Fix GETATTR bitmap verification
Andreas Gruenbacher [Tue, 3 Nov 2015 17:25:33 +0000 (18:25 +0100)]
nfs: Fix GETATTR bitmap verification

When decoding GETATTR replies, the client checks the attribute bitmap
for which attributes the server has sent.  It misses bits at the word
boundaries, though; fix that.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfs: Remove unused xdr page offsets in getacl/setacl arguments
Andreas Gruenbacher [Tue, 3 Nov 2015 17:25:34 +0000 (18:25 +0100)]
nfs: Remove unused xdr page offsets in getacl/setacl arguments

The arguments passed around for getacl and setacl xdr encoding, struct
nfs_setaclargs and struct nfs_getaclargs, both contain an array of
pages, an offset into the first page, and the length of the page data.
The offset is unused as it is always zero; remove it.

Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agofs/nfs: remove unnecessary new_valid_dev check
Yaowei Bai [Mon, 28 Sep 2015 14:30:32 +0000 (22:30 +0800)]
fs/nfs: remove unnecessary new_valid_dev check

As new_valid_dev always returns 1, so !new_valid_dev check is not
needed, remove it.

Signed-off-by: Yaowei Bai <bywxiaobai@163.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoSUNRPC: fix variable type
Andrzej Hajda [Thu, 24 Sep 2015 14:00:09 +0000 (16:00 +0200)]
SUNRPC: fix variable type

Due to incorrect len type bc_send_request returned always zero.

The problem has been detected using proposed semantic patch
scripts/coccinelle/tests/assign_signed_to_unsigned.cocci [1].

[1]: http://permalink.gmane.org/gmane.linux.kernel/2046107

Signed-off-by: Andrzej Hajda <a.hajda@samsung.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoMerge tag 'nfs-rdma-4.4-2' of git://git.linux-nfs.org/projects/anna/nfs-rdma
Trond Myklebust [Mon, 2 Nov 2015 22:09:24 +0000 (17:09 -0500)]
Merge tag 'nfs-rdma-4.4-2' of git://git.linux-nfs.org/projects/anna/nfs-rdma

NFS: NFSoRDMA Client Side Changes

In addition to a variety of bugfixes, these patches are mostly geared at
enabling both swap and backchannel support to the NFS over RDMA client.

Signed-off-by: Anna Schumake <Anna.Schumaker@Netapp.com>
9 years agoNFS: Enable client side NFSv4.1 backchannel to use other transports
Chuck Lever [Sat, 24 Oct 2015 21:28:32 +0000 (17:28 -0400)]
NFS: Enable client side NFSv4.1 backchannel to use other transports

Forechannel transports get their own "bc_up" method to create an
endpoint for the backchannel service.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
[Anna Schumaker: Add forward declaration of struct net to xprt.h]
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agopNFS/flexfiles: Add support for FF_FLAGS_NO_IO_THRU_MDS
Trond Myklebust [Mon, 2 Nov 2015 14:59:00 +0000 (09:59 -0500)]
pNFS/flexfiles: Add support for FF_FLAGS_NO_IO_THRU_MDS

For loosely coupled pNFS/flexfiles systems, there is often no advantage
at all in going through the MDS for I/O, since the MDS is subject to
the same limitations as all other clients when talking to DSes. If a
DS is unresponsive, I/O through the MDS will fail.

For such systems, the only scalable solution is to have the pNFS clients
retry doing pNFS, and so the protocol now provides a flag that allows
the pNFS server to signal this.

If LAYOUTGET returns FF_FLAGS_NO_IO_THRU_MDS, then we should assume that
the MDS wants the client to retry using these devices, even if they were
previously marked as being unavailable. To do so, we add a helper,
ff_layout_mark_devices_valid() that will be called from layoutget.

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agopNFS/flexfiles: When mirrored, retry failed reads by switching mirrors
Trond Myklebust [Mon, 2 Nov 2015 15:11:17 +0000 (10:11 -0500)]
pNFS/flexfiles: When mirrored, retry failed reads by switching mirrors

If the pNFS/flexfiles file is mirrored, and a read to one mirror fails,
then we should bump the mirror index, so that we retry to a different
mirror. Once we've iterated through all mirrors and all failed, we can
return the layout and issue a new LAYOUTGET.

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoSUNRPC: Remove the TCP-only restriction in bc_svc_process()
Chuck Lever [Sat, 24 Oct 2015 21:28:24 +0000 (17:28 -0400)]
SUNRPC: Remove the TCP-only restriction in bc_svc_process()

Allow the use of other transport classes when handling a backward
direction RPC call.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agosvcrdma: Add backward direction service for RPC/RDMA transport
Chuck Lever [Sat, 24 Oct 2015 21:28:16 +0000 (17:28 -0400)]
svcrdma: Add backward direction service for RPC/RDMA transport

On NFSv4.1 mount points, the Linux NFS client uses this transport
endpoint to receive backward direction calls and route replies back
to the NFSv4.1 server.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Acked-by: "J. Bruce Fields" <bfields@fieldses.org>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Handle incoming backward direction RPC calls
Chuck Lever [Sat, 24 Oct 2015 21:28:08 +0000 (17:28 -0400)]
xprtrdma: Handle incoming backward direction RPC calls

Introduce a code path in the rpcrdma_reply_handler() to catch
incoming backward direction RPC calls and route them to the ULP's
backchannel server.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Add support for sending backward direction RPC replies
Chuck Lever [Sat, 24 Oct 2015 21:27:59 +0000 (17:27 -0400)]
xprtrdma: Add support for sending backward direction RPC replies

Backward direction RPC replies are sent via the client transport's
send_request method, the same way forward direction RPC calls are
sent.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Pre-allocate Work Requests for backchannel
Chuck Lever [Sat, 24 Oct 2015 21:27:51 +0000 (17:27 -0400)]
xprtrdma: Pre-allocate Work Requests for backchannel

Pre-allocate extra send and receive Work Requests needed to handle
backchannel receives and sends.

The transport doesn't know how many extra WRs to pre-allocate until
the xprt_setup_backchannel() call, but that's long after the WRs are
allocated during forechannel setup.

So, use a fixed value for now.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Pre-allocate backward rpc_rqst and send/receive buffers
Chuck Lever [Sat, 24 Oct 2015 21:27:43 +0000 (17:27 -0400)]
xprtrdma: Pre-allocate backward rpc_rqst and send/receive buffers

xprtrdma's backward direction send and receive buffers are the same
size as the forechannel's inline threshold, and must be pre-
registered.

The consumer has no control over which receive buffer the adapter
chooses to catch an incoming backwards-direction call. Any receive
buffer can be used for either a forward reply or a backward call.
Thus both types of RPC message must all be the same size.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoSUNRPC: Abstract backchannel operations
Chuck Lever [Sat, 24 Oct 2015 21:27:35 +0000 (17:27 -0400)]
SUNRPC: Abstract backchannel operations

xprt_{setup,destroy}_backchannel() won't be adequate for RPC/RMDA
bi-direction. In particular, receive buffers have to be pre-
registered and posted in order to receive incoming backchannel
requests.

Add a virtual function call to allow the insertion of appropriate
backchannel setup and destruction methods for each transport.

In addition, freeing a backchannel request is a little different
for RPC/RDMA. Introduce an rpc_xprt_op to handle the difference.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Saving IRQs no longer needed for rb_lock
Chuck Lever [Sat, 24 Oct 2015 21:27:27 +0000 (17:27 -0400)]
xprtrdma: Saving IRQs no longer needed for rb_lock

Now that RPC replies are processed in a workqueue, there's no need
to disable IRQs when managing send and receive buffers. This saves
noticeable overhead per RPC.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Remove reply tasklet
Chuck Lever [Sat, 24 Oct 2015 21:27:18 +0000 (17:27 -0400)]
xprtrdma: Remove reply tasklet

Clean up: The reply tasklet is no longer used.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Use workqueue to process RPC/RDMA replies
Chuck Lever [Sat, 24 Oct 2015 21:27:10 +0000 (17:27 -0400)]
xprtrdma: Use workqueue to process RPC/RDMA replies

The reply tasklet is fast, but it's single threaded. After reply
traffic saturates a single CPU, there's no more reply processing
capacity.

Replace the tasklet with a workqueue to spread reply handling across
all CPUs.  This also moves RPC/RDMA reply handling out of the soft
IRQ context and into a context that allows sleeps.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Replace send and receive arrays
Chuck Lever [Sat, 24 Oct 2015 21:27:02 +0000 (17:27 -0400)]
xprtrdma: Replace send and receive arrays

The rb_send_bufs and rb_recv_bufs arrays are used to implement a
pair of stacks for keeping track of free rpcrdma_req and rpcrdma_rep
structs. Replace those arrays with free lists.

To allow more than 512 RPCs in-flight at once, each of these arrays
would be larger than a page (assuming 8-byte addresses and 4KB
pages). Allowing up to 64K in-flight RPCs (as TCP now does), each
buffer array would have to be 128 pages. That's an order-6
allocation. (Not that we're going there.)

A list is easier to expand dynamically. Instead of allocating a
larger array of pointers and copying the existing pointers to the
new array, simply append more buffers to each list.

This also makes it simpler to manage receive buffers that might
catch backwards-direction calls, or to post receive buffers in
bulk to amortize the overhead of ib_post_recv.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Devesh Sharma <devesh.sharma@avagotech.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Refactor reply handler error handling
Chuck Lever [Sat, 24 Oct 2015 21:26:54 +0000 (17:26 -0400)]
xprtrdma: Refactor reply handler error handling

Clean up: The error cases in rpcrdma_reply_handler() almost never
execute. Ensure the compiler places them out of the hot path.

No behavior change expected.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Devesh Sharma <devesh.sharma@avagotech.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Prevent loss of completion signals
Chuck Lever [Sat, 24 Oct 2015 21:26:45 +0000 (17:26 -0400)]
xprtrdma: Prevent loss of completion signals

Commit 8301a2c047cc ("xprtrdma: Limit work done by completion
handler") was supposed to prevent xprtrdma's upcall handlers from
starving other softIRQ work by letting them return to the provider
before all CQEs have been polled.

The logic assumes the provider will call the upcall handler again
immediately if the CQ is re-armed while there are still queued CQEs.

This assumption is invalid. The IBTA spec says that after a CQ is
armed, the hardware must interrupt only when a new CQE is inserted.
xprtrdma can't rely on the provider calling again, even though some
providers do.

Therefore, leaving CQEs on queue makes sense only when there is
another mechanism that ensures all remaining CQEs are consumed in a
timely fashion. xprtrdma does not have such a mechanism. If a CQE
remains queued, the transport can wait forever to send the next RPC.

Finally, move the wcs array back onto the stack to ensure that the
poll array is always local to the CPU where the completion upcall is
running.

Fixes: 8301a2c047cc ("xprtrdma: Limit work done by completion ...")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Devesh Sharma <devesh.sharma@avagotech.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Re-arm after missed events
Chuck Lever [Sat, 24 Oct 2015 21:26:37 +0000 (17:26 -0400)]
xprtrdma: Re-arm after missed events

ib_req_notify_cq(IB_CQ_REPORT_MISSED_EVENTS) returns a positive
value if WCs were added to a CQ after the last completion upcall
but before the CQ has been re-armed.

Commit 7f23f6f6e388 ("xprtrmda: Reduce lock contention in
completion handlers") assumed that when ib_req_notify_cq() returned
a positive RC, the CQ had also been successfully re-armed, making
it safe to return control to the provider without losing any
completion signals. That is an invalid assumption.

Change both completion handlers to continue polling while
ib_req_notify_cq() returns a positive value.

Fixes: 7f23f6f6e388 ("xprtrmda: Reduce lock contention in ...")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Reviewed-by: Devesh Sharma <devesh.sharma@avagotech.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: Enable swap-on-NFS/RDMA
Chuck Lever [Sat, 24 Oct 2015 21:26:29 +0000 (17:26 -0400)]
xprtrdma: Enable swap-on-NFS/RDMA

After adding a swapfile on an NFS/RDMA mount and removing the
normal swap partition, I was able to push the NFS client well
into swap without any issue.

I forgot to swapoff the NFS file before rebooting. This pinned
the NFS mount and the IB core and provider, causing shutdown to
hang. I think this is expected and safe behavior. Probably
shutdown scripts should "swapoff -a" before unmounting any
filesystems.

Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Reviewed-by: Sagi Grimberg <sagig@mellanox.com>
Tested-By: Devesh Sharma <devesh.sharma@avagotech.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoxprtrdma: don't log warnings for flushed completions
Steve Wise [Mon, 21 Sep 2015 17:24:34 +0000 (12:24 -0500)]
xprtrdma: don't log warnings for flushed completions

Unsignaled send WRs can get flushed as part of normal unmount, so don't
log them as warnings.

Signed-off-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
9 years agoMerge branch 'bugfixes'
Trond Myklebust [Wed, 21 Oct 2015 21:07:21 +0000 (16:07 -0500)]
Merge branch 'bugfixes'

* bugfixes:
  NFSv4.1/pnfs: Retry through MDS when getting bad length of data
  nfs/blocklayout: Fix bad using of page offset in bl_read_pagelist
  NFS: Return directly if encode_sessionid fail
  NFS: Fix bad checking of max taglen in callback request
  NFS: Fix bad defines of callback response maxsize
  NFS: Use NFS4_MAX_SESSIONID_LEN directly for decode/encode sessionid
  NFS: Remove unneeded NFS_DEBUG checking before define NFSDBG_FACILITY
  NFS: Remove the left function defines in callback.h
  NFS: Remove the left global variable nfs_callback_tcpport
  NFS: Get rid of the unneeded addr stored in callback arguments
  nfsroot: make nfsroot to accept the 1024 bytes long directory name

9 years agoNFSv4.1/pnfs: Retry through MDS when getting bad length of data
Kinglong Mee [Fri, 16 Oct 2015 09:23:29 +0000 (17:23 +0800)]
NFSv4.1/pnfs: Retry through MDS when getting bad length of data

If non rpc-based layout driver return bad length of data, nfs retries
by calling rpc_restart_call_prepare() that cause an NULL reference panic.

This patch lets nfs retry through MDS for non rpc-based layout driver
return bad length of data.

[13034.883329] BUG: unable to handle kernel NULL pointer dereference at           (null)
[13034.884902] IP: [<ffffffffa00db372>] rpc_restart_call_prepare+0x62/0x90 [sunrpc]
[13034.886558] PGD 0
[13034.888126] Oops: 0000 [#1] KASAN
[13034.889710] Modules linked in: blocklayoutdriver(OE) nfsv4(OE) nfs(OE) fscache(E) nfsd(OE) xfs libcrc32c coretemp btrfs crct10dif_pclmul crc32_pclmul crc32c_intel ghash_clmulni_intel ppdev vmw_balloon auth_rpcgss shpchp nfs_acl lockd vmw_vmci parport_pc xor raid6_pq grace parport sunrpc i2c_piix4 vmwgfx drm_kms_helper ttm drm mptspi e1000 serio_raw scsi_transport_spi mptscsih mptbase ata_generic pata_acpi [last unloaded: fscache]
[13034.898260] CPU: 0 PID: 10112 Comm: kworker/0:1 Tainted: G           OE   4.3.0-rc5+ #279
[13034.899932] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 07/02/2015
[13034.903342] Workqueue: events bl_read_cleanup [blocklayoutdriver]
[13034.905059] task: ffff88006a9148c0 ti: ffff880035e90000 task.ti: ffff880035e90000
[13034.906827] RIP: 0010:[<ffffffffa00db372>]  [<ffffffffa00db372>] rpc_restart_call_prepare+0x62/0x90 [sunrpc]
[13034.910522] RSP: 0018:ffff880035e97b58  EFLAGS: 00010282
[13034.912378] RAX: fffffbfff04a5a94 RBX: ffff880068fe4858 RCX: 0000000000000003
[13034.914339] RDX: dffffc0000000000 RSI: 0000000000000003 RDI: 0000000000000282
[13034.916236] RBP: ffff880035e97b68 R08: 0000000000000001 R09: 0000000000000001
[13034.918229] R10: 0000000000000000 R11: 0000000000000001 R12: 0000000000000000
[13034.920007] R13: ffff880068fe4858 R14: ffff880068fe4a60 R15: 0000000000001000
[13034.921845] FS:  0000000000000000(0000) GS:ffffffff82247000(0000) knlGS:0000000000000000
[13034.923645] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[13034.925525] CR2: 0000000000000000 CR3: 00000000063dd000 CR4: 00000000001406f0
[13034.932808] Stack:
[13034.934813]  ffff880068fe4780 0000000000001000 ffff880035e97ba8 ffffffffa08800d2
[13034.936675]  ffffffffa088029d ffff880068fe4780 ffff880068fe4858 ffffffffa089c0a0
[13034.938593]  ffff880068fe47e0 ffff88005d59faf0 ffff880035e97be0 ffffffffa087e08f
[13034.940454] Call Trace:
[13034.942388]  [<ffffffffa08800d2>] nfs_readpage_result+0x112/0x200 [nfs]
[13034.944317]  [<ffffffffa088029d>] ? nfs_readpage_done+0xdd/0x160 [nfs]
[13034.946267]  [<ffffffffa087e08f>] nfs_pgio_result+0x9f/0x120 [nfs]
[13034.948166]  [<ffffffffa09266cc>] pnfs_ld_read_done+0x7c/0x1e0 [nfsv4]
[13034.950247]  [<ffffffffa03b07ee>] bl_read_cleanup+0x2e/0x60 [blocklayoutdriver]
[13034.952156]  [<ffffffff810ebf62>] process_one_work+0x412/0x870
[13034.954102]  [<ffffffff810ebe84>] ? process_one_work+0x334/0x870
[13034.955949]  [<ffffffff810ebb50>] ? queue_delayed_work_on+0x40/0x40
[13034.957985]  [<ffffffff810ec441>] worker_thread+0x81/0x6a0
[13034.959817]  [<ffffffff810ec3c0>] ? process_one_work+0x870/0x870
[13034.961785]  [<ffffffff810f43bd>] kthread+0x17d/0x1a0
[13034.963544]  [<ffffffff810f4240>] ? kthread_create_on_node+0x330/0x330
[13034.965479]  [<ffffffff81100428>] ? finish_task_switch+0x88/0x220
[13034.967223]  [<ffffffff810f4240>] ? kthread_create_on_node+0x330/0x330
[13034.968929]  [<ffffffff81b6ae5f>] ret_from_fork+0x3f/0x70
[13034.970534]  [<ffffffff810f4240>] ? kthread_create_on_node+0x330/0x330
[13034.972176] Code: c7 43 50 40 84 0d a0 e8 3d fe 1c e1 48 8d 7b 58 c7 83 e4 00 00 00 00 00 00 00 e8 ca fe 1c e1 4c 8b 63 58 4c 89 e7 e8 be fe 1c e1 <49> 83 3c 24 00 74 12 48 c7 43 50 f0 a2 0e a0 b8 01 00 00 00 5b
[13034.977148] RIP  [<ffffffffa00db372>] rpc_restart_call_prepare+0x62/0x90 [sunrpc]
[13034.978780]  RSP <ffff880035e97b58>
[13034.980399] CR2: 0000000000000000

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfs/blocklayout: Fix bad using of page offset in bl_read_pagelist
Kinglong Mee [Fri, 16 Oct 2015 09:22:50 +0000 (17:22 +0800)]
nfs/blocklayout: Fix bad using of page offset in bl_read_pagelist

Blocklayout uses file offset for the read-back page's offset of first writing,
it's definitely wrong, it writes data to bad address of page that cause userspace
application segment fault. It must be the page base stored in header->args.pgbase.

Also, the pg_offset has no influence with isect and extent length.

Note: The offset of the non-first page is always zero.

Ps: A test program will segment fault at read() as,
#define _GNU_SOURCE

#include <stdio.h>
#include <stdlib.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include <fcntl.h>
#include <errno.h>

int main(int argc, char **argv)
{
        char buf[2049];
        char *filename = NULL;
        int fd = -1;

        if (argc < 2) {
                printf("Usage: %s filename\n", argv[0]);
                return 0;
        }

        filename = argv[1];
        fd = open(filename, O_RDONLY | O_DIRECT);
        if (fd < 0) {
                printf("Open %s fail: %m\n", filename);
                return 1;
        }

        lseek(fd, 2048, SEEK_SET);
        if (read(fd, buf, sizeof(buf) - 1) != (sizeof(buf) - 1))
                printf("Read 4096 bityes data from %s fail: %m\n", filename);
out:
        close(fd);
        return 0;
}

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoNFS: Return directly if encode_sessionid fail
Kinglong Mee [Thu, 24 Sep 2015 12:58:38 +0000 (20:58 +0800)]
NFS: Return directly if encode_sessionid fail

encode_sessionid() may return error, nfs needs process the return value.

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoNFS: Fix bad checking of max taglen in callback request
Kinglong Mee [Thu, 24 Sep 2015 12:58:16 +0000 (20:58 +0800)]
NFS: Fix bad checking of max taglen in callback request

The taglen should be checked with CB_OP_TAGLEN_MAXSZ directly.

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoNFS: Fix bad defines of callback response maxsize
Kinglong Mee [Thu, 24 Sep 2015 12:57:58 +0000 (20:57 +0800)]
NFS: Fix bad defines of callback response maxsize

As CB_OP_TAGLEN_MAXSZ, all XXX_MAXSZ should be defined as bit.
Each operation should not cantains CB_OP_TAGLEN_MAXSZ.

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoNFS: Use NFS4_MAX_SESSIONID_LEN directly for decode/encode sessionid
Kinglong Mee [Thu, 24 Sep 2015 12:57:37 +0000 (20:57 +0800)]
NFS: Use NFS4_MAX_SESSIONID_LEN directly for decode/encode sessionid

It's no need to define a temporary variables for NFS4_MAX_SESSIONID_LEN.

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoNFS: Remove unneeded NFS_DEBUG checking before define NFSDBG_FACILITY
Kinglong Mee [Thu, 24 Sep 2015 12:57:03 +0000 (20:57 +0800)]
NFS: Remove unneeded NFS_DEBUG checking before define NFSDBG_FACILITY

It's not needed to checking NFS_DEBUG before define NFSDBG_FACILITY, remove it.

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoNFS: Remove the left function defines in callback.h
Kinglong Mee [Thu, 24 Sep 2015 12:56:29 +0000 (20:56 +0800)]
NFS: Remove the left function defines in callback.h

Commit 778be232a207 "NFS do not find client in NFSv4 pg_authenticate" has remove
the define and using of nfs4_set_callback_sessionid(), and
commit 36281caa839f "NFSv4: Further clean-ups of delegation stateid validation"
has update the checking of stateid, and move the code to nfs4proc.c.

This patch remove those function defines left in callback.h

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoNFS: Remove the left global variable nfs_callback_tcpport
Kinglong Mee [Thu, 24 Sep 2015 12:55:59 +0000 (20:55 +0800)]
NFS: Remove the left global variable nfs_callback_tcpport

Commit bbe0a3aa4e22 "NFS: make nfs_callback_tcpport per network context" has
make nfs_callback_tcpport per network, but left the global nfs_callback_tcpport,
remove it.

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoNFS: Get rid of the unneeded addr stored in callback arguments
Kinglong Mee [Thu, 24 Sep 2015 12:55:21 +0000 (20:55 +0800)]
NFS: Get rid of the unneeded addr stored in callback arguments

Commit c36fca52f5 "NFS refactor nfs_find_client and reference client
across callback processing" has store clp in cb_process_state
which is set in cb_sequence.

So that, it's unneeded to store address pointer in any callback arguments.

Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfsroot: make nfsroot to accept the 1024 bytes long directory name
Li RongQing [Thu, 24 Sep 2015 02:19:09 +0000 (10:19 +0800)]
nfsroot: make nfsroot to accept the 1024 bytes long directory name

although NFS_MAXPATHLEN is defined to 1024, nfs client hopes to accept
a 1024 byte path, but nfs_root_parms is limited to 256, and the nfs path
will truncated when a user inputs nfs path from kernel cmdline

enlarge nfs_root_parms to 1024, to make it accept the 1024 bytes long
directory name, since nfs_root_parms is defined as _initdata, it will
be released after system bootup

Signed-off-by: Li RongQing <roy.qing.li@gmail.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoMerge branch 'nfsclone'
Trond Myklebust [Wed, 21 Oct 2015 20:42:20 +0000 (15:42 -0500)]
Merge branch 'nfsclone'

* nfsclone:
  nfs: add missing linux/types.h
  NFS: Fix an 'unused variable' complaint when #ifndef CONFIG_NFS_V4_2
  nfs42: add NFS_IOC_CLONE_RANGE ioctl
  nfs42: respect clone_blksize
  nfs: get clone_blksize when probing fsinfo
  nfs42: add NFS_IOC_CLONE ioctl
  nfs42: add CLONE proc functions
  nfs42: add CLONE xdr functions

9 years agonfs: add missing linux/types.h
Peng Tao [Tue, 20 Oct 2015 19:18:16 +0000 (03:18 +0800)]
nfs: add missing linux/types.h

After merging the nfs tree, today's linux-next build (powerpc allyesconfig
produced this warning:

./usr/include/linux/nfs.h:40: found __[us]{8,16,32,64} type without #include <linux/types.h>

Reported-by: Stephen Rothwell <sfr@canb.auug.org.au>
Signed-off-by: Peng Tao <tao.peng@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agoLinux 4.3-rc6
Linus Torvalds [Sun, 18 Oct 2015 23:08:42 +0000 (16:08 -0700)]
Linux 4.3-rc6

9 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sun, 18 Oct 2015 19:07:48 +0000 (12:07 -0700)]
Merge branch 'i2c/for-current' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "Here are some bugfixes for the I2C subsystem.

  Kieran found a flaw in the recently renewed wake irq handling.  Mika
  handled a user bug report where the ACPI info turned out to be
  unusable.  I updated MAINTAINERS so that such bug reports will sooner
  get to the right people.  Geert pointed me to a problem of some i2c
  drivers regarding PM which I fixed"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  i2c: designware: Do not use parameters from ACPI on Dell Inspiron 7348
  MAINTAINERS: add maintainers for Synopsis Designware I2C drivers
  i2c: designware-platdrv: enable RuntimePM before registering to the core
  i2c: s3c2410: enable RuntimePM before registering to the core
  i2c: rcar: enable RuntimePM before registering to the core
  i2c: return probe deferred status on dev_pm_domain_attach

9 years agoi2c: designware: Do not use parameters from ACPI on Dell Inspiron 7348
Mika Westerberg [Thu, 24 Sep 2015 09:06:54 +0000 (12:06 +0300)]
i2c: designware: Do not use parameters from ACPI on Dell Inspiron 7348

ACPI SSCN/FMCN methods were originally added because then the platform can
provide the most accurate HCNT/LCNT values to the driver. However, this
seems not to be true for Dell Inspiron 7348 where using these causes the
touchpad to fail in boot:

  i2c_hid i2c-DLL0675:00: failed to retrieve report from device.
  i2c_designware INT3433:00: i2c_dw_handle_tx_abort: lost arbitration
  i2c_hid i2c-DLL0675:00: failed to retrieve report from device.
  i2c_designware INT3433:00: controller timed out

The values received from ACPI are (in fast mode):

  HCNT: 72
  LCNT: 160

this translates to following timings (input clock is 100MHz on Broadwell):

  tHIGH: 720 ns (spec min 600 ns)
  tLOW: 1600 ns (spec min 1300 ns)
  Bus period: 2920 ns (assuming 300 ns tf and tr)
  Bus speed: 342.5 kHz

Both tHIGH and tLOW are within the I2C specification.

The calculated values when ACPI parameters are not used are (in fast mode):

  HCNT: 87
  LCNT: 159

which translates to:

  tHIGH: 870 ns (spec min 600 ns)
  tLOW: 1590 ns (spec min 1300 ns)
  Bus period 3060 ns (assuming 300 ns tf and tr)
  Bus speed 326.8 kHz

These values are also within the I2C specification.

Since both ACPI and calculated values meet the I2C specification timing
requirements it is hard to say why the touchpad does not function properly
with the ACPI values except that the bus speed is higher in this case (but
still well below the max 400kHz).

Solve this by adding DMI quirk to the driver that disables using ACPI
parameters on this particulare machine.

Reported-by: Pavel Roskin <plroskin@gmail.com>
Signed-off-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Tested-by: Pavel Roskin <plroskin@gmail.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Cc: stable@kernel.org
9 years agoMerge branches 'irq-urgent-for-linus' and 'timers-urgent-for-linus' of git://git...
Linus Torvalds [Sat, 17 Oct 2015 15:47:27 +0000 (08:47 -0700)]
Merge branches 'irq-urgent-for-linus' and 'timers-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull irq/timer fixes from Thomas Gleixner:
 "irq: a fix for the new hierarchical MSI interrupt handling which
  unbreaks PCI=n configurations.

  timers: a fix for the new hrtimer clock offset update mechanism to
  ensure that the boot time offset is respected"

* 'irq-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  genirq/msi: Do not use pci_msi_[un]mask_irq as default methods

* 'timers-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  timekeeping: Increment clock_was_set_seq in timekeeping_init()

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sat, 17 Oct 2015 00:39:27 +0000 (17:39 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixes from Dmitry Torokhov:
 "Just two small fixups to ads7846 touchscreen controller driver and
  Cypress touchpad driver"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: cyapa - fix the copy paste error on electrodes_rx value
  Input: ads7846 - correct the value got from SPI

9 years agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 17 Oct 2015 00:11:14 +0000 (17:11 -0700)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fix from Stephen Boyd:
 "Just one revert for Armada XP devices: the conversion to
  of_clk_get_parent_name() wasn't a direct translation, so we
  revert back to of_clk_get() + __clk_get_name().

  We could make of_clk_get_parent_name() more robust, but that
  may have unintended side-effects, so we'll do that in the
  next version"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  Partially revert "clk: mvebu: Convert to clk_hw based provider APIs"

9 years agoMerge tag 'dm-4.3-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/device...
Linus Torvalds [Fri, 16 Oct 2015 20:03:05 +0000 (13:03 -0700)]
Merge tag 'dm-4.3-fixes-3' of git://git./linux/kernel/git/device-mapper/linux-dm

Pull device mapper fixes from Mike Snitzer:
 "Two DM target error path cleanup fixes (one for stable in DM thinp and
  one for a v4.3-rc5 thinko in DM snapshot)"

* tag 'dm-4.3-fixes-3' of git://git.kernel.org/pub/scm/linux/kernel/git/device-mapper/linux-dm:
  dm thin: fix missing pool reference count decrement in pool_ctr error path
  dm snapshot persistent: fix missing cleanup in persistent_ctr error path

9 years agoMerge branch 'for-linus-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/mason...
Linus Torvalds [Fri, 16 Oct 2015 19:55:34 +0000 (12:55 -0700)]
Merge branch 'for-linus-4.3' of git://git./linux/kernel/git/mason/linux-btrfs

Pull btrfs fixes from Chris Mason:
 "I have two more bug fixes for btrfs.

  My commit fixes a bug we hit last week at FB, a combination of lots of
  hard links and an admin command to resolve inode numbers.

  Dave is adding checks to make sure balance on current kernels ignores
  filters it doesn't understand.  The penalty for being wrong is just
  doing more work (not crashing etc), but it's a good fix"

* 'for-linus-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs:
  btrfs: fix use after free iterating extrefs
  btrfs: check unsupported filters in balance arguments

9 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph...
Linus Torvalds [Fri, 16 Oct 2015 19:47:02 +0000 (12:47 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/sage/ceph-client

Pull Ceph fixes from Sage Weil:
 "Just two small items from Ilya:

  The first patch fixes the RBD readahead to grab full objects.  The
  second fixes the write ops to prevent undue promotion when a cache
  tier is configured on the server side"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
  rbd: use writefull op for object size writes
  rbd: set max_sectors explicitly

9 years agoMerge tag 'pm+acpi-4.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael...
Linus Torvalds [Fri, 16 Oct 2015 19:25:54 +0000 (12:25 -0700)]
Merge tag 'pm+acpi-4.3-rc6' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management and ACPI fixes from Rafael Wysocki:
 "These fix two recent regressions (ACPICA, the generic power domains
  framework) and one crash that may happen on specific hardware
  supported since 4.1 (intel_pstate).

  Specifics:

   - Fix a regression introduced by a recent ACPICA cleanup that
     uncovered a latent bug (Lv Zheng).

   - Fix a recent regression in the generic power domains framework that
     may cause it to violate PM QoS latency constraints in some cases
     (Ulf Hansson).

   - Fix an intel_pstate driver crash on the Knights Landing chips that
     do not update the MPERF counter as often as expected by the driver
     which may result in a divide by 0 (Srinivas Pandruvada)"

* tag 'pm+acpi-4.3-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  cpufreq: intel_pstate: Fix divide by zero on Knights Landing (KNL)
  ACPICA: Tables: Fix FADT dependency regression
  PM / Domains: Fix validation of latency constraints in genpd governor

9 years agoMerge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux
Linus Torvalds [Fri, 16 Oct 2015 19:19:11 +0000 (12:19 -0700)]
Merge branch 'drm-fixes' of git://people.freedesktop.org/~airlied/linux

Pull drm fixes from Dave Airlie:
 "Nothing too crazy or exciting:

   - two MAINTAINERS entries that I didn't see the point in delaying.
   - one drm mst fix to stop sending uninitialised data to monitors
   - two amdgpu fixes
   - one radeon mst tiling fix
   - one vmwgfx regression fix
   - one virtio warning fix.

  I have found one locking problem that needs a bit of reorg to fix, but
  I'm not sure it's worth putting in -fixes as I don't think we've seen
  it hit in the real world ever, I just found it using the virtio-gpu
  driver when working on it.  I'll possibly send it next week once I've
  time to discuss with Daniel"

* 'drm-fixes' of git://people.freedesktop.org/~airlied/linux:
  drm/virtio: use %llu format string form atomic64_t
  MAINTAINERS: Add myself as maintainer for the gma500 driver
  MAINTAINERS: add a maintainer for the atmel-hlcdc DRM driver
  drm/amdgpu: Keep the pflip interrupts always enabled v7
  drm/amdgpu: adjust default dispclk (v2)
  drm/dp/mst: make mst i2c transfer code more robust.
  drm/radeon: attach tile property to mst connector
  drm/vmwgfx: Fix kernel NULL pointer dereference on older hardware

9 years agoMerge tag 'powerpc-4.3-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Fri, 16 Oct 2015 19:07:43 +0000 (12:07 -0700)]
Merge tag 'powerpc-4.3-4' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 - Re-enable CONFIG_SCSI_DH in our defconfigs
 - Remove unused os_area_db_id_video_mode
 - cxl: fix leak of IRQ names in cxl_free_afu_irqs() from Andrew
 - cxl: fix leak of ctx->irq_bitmap when releasing context via kernel API from Andrew
 - cxl: fix leak of ctx->mapping when releasing kernel API contexts from Andrew
 - cxl: Workaround malformed pcie packets on some cards from Philippe
 - cxl: Fix number of allocated pages in SPA from Christophe Lombard
 - Fix checkstop in native_hpte_clear() with lockdep from Cyril
 - Panic on unhandled Machine Check on powernv from Daniel
 - selftests/powerpc: Fix build failure of load_unaligned_zeropad test

* tag 'powerpc-4.3-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  selftests/powerpc: Fix build failure of load_unaligned_zeropad test
  powerpc/powernv: Panic on unhandled Machine Check
  powerpc: Fix checkstop in native_hpte_clear() with lockdep
  cxl: Fix number of allocated pages in SPA
  cxl: Workaround malformed pcie packets on some cards
  cxl: fix leak of ctx->mapping when releasing kernel API contexts
  cxl: fix leak of ctx->irq_bitmap when releasing context via kernel API
  cxl: fix leak of IRQ names in cxl_free_afu_irqs()
  powerpc/ps3: Remove unused os_area_db_id_video_mode
  powerpc/configs: Re-enable CONFIG_SCSI_DH

9 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 16 Oct 2015 18:42:37 +0000 (11:42 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "6 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  sh: add copy_user_page() alias for __copy_user()
  lib/Kconfig: ZLIB_DEFLATE must select BITREVERSE
  mm, dax: fix DAX deadlocks
  memcg: convert threshold to bytes
  builddeb: remove debian/files before build
  mm, fs: obey gfp_mapping for add_to_page_cache()

9 years agosh: add copy_user_page() alias for __copy_user()
Ross Zwisler [Thu, 15 Oct 2015 22:28:38 +0000 (15:28 -0700)]
sh: add copy_user_page() alias for __copy_user()

copy_user_page() is needed by DAX.  Without this we get a compile error
for DAX on SH:

  fs/dax.c:280:2: error: implicit declaration of function `copy_user_page' [-Werror=implicit-function-declaration]
    copy_user_page(vto, (void __force *)vfrom, vaddr, to);
      ^

This was done with a random config that happened to include DAX support.

This patch has only been compile tested.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Matthew Wilcox <willy@linux.intel.com>
Cc: Matt Fleming <matt@codeblueprint.co.uk>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agolib/Kconfig: ZLIB_DEFLATE must select BITREVERSE
Andrew Morton [Thu, 15 Oct 2015 22:28:35 +0000 (15:28 -0700)]
lib/Kconfig: ZLIB_DEFLATE must select BITREVERSE

lib/built-in.o: In function `__bitrev32':
deftree.c:(.text+0x1e799): undefined reference to `byte_rev_table'
deftree.c:(.text+0x1e7a0): undefined reference to `byte_rev_table'
deftree.c:(.text+0x1e7b4): undefined reference to `byte_rev_table'
deftree.c:(.text+0x1e7c1): undefined reference to `byte_rev_table'

Anything which uses bitrevX() has to select BITREVERSE, to grab
lib/bitrev.o.

Reported-by: Jim Davis <jim.epost@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomm, dax: fix DAX deadlocks
Ross Zwisler [Thu, 15 Oct 2015 22:28:32 +0000 (15:28 -0700)]
mm, dax: fix DAX deadlocks

The following two locking commits in the DAX code:

commit 843172978bb9 ("dax: fix race between simultaneous faults")
commit 46c043ede471 ("mm: take i_mmap_lock in unmap_mapping_range() for DAX")

introduced a number of deadlocks and other issues which need to be fixed
for the v4.3 kernel.  The list of issues in DAX after these commits
(some newly introduced by the commits, some preexisting) can be found
here:

  https://lkml.org/lkml/2015/9/25/602 (Subject: "Re: [PATCH] dax: fix deadlock in __dax_fault").

This undoes most of the changes introduced by those two commits,
essentially returning us to the DAX locking scheme that was used in
v4.2.

Signed-off-by: Ross Zwisler <ross.zwisler@linux.intel.com>
Cc: Alexander Viro <viro@zeniv.linux.org.uk>
Cc: Dan Williams <dan.j.williams@intel.com>
Tested-by: Dave Chinner <dchinner@redhat.com>
Cc: Jan Kara <jack@suse.com>
Cc: "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>
Cc: Matthew Wilcox <matthew.r.wilcox@intel.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomemcg: convert threshold to bytes
Shaohua Li [Thu, 15 Oct 2015 22:28:29 +0000 (15:28 -0700)]
memcg: convert threshold to bytes

page_counter_memparse() returns pages for the threshold, while
mem_cgroup_usage() returns bytes for memory usage.  Convert the
threshold to bytes.

Fixes: 3e32cb2e0a12b6915 ("memcg: rename cgroup_event to mem_cgroup_event").
Signed-off-by: Shaohua Li <shli@fb.com>
Cc: Johannes Weiner <hannes@cmpxchg.org>
Acked-by: Michal Hocko <mhocko@suse.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agobuilddeb: remove debian/files before build
Riku Voipio [Thu, 15 Oct 2015 22:28:26 +0000 (15:28 -0700)]
builddeb: remove debian/files before build

Commit 3716001bcb7f ("deb-pkg: add source package") added the ability to
create a debian changelog file.  This exposed that previously the
builddeb script hasn't cleared debian/files between builds.

As debian/files keeps accumulating entries, the changes file will end up
growing indefinelty.  With outdated entries in debian/files, builddeb
script will exit with failure.  This regression impacts those who use
"make deb-pkg" target to build kernel into a .deb package and never use
"make mrproper" or other means to clean kernel tree from generated
directories.

To fix the regression, remove debian/files before starting build and in
the generated clean rule.

Fixes: 3716001bcb7f ("deb-pkg: add source package")
Signed-off-by: Riku Voipio <riku.voipio@linaro.org>
Reported-by: Doug Smythies <dsmythies@telus.net>
Tested-by: Doug Smythies <dsmythies@telus.net>
Tested-by: Kalle Valo <kvalo@codeaurora.org>
Acked-by: Ben Hutchings <ben@decadent.org.uk>
Cc: Michal Marek <mmarek@suse.cz>
Cc: maximilian attems <maks@stro.at>
Cc: Chris J Arges <chris.j.arges@canonical.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agomm, fs: obey gfp_mapping for add_to_page_cache()
Michal Hocko [Thu, 15 Oct 2015 22:28:24 +0000 (15:28 -0700)]
mm, fs: obey gfp_mapping for add_to_page_cache()

Commit 6afdb859b710 ("mm: do not ignore mapping_gfp_mask in page cache
allocation paths") has caught some users of hardcoded GFP_KERNEL used in
the page cache allocation paths.  This, however, wasn't complete and
there were others which went unnoticed.

Dave Chinner has reported the following deadlock for xfs on loop device:
: With the recent merge of the loop device changes, I'm now seeing
: XFS deadlock on my single CPU, 1GB RAM VM running xfs/073.
:
: The deadlocked is as follows:
:
: kloopd1: loop_queue_read_work
:       xfs_file_iter_read
:       lock XFS inode XFS_IOLOCK_SHARED (on image file)
:       page cache read (GFP_KERNEL)
:       radix tree alloc
:       memory reclaim
:       reclaim XFS inodes
:       log force to unpin inodes
:       <wait for log IO completion>
:
: xfs-cil/loop1: <does log force IO work>
:       xlog_cil_push
:       xlog_write
:       <loop issuing log writes>
:               xlog_state_get_iclog_space()
:               <blocks due to all log buffers under write io>
:               <waits for IO completion>
:
: kloopd1: loop_queue_write_work
:       xfs_file_write_iter
:       lock XFS inode XFS_IOLOCK_EXCL (on image file)
:       <wait for inode to be unlocked>
:
: i.e. the kloopd, with it's split read and write work queues, has
: introduced a dependency through memory reclaim. i.e. that writes
: need to be able to progress for reads make progress.
:
: The problem, fundamentally, is that mpage_readpages() does a
: GFP_KERNEL allocation, rather than paying attention to the inode's
: mapping gfp mask, which is set to GFP_NOFS.
:
: The didn't used to happen, because the loop device used to issue
: reads through the splice path and that does:
:
:       error = add_to_page_cache_lru(page, mapping, index,
:                       GFP_KERNEL & mapping_gfp_mask(mapping));

This has changed by commit aa4d86163e4 ("block: loop: switch to VFS
ITER_BVEC").

This patch changes mpage_readpage{s} to follow gfp mask set for the
mapping.  There are, however, other places which are doing basically the
same.

lustre:ll_dir_filler is doing GFP_KERNEL from the function which
apparently uses GFP_NOFS for other allocations so let's make this
consistent.

cifs:readpages_get_pages is called from cifs_readpages and
__cifs_readpages_from_fscache called from the same path obeys mapping
gfp.

ramfs_nommu_expand_for_mapping is hardcoding GFP_KERNEL as well
regardless it uses mapping_gfp_mask for the page allocation.

ext4_mpage_readpages is the called from the page cache allocation path
same as read_pages and read_cache_pages

As I've noticed in my previous post I cannot say I would be happy about
sprinkling mapping_gfp_mask all over the place and it sounds like we
should drop gfp_mask argument altogether and use it internally in
__add_to_page_cache_locked that would require all the filesystems to use
mapping gfp consistently which I am not sure is the case here.  From a
quick glance it seems that some file system use it all the time while
others are selective.

Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: Dave Chinner <david@fromorbit.com>
Cc: "Theodore Ts'o" <tytso@mit.edu>
Cc: Ming Lei <ming.lei@canonical.com>
Cc: Andreas Dilger <andreas.dilger@intel.com>
Cc: Oleg Drokin <oleg.drokin@intel.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Christoph Hellwig <hch@lst.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agorbd: use writefull op for object size writes
Ilya Dryomov [Wed, 7 Oct 2015 15:27:17 +0000 (17:27 +0200)]
rbd: use writefull op for object size writes

This covers only the simplest case - an object size sized write, but
it's still useful in tiering setups when EC is used for the base tier
as writefull op can be proxied, saving an object promotion.

Even though updating ceph_osdc_new_request() to allow writefull should
just be a matter of fixing an assert, I didn't do it because its only
user is cephfs.  All other sites were updated.

Reflects ceph.git commit 7bfb7f9025a8ee0d2305f49bf0336d2424da5b5b.

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>
9 years agorbd: set max_sectors explicitly
Ilya Dryomov [Wed, 7 Oct 2015 14:09:35 +0000 (16:09 +0200)]
rbd: set max_sectors explicitly

Commit 30e2bc08b2bb ("Revert "block: remove artifical max_hw_sectors
cap"") restored a clamp on max_sectors.  It's now 2560 sectors instead
of 1024, but it's not good enough: we set max_hw_sectors to rbd object
size because we don't want object sized I/Os to be split, and the
default object size is 4M.

So, set max_sectors to max_hw_sectors in rbd at queue init time.

Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>
9 years agotimekeeping: Increment clock_was_set_seq in timekeeping_init()
Thomas Gleixner [Fri, 16 Oct 2015 13:50:22 +0000 (15:50 +0200)]
timekeeping: Increment clock_was_set_seq in timekeeping_init()

timekeeping_init() can set the wall time offset, so we need to
increment the clock_was_set_seq counter. That way hrtimers will pick
up the early offset immediately. Otherwise on a machine which does not
set wall time later in the boot process the hrtimer offset is stale at
0 and wall time timers are going to expire with a delay of 45 years.

Fixes: 868a3e915f7f "hrtimer: Make offset update smarter"
Reported-and-tested-by: Heiko Carstens <heiko.carstens@de.ibm.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Stefan Liebler <stli@linux.vnet.ibm.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: John Stultz <john.stultz@linaro.org>
9 years agoMerge branches 'acpica', 'pm-domains' and 'pm-cpufreq'
Rafael J. Wysocki [Fri, 16 Oct 2015 12:32:27 +0000 (14:32 +0200)]
Merge branches 'acpica', 'pm-domains' and 'pm-cpufreq'

* acpica:
  ACPICA: Tables: Fix FADT dependency regression

* pm-domains:
  PM / Domains: Fix validation of latency constraints in genpd governor

* pm-cpufreq:
  cpufreq: intel_pstate: Fix divide by zero on Knights Landing (KNL)

9 years agogenirq/msi: Do not use pci_msi_[un]mask_irq as default methods
Marc Zyngier [Tue, 13 Oct 2015 18:14:45 +0000 (19:14 +0100)]
genirq/msi: Do not use pci_msi_[un]mask_irq as default methods

When we create a generic MSI domain, that MSI_FLAG_USE_DEF_CHIP_OPS
is set, and that any of .mask or .unmask are NULL in the irq_chip
structure, we set them to pci_msi_[un]mask_irq.

This is a bad idea for at least two reasons:
- PCI_MSI might not be selected, kernel fails to build (yes, this is
  legitimate, at least on arm64!)
- This may not be a PCI/MSI domain at all (platform MSI, for example)

Either way, this looks wrong. Move the overriding of mask/unmask to
the PCI counterpart, and panic is any of these two methods is not
set in the core code (they really should be present).

Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Cc: Jiang Liu <jiang.liu@linux.intel.com>
Cc: Bjorn Helgaas <bhelgaas@google.com>
Link: http://lkml.kernel.org/r/1444760085-27857-1-git-send-email-marc.zyngier@arm.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
9 years agodrm/virtio: use %llu format string form atomic64_t
Arnd Bergmann [Wed, 7 Oct 2015 10:41:21 +0000 (12:41 +0200)]
drm/virtio: use %llu format string form atomic64_t

The virtgpu driver prints the last_seq variable using the %ld or
%lu format string, which does not work correctly on all architectures
and causes this compiler warning on ARM:

drivers/gpu/drm/virtio/virtgpu_fence.c: In function 'virtio_timeline_value_str':
drivers/gpu/drm/virtio/virtgpu_fence.c:64:22: warning: format '%lu' expects argument of type 'long unsigned int', but argument 4 has type 'long long int' [-Wformat=]
  snprintf(str, size, "%lu", atomic64_read(&fence->drv->last_seq));
                      ^
drivers/gpu/drm/virtio/virtgpu_debugfs.c: In function 'virtio_gpu_debugfs_irq_info':
drivers/gpu/drm/virtio/virtgpu_debugfs.c:37:16: warning: format '%ld' expects argument of type 'long int', but argument 3 has type 'long long int' [-Wformat=]
  seq_printf(m, "fence %ld %lld\n",
                ^

In order to avoid the warnings, this changes the format strings to %llu
and adds a cast to u64, which makes it work the same way everywhere.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoMAINTAINERS: Add myself as maintainer for the gma500 driver
Patrik Jakobsson [Mon, 12 Oct 2015 21:14:44 +0000 (23:14 +0200)]
MAINTAINERS: Add myself as maintainer for the gma500 driver

Signed-off-by: Patrik Jakobsson <patrik.r.jakobsson@gmail.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoMAINTAINERS: add a maintainer for the atmel-hlcdc DRM driver
Boris BREZILLON [Tue, 13 Oct 2015 09:55:42 +0000 (11:55 +0200)]
MAINTAINERS: add a maintainer for the atmel-hlcdc DRM driver

Add myself as the maintainer of the atmel-hlcdc DRM driver.

Signed-off-by: Boris Brezillon <boris.brezillon@free-electrons.com>
Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoMerge branch 'drm-fixes-4.3' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Fri, 16 Oct 2015 00:01:24 +0000 (10:01 +1000)]
Merge branch 'drm-fixes-4.3' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

Just two fixes for amdgpu:
- fix pageflip interrupt issue
- fix display clock handling on certain fiji boards

* 'drm-fixes-4.3' of git://people.freedesktop.org/~agd5f/linux:
  drm/amdgpu: Keep the pflip interrupts always enabled v7
  drm/amdgpu: adjust default dispclk (v2)

9 years agoMerge tag 'vmwgfx-fixes-4.3-151014' of git://people.freedesktop.org/~thomash/linux...
Dave Airlie [Thu, 15 Oct 2015 22:10:21 +0000 (08:10 +1000)]
Merge tag 'vmwgfx-fixes-4.3-151014' of git://people.freedesktop.org/~thomash/linux into drm-fixes

Pull request of 2015-10-14

* tag 'vmwgfx-fixes-4.3-151014' of git://people.freedesktop.org/~thomash/linux:
  drm/vmwgfx: Fix kernel NULL pointer dereference on older hardware

9 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Thu, 15 Oct 2015 21:03:38 +0000 (14:03 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Will Deacon:
 "Here are a few more arm64 fixes for 4.3.  Again, nothing too
  significant, but worth having nonetheless.  The MINSIGSTKSZ update is
  a bit grotty, but the value we currently have is wrong (too small), so
  anybody using that will have issues already.  It has Arnd's ack for
  the asm-generic change.

  Summary:

   - Fix module CFLAGS setting in workaround for erratum #843419
   - Update MINSIGSTKSZ and SIGSTKSZ to match glibc
   - Wire up some new compat syscalls"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64: compat: wire up new syscalls
  arm64: Fix MINSIGSTKSZ and SIGSTKSZ
  arm64: errata: use KBUILD_CFLAGS_MODULE for erratum #843419

9 years agoMerge tag 'pinctrl-v4.3-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw...
Linus Torvalds [Thu, 15 Oct 2015 20:58:22 +0000 (13:58 -0700)]
Merge tag 'pinctrl-v4.3-3' of git://git./linux/kernel/git/linusw/linux-pinctrl

Pull pinctrl fixes from Linus Walleij:
 "Here are some overdue (what can I say, I was on a short vacation)
  driver fixes for the pin control subsystem:

   - Allwinner sun5i A10s had a faulty mapping

   - Freescale i.MX25 had some bad arithmetics

   - Uniphier PH1-sLD8 missed some input enable settings"

* tag 'pinctrl-v4.3-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-pinctrl:
  pinctrl: uniphier: fix input enable settings for PH1-sLD8
  pinctrl: imx25: ensure that a pin with id i is at position i in the info array
  pinctrl: sun5i: Fix a10s pwm1 pinctrl mapping

9 years agocpufreq: intel_pstate: Fix divide by zero on Knights Landing (KNL)
Srinivas Pandruvada [Thu, 15 Oct 2015 19:34:21 +0000 (12:34 -0700)]
cpufreq: intel_pstate: Fix divide by zero on Knights Landing (KNL)

This is a workaround for KNL platform, where in some cases MPERF counter
will not have updated value before next read of MSR_IA32_MPERF. In this
case divide by zero will occur. This change ignores current sample for
busy calculation in this case.

Fixes: b34ef932d79a (intel_pstate: Knights Landing support)
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>
Acked-by: Kristen Carlson Accardi <kristen@linux.intel.com>
Cc: 4.1+ <stable@vger.kernel.org> # 4.1+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
9 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma
Linus Torvalds [Thu, 15 Oct 2015 20:44:35 +0000 (13:44 -0700)]
Merge tag 'for-linus' of git://git./linux/kernel/git/dledford/rdma

Pull rdma updates from Doug Ledford:
 "We have four batched up patches for the current rc kernel.

  Two of them are small fixes that are obvious.

  One of them is larger than I would like for a late stage rc pull, but
  we found an issue in the namespace lookup code related to RoCE and
  this works around the issue for now (we allow a lookup with a
  namespace to succeed on RoCE since RoCE namespaces aren't implemented
  yet).  This will go away in 4.4 when we put in support for namespaces
  in RoCE devices.

  The last one is large in terms of lines, but is all legal and no
  functional changes.  Cisco needed to update their files to be more
  specific about their license.  They had intended the files to be dual
  licensed as GPL/BSD all along, and specified that in their module
  license tag, but their file headers were not up to par.  They
  contacted all of the contributors to get agreement and then submitted
  a patch to update the license headers in the files.

  Summary:

   - Work around connection namespace lookup bug related to RoCE

   - Change usnic license to Dual GPL/BSD (was intended to be that way
     all along, but wasn't clear, permission from contributors was
     chased down)

   - Fix an issue between NFSoRDMA and mlx5 that could cause an oops

   - Fix leak of sendonly multicast groups"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma:
  IB/ipoib: For sendonly join free the multicast group on leave
  IB/cma: Accept connection without a valid netdev on RoCE
  xprtrdma: Don't require LOCAL_DMA_LKEY support for fastreg
  usnic: add missing clauses to BSD license

9 years agoMerge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs
Linus Torvalds [Thu, 15 Oct 2015 20:31:00 +0000 (13:31 -0700)]
Merge branch 'for_linus' of git://git./linux/kernel/git/jack/linux-fs

Pull ext4 Kconfig description fixup from Jan Kara:
 "A small fixup in description of EXT4_USE_FOR_EXT2 config option"

* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  ext4: Update EXT4_USE_FOR_EXT2 description

9 years agoNFS: Fix an 'unused variable' complaint when #ifndef CONFIG_NFS_V4_2
Trond Myklebust [Thu, 15 Oct 2015 20:20:11 +0000 (16:20 -0400)]
NFS: Fix an 'unused variable' complaint when #ifndef CONFIG_NFS_V4_2

Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfs42: add NFS_IOC_CLONE_RANGE ioctl
Peng Tao [Fri, 25 Sep 2015 18:24:39 +0000 (02:24 +0800)]
nfs42: add NFS_IOC_CLONE_RANGE ioctl

It follows btrfs BTRFS_IOC_CLONE_RANGE lead on ioctl number and
arguments.

Signed-off-by: Peng Tao <tao.peng@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfs42: respect clone_blksize
Peng Tao [Fri, 25 Sep 2015 18:24:38 +0000 (02:24 +0800)]
nfs42: respect clone_blksize

draft-ietf-nfsv4-minorversion2-38.txt says:
   Both cl_src_offset and
   cl_dst_offset must be aligned to the clone block size Section 12.2.1.
   The number of bytes to be cloned must be a multiple of the clone
   block size, except in the case in which cl_src_offset plus the number
   of bytes to be cloned is equal to the source file size.

Signed-off-by: Peng Tao <tao.peng@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfs: get clone_blksize when probing fsinfo
Peng Tao [Fri, 25 Sep 2015 18:24:37 +0000 (02:24 +0800)]
nfs: get clone_blksize when probing fsinfo

NFSv42 CLONE operation is supposed to respect it.

Signed-off-by: Peng Tao <tao.peng@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfs42: add NFS_IOC_CLONE ioctl
Peng Tao [Fri, 25 Sep 2015 18:24:36 +0000 (02:24 +0800)]
nfs42: add NFS_IOC_CLONE ioctl

It can be called by user space to CLONE two files.
Follow btrfs lead and define NFS_IOC_CLONE same as BTRFS_IOC_CLONE.
Thus we don't mess up userspace with too many ioctls.

Signed-off-by: Peng Tao <tao.peng@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfs42: add CLONE proc functions
Peng Tao [Fri, 25 Sep 2015 18:24:35 +0000 (02:24 +0800)]
nfs42: add CLONE proc functions

Signed-off-by: Peng Tao <tao.peng@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agonfs42: add CLONE xdr functions
Peng Tao [Fri, 25 Sep 2015 18:24:34 +0000 (02:24 +0800)]
nfs42: add CLONE xdr functions

xdr definitions per draft-ietf-nfsv4-minorversion2-38.txt

Signed-off-by: Peng Tao <tao.peng@primarydata.com>
Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
9 years agovmstat: explicitly schedule per-cpu work on the CPU we need it to run on
Linus Torvalds [Thu, 15 Oct 2015 20:01:50 +0000 (13:01 -0700)]
vmstat: explicitly schedule per-cpu work on the CPU we need it to run on

The vmstat code uses "schedule_delayed_work_on()" to do the initial
startup of the delayed work on the right CPU, but then once it was
started it would use the non-cpu-specific "schedule_delayed_work()" to
re-schedule it on that CPU.

That just happened to schedule it on the same CPU historically (well, in
almost all situations), but the code _requires_ this work to be per-cpu,
and should say so explicitly rather than depend on the non-cpu-specific
scheduling to schedule on the current CPU.

The timer code is being changed to not be as single-minded in always
running things on the calling CPU.

See also commit 874bbfe600a6 ("workqueue: make sure delayed work run in
local cpu") that for now maintains the local CPU guarantees just in case
there are other broken users that depended on the accidental behavior.

Cc: Christoph Lameter <cl@linux.com>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
9 years agoMerge branch 'for-4.3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq
Linus Torvalds [Thu, 15 Oct 2015 19:58:37 +0000 (12:58 -0700)]
Merge branch 'for-4.3-fixes' of git://git./linux/kernel/git/tj/wq

Pull workqueue fixlet from Tejun Heo:
 "Single patch to make delayed work always be queued on the local CPU"

This is not actually something we should guarantee, but it's something
we by accident have historically done, and at least one call site has
grown to depend on it.

I'm going to fix that known broken callsite, but in the meantime this
makes the accidental behavior be explicit, just in case there are other
cases that might depend on it.

* 'for-4.3-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj/wq:
  workqueue: make sure delayed work run in local cpu

9 years agoMAINTAINERS: add maintainers for Synopsis Designware I2C drivers
Wolfram Sang [Thu, 15 Oct 2015 13:50:15 +0000 (15:50 +0200)]
MAINTAINERS: add maintainers for Synopsis Designware I2C drivers

Those guys already have been helpful in the past and are actively
working on this driver, unlike me.

Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Acked-by: Jarkko Nikula <jarkko.nikula@linux.intel.com>
Acked-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
9 years agodrm/amdgpu: Keep the pflip interrupts always enabled v7
Michel Dänzer [Thu, 8 Oct 2015 08:48:04 +0000 (17:48 +0900)]
drm/amdgpu: Keep the pflip interrupts always enabled v7

This fixes flickering issues caused by prematurely firing pflip
interrupts.

v2 (chk): add commit message, fix DCE V10/V11 and DM as well
v3: Re-enable pflip interrupt wherever we re-enable a CRTC
v4: Enable pflip interrupt in DAL as well
v5: drop DAL changes for upstream
v6: (agd): only enable interrupts on crtcs that exist
v7: (agd): integrate suggestions from Michel

Signed-off-by: Michel Dänzer <michel.daenzer@amd.com>
Signed-off-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
9 years agodrm/amdgpu: adjust default dispclk (v2)
Alex Deucher [Mon, 12 Oct 2015 14:38:02 +0000 (10:38 -0400)]
drm/amdgpu: adjust default dispclk (v2)

Set the default to 600Mhz if it's not set in the bios,
and bump the default to 600Mhz if it's lower than that.

Port of radeon commit:
9368931db826d57b6b88b3145a00276626b48df0

v2: clean up the code a bit

bug:
https://bugs.freedesktop.org/show_bug.cgi?id=91896

Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
9 years agoi2c: designware-platdrv: enable RuntimePM before registering to the core
Wolfram Sang [Fri, 9 Oct 2015 09:39:24 +0000 (10:39 +0100)]
i2c: designware-platdrv: enable RuntimePM before registering to the core

The core may register clients attached to this master which may use
funtionality from the master. So, RuntimePM must be enabled before, otherwise
this will fail.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Acked-by: Mika Westerberg <mika.westerberg@linux.intel.com>
Cc: stable@kernel.org
9 years agoi2c: s3c2410: enable RuntimePM before registering to the core
Wolfram Sang [Sat, 10 Oct 2015 07:24:23 +0000 (08:24 +0100)]
i2c: s3c2410: enable RuntimePM before registering to the core

The core may register clients attached to this master which may use
funtionality from the master. So, RuntimePM must be enabled before, otherwise
this will fail. While here, move drvdata, too.

Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Tested-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Acked-by: Kukjin Kim <kgene@kernel.org>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Cc: stable@kernel.org
9 years agoi2c: rcar: enable RuntimePM before registering to the core
Wolfram Sang [Fri, 9 Oct 2015 09:39:25 +0000 (10:39 +0100)]
i2c: rcar: enable RuntimePM before registering to the core

The core may register clients attached to this master which may use
funtionality from the master. So, RuntimePM must be enabled before, otherwise
this will fail. While here, move drvdata, too.

Reported-by: Geert Uytterhoeven <geert+renesas@glider.be>
Signed-off-by: Wolfram Sang <wsa+renesas@sang-engineering.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
Cc: stable@kernel.org
9 years agoi2c: return probe deferred status on dev_pm_domain_attach
Kieran Bingham [Mon, 12 Oct 2015 20:54:43 +0000 (21:54 +0100)]
i2c: return probe deferred status on dev_pm_domain_attach

A change of return status was introduced in commit 3fffd1283927
("i2c: allow specifying separate wakeup interrupt in device tree")

The commit prevents the defer status being passed up the call stack
appropriately when dev_pm_domain_attach returns -EPROBE_DEFER.

Catch the PROBE_DEFER and clear up the IRQ wakeup status

Signed-off-by: Kieran Bingham <kieranbingham@gmail.com>
Fixes: 3fffd1283927 ("i2c: allow specifying separate wakeup interrupt in device tree")
Reviewed-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
Signed-off-by: Wolfram Sang <wsa@the-dreams.de>
9 years agodrm/dp/mst: make mst i2c transfer code more robust.
Dave Airlie [Wed, 14 Oct 2015 08:51:17 +0000 (18:51 +1000)]
drm/dp/mst: make mst i2c transfer code more robust.

This zeroes the msg so no random stack data ends up getting
sent, it also limits the function to not accepting > 4
i2c msgs.

Cc: stable@vger.kernel.org
Reviewed-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agodrm/radeon: attach tile property to mst connector
Dave Airlie [Wed, 14 Oct 2015 23:04:21 +0000 (09:04 +1000)]
drm/radeon: attach tile property to mst connector

This allows tiled monitors to work with radeon once mst is enabled.

Cc: stable@vger.kernel.org
Signed-off-by: Dave Airlie <airlied@redhat.com>
9 years agoACPICA: Tables: Fix FADT dependency regression
Lv Zheng [Wed, 14 Oct 2015 05:53:57 +0000 (13:53 +0800)]
ACPICA: Tables: Fix FADT dependency regression

Some logics actually relying on the existence of FADT, currently relies on
the number of loaded tables. This false dependency can easily trigger
regressions. One of them has been introduced by commit 8ec3f459073e
(ACPICA: Tables: Fix global table list issues by removing fixed table).

The commit changing the fixed table indexes results in the change of FADT
table index, originally, it was 3 (thus the installed table count should be
greater than 4), while currently it is 0 (and the installed table count may
be 3).

This patch fixes this regression by cleaning up the code. Lv Zheng.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=105351
Fixes: 8ec3f459073e (ACPICA: Tables: Fix global table list issues by removing fixed table)
Reported-and-tested-by: Meelis Roos <mroos@linux.ee>
Signed-off-by: Lv Zheng <lv.zheng@intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
9 years agoPartially revert "clk: mvebu: Convert to clk_hw based provider APIs"
Stephen Boyd [Wed, 14 Oct 2015 18:23:09 +0000 (11:23 -0700)]
Partially revert "clk: mvebu: Convert to clk_hw based provider APIs"

This partially reverts commit eca61c9ff2588e1df373e61078e1874976315839.

Thomas reports that it causes regressions on Armada XP devices.
This is because of_clk_get_parent_name() relies on the property
'clock-output-names' to resolve the name of a clock's parent,
without trying to get the clock from the framework and call
__clk_get_name(). Given that Armada XP devices don't have the
'clock-output-names' property, of_clk_get_parent_name() returns
the name of the node which doesn't match the actual parent
clock's name at all, causing CPU clocks to never link up with
their parents.

Reported-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
9 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Wed, 14 Oct 2015 17:01:32 +0000 (10:01 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull KVM fixes from Paolo Bonzini:
 "Bug fixes for system management mode emulation.

  The first two patches fix SMM emulation on Nehalem processors.  The
  others fix some cases that became apparent as work progressed on the
  firmware side"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: x86: fix RSM into 64-bit protected mode
  KVM: x86: fix previous commit for 32-bit
  KVM: x86: fix SMI to halted VCPU
  KVM: x86: clean up kvm_arch_vcpu_runnable
  KVM: x86: map/unmap private slots in __x86_set_memory_region
  KVM: x86: build kvm_userspace_memory_region in x86_set_memory_region

9 years agoMerge tag 'mfd-fixes-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/mfd
Linus Torvalds [Wed, 14 Oct 2015 16:47:38 +0000 (09:47 -0700)]
Merge tag 'mfd-fixes-4.3' of git://git./linux/kernel/git/lee/mfd

Pull MFD fixes from Lee Jones:
 "Bug Fixes:

   - Return correct error code i.e. not zero

   - Fix build error when !CONFIG_PM_SLEEP"

* tag 'mfd-fixes-4.3' of git://git.kernel.org/pub/scm/linux/kernel/git/lee/mfd:
  mfd: max77843: Fix max77843_chg_init() return on error
  mfd: intel-lpss: Fix build error when !CONFIG_PM_SLEEP

9 years agoKVM: x86: fix RSM into 64-bit protected mode
Paolo Bonzini [Wed, 14 Oct 2015 13:25:52 +0000 (15:25 +0200)]
KVM: x86: fix RSM into 64-bit protected mode

In order to get into 64-bit protected mode, you need to enable
paging while EFER.LMA=1.  For this to work, CS.L must be 0.
Currently, we load the segments before CR0 and CR4, which means
that if RSM returns into 64-bit protected mode CS.L is already 1
and everything breaks.

Luckily, CS.L=0 is always the case when executing RSM, because it
is forbidden to execute RSM from 64-bit protected mode.  Hence it
is enough to load CR0 and CR4 first, and only then the segments.

Fixes: 660a5d517aaab9187f93854425c4c63f4a09195c
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoKVM: x86: fix previous commit for 32-bit
Paolo Bonzini [Wed, 14 Oct 2015 13:51:08 +0000 (15:51 +0200)]
KVM: x86: fix previous commit for 32-bit

Unfortunately I only noticed this after pushing.

Fixes: f0d648bdf0a5bbc91da6099d5282f77996558ea4
Cc: stable@vger.kernel.org
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
9 years agoarm64: compat: wire up new syscalls
Will Deacon [Wed, 14 Oct 2015 12:20:36 +0000 (13:20 +0100)]
arm64: compat: wire up new syscalls

Commit 208473c1f3ac ("ARM: wire up new syscalls") hooked up the new
userfaultfd and membarrier syscalls for ARM, so do the same for our
compat syscall table in arm64.

Signed-off-by: Will Deacon <will.deacon@arm.com>
9 years agoInput: cyapa - fix the copy paste error on electrodes_rx value
Dudley Du [Mon, 12 Oct 2015 17:16:16 +0000 (10:16 -0700)]
Input: cyapa - fix the copy paste error on electrodes_rx value

Fix the copy paste error on the electrodes_rx value set code which will
cause the electrodes_rx value be always set to the value of electrodes_y.

Reported-by: Sudip Mukherjee <sudipm.mukherjee@gmail.com>
Signed-off-by: Dudley Du <dudl@cypress.com>
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
9 years agobtrfs: fix use after free iterating extrefs
Chris Mason [Tue, 13 Oct 2015 18:06:48 +0000 (14:06 -0400)]
btrfs: fix use after free iterating extrefs

The code for btrfs inode-resolve has never worked properly for
files with enough hard links to trigger extrefs.  It was trying to
get the leaf out of a path after freeing the path:

btrfs_release_path(path);
leaf = path->nodes[0];
item_size = btrfs_item_size_nr(leaf, slot);

The fix here is to use the extent buffer we cloned just a little higher
up to avoid deadlocks caused by using the leaf in the path.

Signed-off-by: Chris Mason <clm@fb.com>
cc: stable@vger.kernel.org # v3.7+
cc: Mark Fasheh <mfasheh@suse.de>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Mark Fasheh <mfasheh@suse.de>