sfrench/cifs-2.6.git
5 years agonvme-multipath: drop optimization for static ANA group IDs
Hannes Reinecke [Wed, 9 Jan 2019 08:45:15 +0000 (09:45 +0100)]
nvme-multipath: drop optimization for static ANA group IDs

Bit 6 in the ANACAP field is used to indicate that the ANA group ID
doesn't change while the namespace is attached to the controller.
There is an optimisation in the code to only allocate space
for the ANA group header, as the namespace list won't change and
hence would not need to be refreshed.
However, this optimisation was never carried over to the actual
workflow, which always assumes that the buffer is large enough
to hold the ANA header _and_ the namespace list.
So drop this optimisation and always allocate enough space.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Hannes Reinecke <hare@suse.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agonvmet-rdma: fix null dereference under heavy load
Raju Rangoju [Thu, 3 Jan 2019 17:35:31 +0000 (23:05 +0530)]
nvmet-rdma: fix null dereference under heavy load

Under heavy load if we don't have any pre-allocated rsps left, we
dynamically allocate a rsp, but we are not actually allocating memory
for nvme_completion (rsp->req.rsp). In such a case, accessing pointer
fields (req->rsp->status) in nvmet_req_init() will result in crash.

To fix this, allocate the memory for nvme_completion by calling
nvmet_rdma_alloc_rsp()

Fixes: 8407879c("nvmet-rdma:fix possible bogus dereference under heavy load")
Cc: <stable@vger.kernel.org>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Raju Rangoju <rajur@chelsio.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agonvme-rdma: rework queue maps handling
Sagi Grimberg [Sat, 19 Jan 2019 00:43:24 +0000 (16:43 -0800)]
nvme-rdma: rework queue maps handling

If the device supports less queues than provided (if the device has less
completion vectors), we might hit a bug due to the fact that we ignore
that in nvme_rdma_map_queues (we override the maps nr_queues with user
opts).

Instead, keep track of how many default/read/poll queues we actually
allocated (rather than asked by the user) and use that to assign our
queue mappings.

Fixes: b65bb777ef22 (" nvme-rdma: support separate queue maps for read and write")
Reported-by: Saleem, Shiraz <shiraz.saleem@intel.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agonvme-tcp: fix timeout handler
Sagi Grimberg [Tue, 8 Jan 2019 09:01:30 +0000 (01:01 -0800)]
nvme-tcp: fix timeout handler

Currently, we have several problems with the timeout
handler:
1. If we timeout on the controller establishment flow, we will hang
because we don't execute the error recovery (and we shouldn't because
the create_ctrl flow needs to fail and cleanup on its own)
2. We might also hang if we get a disconnet on a queue while the
controller is already deleting. This racy flow can cause the controller
disable/shutdown admin command to hang.

We cannot complete a timed out request from the timeout handler without
mutual exclusion from the teardown flow (e.g. nvme_rdma_error_recovery_work).
So we serialize it in the timeout handler and teardown io and admin
queues to guarantee that no one races with us from completing the
request.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agonvme-rdma: fix timeout handler
Sagi Grimberg [Tue, 8 Jan 2019 08:53:22 +0000 (00:53 -0800)]
nvme-rdma: fix timeout handler

Currently, we have several problems with the timeout
handler:
1. If we timeout on the controller establishment flow, we will hang
because we don't execute the error recovery (and we shouldn't because
the create_ctrl flow needs to fail and cleanup on its own)
2. We might also hang if we get a disconnet on a queue while the
controller is already deleting. This racy flow can cause the controller
disable/shutdown admin command to hang.

We cannot complete a timed out request from the timeout handler without
mutual exclusion from the teardown flow (e.g. nvme_rdma_error_recovery_work).
So we serialize it in the timeout handler and teardown io and admin
queues to guarantee that no one races with us from completing the
request.

Reported-by: Jaesoo Lee <jalee@purestorage.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoarm64/xen: fix xen-swiotlb cache flushing
Christoph Hellwig [Wed, 16 Jan 2019 18:01:48 +0000 (19:01 +0100)]
arm64/xen: fix xen-swiotlb cache flushing

Xen-swiotlb hooks into the arm/arm64 arch code through a copy of the DMA
DMA mapping operations stored in the struct device arch data.

Switching arm64 to use the direct calls for the merged DMA direct /
swiotlb code broke this scheme.  Replace the indirect calls with
direct-calls in xen-swiotlb as well to fix this problem.

Fixes: 356da6d0cde3 ("dma-mapping: bypass indirect calls for dma-direct")
Reported-by: Julien Grall <julien.grall@arm.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
5 years agoRevert "Change mincore() to count "mapped" pages rather than "cached" pages"
Linus Torvalds [Wed, 23 Jan 2019 20:04:37 +0000 (09:04 +1300)]
Revert "Change mincore() to count "mapped" pages rather than "cached" pages"

This reverts commit 574823bfab82d9d8fa47f422778043fbb4b4f50e.

It turns out that my hope that we could just remove the code that
exposes the cache residency status from mincore() was too optimistic.

There are various random users that want it, and one example would be
the Netflix database cluster maintenance. To quote Josh Snyder:

 "For Netflix, losing accurate information from the mincore syscall
  would lengthen database cluster maintenance operations from days to
  months. We rely on cross-process mincore to migrate the contents of a
  page cache from machine to machine, and across reboots.

  To do this, I wrote and maintain happycache [1], a page cache
  dumper/loader tool. It is quite similar in architecture to pgfincore,
  except that it is agnostic to workload. The gist of happycache's
  operation is "produce a dump of residence status for each page, do
  some operation, then reload exactly the same pages which were present
  before." happycache is entirely dependent on accurate reporting of the
  in-core status of file-backed pages, as accessed by another process.

  We primarily use happycache with Cassandra, which (like Postgres +
  pgfincore) relies heavily on OS page cache to reduce disk accesses.
  Because our workloads never experience a cold page cache, we are able
  to provision hardware for a peak utilization level that is far lower
  than the hypothetical "every query is a cache miss" peak.

  A database warmed by happycache can be ready for service in seconds
  (bounded only by the performance of the drives and the I/O subsystem),
  with no period of in-service degradation. By contrast, putting a
  database in service without a page cache entails a potentially
  unbounded period of degradation (at Netflix, the time to populate a
  single node's cache via natural cache misses varies by workload from
  hours to weeks). If a single node upgrade were to take weeks, then
  upgrading an entire cluster would take months. Since we want to apply
  security upgrades (and other things) on a somewhat tighter schedule,
  we would have to develop more complex solutions to provide the same
  functionality already provided by mincore.

  At the bottom line, happycache is designed to benignly exploit the
  same information leak documented in the paper [2]. I think it makes
  perfect sense to remove cross-process mincore functionality from
  unprivileged users, but not to remove it entirely"

We do have an alternate approach that limits the cache residency
reporting only to processes that have write permissions to the file, so
we can fix the original information leak issue that way.  It involves
_adding_ code rather than removing it, which is sad, but hey, at least
we haven't found any users that would find the restrictions
unacceptable.

So revert the optimistic first approach to make room for that alternate
fix instead.

Reported-by: Josh Snyder <joshs@netflix.com>
Cc: Jiri Kosina <jikos@kernel.org>
Cc: Dominique Martinet <asmadeus@codewreck.org>
Cc: Andy Lutomirski <luto@amacapital.net>
Cc: Dave Chinner <david@fromorbit.com>
Cc: Kevin Easton <kevin@guarana.org>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Cyril Hrubis <chrubis@suse.cz>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Tejun Heo <tj@kernel.org>
Cc: Kirill A. Shutemov <kirill@shutemov.name>
Cc: Daniel Gruss <daniel@gruss.cc>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoMerge tag 'for-linus-5.0' of git://github.com/cminyard/linux-ipmi
Linus Torvalds [Wed, 23 Jan 2019 20:00:19 +0000 (09:00 +1300)]
Merge tag 'for-linus-5.0' of git://github.com/cminyard/linux-ipmi

Pull IPMI fixes from Corey Minyard:
 "I missed the merge window, which wasn't really important at the time
  as there was nothing that critical that I had for 5.0.

  However, I say that,and then a number of critical fixes come in:

   - ipmi: fix use-after-free of user->release_barrier.rda
   - ipmi: Prevent use-after-free in deliver_response
   - ipmi: msghandler: Fix potential Spectre v1 vulnerabilities

  which are obvious candidates for 5.0.  Then there is:

   - ipmi:ssif: Fix handling of multi-part return messages

  which is less critical, but it still has some off-by-one things that
  are not great, so it seemed appropriate. Some machines are broken
  without it. Then:

   - ipmi: Don't initialize anything in the core until something uses it

  It turns out that using SRCU causes large chunks of memory to be used
  on big iron machines, even if IPMI is never used. This was causing
  some issues for people on those machines.

  Everything here is destined for stable"

* tag 'for-linus-5.0' of git://github.com/cminyard/linux-ipmi:
  ipmi: Don't initialize anything in the core until something uses it
  ipmi: fix use-after-free of user->release_barrier.rda
  ipmi: Prevent use-after-free in deliver_response
  ipmi: msghandler: Fix potential Spectre v1 vulnerabilities
  ipmi:ssif: Fix handling of multi-part return messages

5 years agoMerge tag 's390-5.0-2' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Wed, 23 Jan 2019 19:58:01 +0000 (08:58 +1300)]
Merge tag 's390-5.0-2' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Martin Schwidefsky:

 - Do not claim to run under z/VM if the hypervisor can not be
   identified

 - Fix crashes due to outdated ASCEs in CR1

 - Avoid a deadlock in regard to CPU hotplug

 - Really fix the vdso mapping issue for compat tasks

 - Avoid crash on restart due to an incorrect stack address

* tag 's390-5.0-2' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/smp: Fix calling smp_call_ipl_cpu() from ipl CPU
  s390/vdso: correct vdso mapping for compat tasks
  s390/smp: fix CPU hotplug deadlock with CPU rescan
  s390/mm: always force a load of the primary ASCE on context switch
  s390/early: improve machine detection

5 years agoax25: fix possible use-after-free
Eric Dumazet [Tue, 22 Jan 2019 18:40:59 +0000 (10:40 -0800)]
ax25: fix possible use-after-free

syzbot found that ax25 routes where not properly protected
against concurrent use [1].

In this particular report the bug happened while
copying ax25->digipeat.

Fix this problem by making sure we call ax25_get_route()
while ax25_route_lock is held, so that no modification
could happen while using the route.

The current two ax25_get_route() callers do not sleep,
so this change should be fine.

Once we do that, ax25_get_route() no longer needs to
grab a reference on the found route.

[1]
ax25_connect(): syz-executor0 uses autobind, please contact jreuter@yaina.de
BUG: KASAN: use-after-free in memcpy include/linux/string.h:352 [inline]
BUG: KASAN: use-after-free in kmemdup+0x42/0x60 mm/util.c:113
Read of size 66 at addr ffff888066641a80 by task syz-executor2/531

ax25_connect(): syz-executor0 uses autobind, please contact jreuter@yaina.de
CPU: 1 PID: 531 Comm: syz-executor2 Not tainted 5.0.0-rc2+ #10
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
Call Trace:
 __dump_stack lib/dump_stack.c:77 [inline]
 dump_stack+0x1db/0x2d0 lib/dump_stack.c:113
 print_address_description.cold+0x7c/0x20d mm/kasan/report.c:187
 kasan_report.cold+0x1b/0x40 mm/kasan/report.c:317
 check_memory_region_inline mm/kasan/generic.c:185 [inline]
 check_memory_region+0x123/0x190 mm/kasan/generic.c:191
 memcpy+0x24/0x50 mm/kasan/common.c:130
 memcpy include/linux/string.h:352 [inline]
 kmemdup+0x42/0x60 mm/util.c:113
 kmemdup include/linux/string.h:425 [inline]
 ax25_rt_autobind+0x25d/0x750 net/ax25/ax25_route.c:424
 ax25_connect.cold+0x30/0xa4 net/ax25/af_ax25.c:1224
 __sys_connect+0x357/0x490 net/socket.c:1664
 __do_sys_connect net/socket.c:1675 [inline]
 __se_sys_connect net/socket.c:1672 [inline]
 __x64_sys_connect+0x73/0xb0 net/socket.c:1672
 do_syscall_64+0x1a3/0x800 arch/x86/entry/common.c:290
 entry_SYSCALL_64_after_hwframe+0x49/0xbe
RIP: 0033:0x458099
Code: 6d b7 fb ff c3 66 2e 0f 1f 84 00 00 00 00 00 66 90 48 89 f8 48 89 f7 48 89 d6 48 89 ca 4d 89 c2 4d 89 c8 4c 8b 4c 24 08 0f 05 <48> 3d 01 f0 ff ff 0f 83 3b b7 fb ff c3 66 2e 0f 1f 84 00 00 00 00
RSP: 002b:00007f870ee22c78 EFLAGS: 00000246 ORIG_RAX: 000000000000002a
RAX: ffffffffffffffda RBX: 0000000000000003 RCX: 0000000000458099
RDX: 0000000000000048 RSI: 0000000020000080 RDI: 0000000000000005
RBP: 000000000073bf00 R08: 0000000000000000 R09: 0000000000000000
ax25_connect(): syz-executor4 uses autobind, please contact jreuter@yaina.de
R10: 0000000000000000 R11: 0000000000000246 R12: 00007f870ee236d4
R13: 00000000004be48e R14: 00000000004ce9a8 R15: 00000000ffffffff

Allocated by task 526:
 save_stack+0x45/0xd0 mm/kasan/common.c:73
 set_track mm/kasan/common.c:85 [inline]
 __kasan_kmalloc mm/kasan/common.c:496 [inline]
 __kasan_kmalloc.constprop.0+0xcf/0xe0 mm/kasan/common.c:469
 kasan_kmalloc+0x9/0x10 mm/kasan/common.c:504
ax25_connect(): syz-executor5 uses autobind, please contact jreuter@yaina.de
 kmem_cache_alloc_trace+0x151/0x760 mm/slab.c:3609
 kmalloc include/linux/slab.h:545 [inline]
 ax25_rt_add net/ax25/ax25_route.c:95 [inline]
 ax25_rt_ioctl+0x3b9/0x1270 net/ax25/ax25_route.c:233
 ax25_ioctl+0x322/0x10b0 net/ax25/af_ax25.c:1763
 sock_do_ioctl+0xe2/0x400 net/socket.c:950
 sock_ioctl+0x32f/0x6c0 net/socket.c:1074
 vfs_ioctl fs/ioctl.c:46 [inline]
 file_ioctl fs/ioctl.c:509 [inline]
 do_vfs_ioctl+0x107b/0x17d0 fs/ioctl.c:696
 ksys_ioctl+0xab/0xd0 fs/ioctl.c:713
 __do_sys_ioctl fs/ioctl.c:720 [inline]
 __se_sys_ioctl fs/ioctl.c:718 [inline]
 __x64_sys_ioctl+0x73/0xb0 fs/ioctl.c:718
 do_syscall_64+0x1a3/0x800 arch/x86/entry/common.c:290
 entry_SYSCALL_64_after_hwframe+0x49/0xbe

ax25_connect(): syz-executor5 uses autobind, please contact jreuter@yaina.de
Freed by task 550:
 save_stack+0x45/0xd0 mm/kasan/common.c:73
 set_track mm/kasan/common.c:85 [inline]
 __kasan_slab_free+0x102/0x150 mm/kasan/common.c:458
 kasan_slab_free+0xe/0x10 mm/kasan/common.c:466
 __cache_free mm/slab.c:3487 [inline]
 kfree+0xcf/0x230 mm/slab.c:3806
 ax25_rt_add net/ax25/ax25_route.c:92 [inline]
 ax25_rt_ioctl+0x304/0x1270 net/ax25/ax25_route.c:233
 ax25_ioctl+0x322/0x10b0 net/ax25/af_ax25.c:1763
 sock_do_ioctl+0xe2/0x400 net/socket.c:950
 sock_ioctl+0x32f/0x6c0 net/socket.c:1074
 vfs_ioctl fs/ioctl.c:46 [inline]
 file_ioctl fs/ioctl.c:509 [inline]
 do_vfs_ioctl+0x107b/0x17d0 fs/ioctl.c:696
 ksys_ioctl+0xab/0xd0 fs/ioctl.c:713
 __do_sys_ioctl fs/ioctl.c:720 [inline]
 __se_sys_ioctl fs/ioctl.c:718 [inline]
 __x64_sys_ioctl+0x73/0xb0 fs/ioctl.c:718
 do_syscall_64+0x1a3/0x800 arch/x86/entry/common.c:290
 entry_SYSCALL_64_after_hwframe+0x49/0xbe

The buggy address belongs to the object at ffff888066641a80
 which belongs to the cache kmalloc-96 of size 96
The buggy address is located 0 bytes inside of
 96-byte region [ffff888066641a80ffff888066641ae0)
The buggy address belongs to the page:
page:ffffea0001999040 count:1 mapcount:0 mapping:ffff88812c3f04c0 index:0x0
flags: 0x1fffc0000000200(slab)
ax25_connect(): syz-executor4 uses autobind, please contact jreuter@yaina.de
raw: 01fffc0000000200 ffffea0001817948 ffffea0002341dc8 ffff88812c3f04c0
raw: 0000000000000000 ffff888066641000 0000000100000020 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
 ffff888066641980: fb fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
 ffff888066641a00: 00 00 00 00 00 00 00 00 02 fc fc fc fc fc fc fc
>ffff888066641a80: fb fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
                   ^
 ffff888066641b00: fb fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
 ffff888066641b80: 00 00 00 00 00 00 00 00 00 00 00 00 fc fc fc fc

Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Ralf Baechle <ralf@linux-mips.org>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agosfc: suppress duplicate nvmem partition types in efx_ef10_mtd_probe
Edward Cree [Tue, 22 Jan 2019 19:02:17 +0000 (19:02 +0000)]
sfc: suppress duplicate nvmem partition types in efx_ef10_mtd_probe

Use a bitmap to keep track of which partition types we've already seen;
 for duplicates, return -EEXIST from efx_ef10_mtd_probe_partition() and
 thus skip adding that partition.
Duplicate partitions occur because of the A/B backup scheme used by newer
 sfc NICs.  Prior to this patch they cause sysfs_warn_dup errors because
 they have the same name, causing us not to expose any MTDs at all.

Signed-off-by: Edward Cree <ecree@solarflare.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agohv_netvsc: fix typos in code comments
Adrian Vladu [Thu, 3 Jan 2019 19:43:08 +0000 (19:43 +0000)]
hv_netvsc: fix typos in code comments

Fix all typos from hyperv netvsc code comments.

Signed-off-by: Adrian Vladu <avladu@cloudbasesolutions.com>
Cc: "K. Y. Srinivasan" <kys@microsoft.com>
Cc: Haiyang Zhang <haiyangz@microsoft.com>
Cc: Stephen Hemminger <sthemmin@microsoft.com>
Cc: Sasha Levin <sashal@kernel.org>
Cc: "David S. Miller" <davem@davemloft.net>
Cc: "Alessandro Pilotti" <apilotti@cloudbasesolutions.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
5 years agohv_netvsc: Fix hash key value reset after other ops
Haiyang Zhang [Tue, 15 Jan 2019 00:51:44 +0000 (00:51 +0000)]
hv_netvsc: Fix hash key value reset after other ops

Changing mtu, channels, or buffer sizes ops call to netvsc_attach(),
rndis_set_subchannel(), which always reset the hash key to default
value. That will override hash key changed previously. This patch
fixes the problem by save the hash key, then restore it when we re-
add the netvsc device.

Fixes: ff4a44199012 ("netvsc: allow get/set of RSS indirection table")
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
[sl: fix up subject line]
Signed-off-by: Sasha Levin <sashal@kernel.org>
5 years agohv_netvsc: Refactor assignments of struct netvsc_device_info
Haiyang Zhang [Tue, 15 Jan 2019 00:51:43 +0000 (00:51 +0000)]
hv_netvsc: Refactor assignments of struct netvsc_device_info

These assignments occur in multiple places. The patch refactor them
to a function for simplicity. It also puts the struct to heap area
for future expension.

Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
[sl: fix up subject line]
Signed-off-by: Sasha Levin <sashal@kernel.org>
5 years agohv_netvsc: Fix ethtool change hash key error
Haiyang Zhang [Tue, 15 Jan 2019 00:51:42 +0000 (00:51 +0000)]
hv_netvsc: Fix ethtool change hash key error

Hyper-V hosts require us to disable RSS before changing RSS key,
otherwise the changing request will fail. This patch fixes the
coding error.

Fixes: ff4a44199012 ("netvsc: allow get/set of RSS indirection table")
Reported-by: Wei Hu <weh@microsoft.com>
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Michael Kelley <mikelley@microsoft.com>
[sl: fix up subject line]
Signed-off-by: Sasha Levin <sashal@kernel.org>
5 years agoravb: expand rx descriptor data to accommodate hw checksum
Simon Horman [Wed, 23 Jan 2019 11:14:52 +0000 (12:14 +0100)]
ravb: expand rx descriptor data to accommodate hw checksum

EtherAVB may provide a checksum of packet data appended to packet data. In
order to allow this checksum to be received by the host descriptor data
needs to be enlarged by 2 bytes to accommodate the checksum.

In the case of MTU-sized packets without a VLAN tag the
checksum were already accommodated by virtue of the space reserved for the
VLAN tag. However, a packet of MTU-size with a  VLAN tag consumed all
packet data space provided by a descriptor leaving no space for the
trailing checksum.

This was not detected by the driver which incorrectly used the last two
bytes of packet data as the checksum and truncate the packet by two bytes.
This resulted all such packets being dropped.

A work around is to disable RX checksum offload
 # ethtool -K eth0 rx off

This patch resolves this problem by increasing the size available for
packet data in RX descriptors by two bytes.

Tested on R-Car E3 (r8a77990) ES1.0 based Ebisu-4D board

v2
* Use sizeof(__sum16) directly rather than adding a driver-local
  #define for the size of the checksum provided by the hw (2 bytes).

Fixes: 4d86d3818627 ("ravb: RX checksum offload")
Signed-off-by: Simon Horman <horms+renesas@verge.net.au>
Reviewed-by: Sergei Shtylyov <sergei.shtylyov@cogentembedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoipmi: Don't initialize anything in the core until something uses it
Corey Minyard [Thu, 20 Dec 2018 22:50:23 +0000 (16:50 -0600)]
ipmi: Don't initialize anything in the core until something uses it

The IPMI driver was recently modified to use SRCU, but it turns out
this uses a chunk of percpu memory, even if IPMI is never used.

So modify thing to on initialize on the first use.  There was already
code to sort of handle this for handling init races, so piggy back
on top of that, and simplify it in the process.

Signed-off-by: Corey Minyard <cminyard@mvista.com>
Reported-by: Tejun Heo <tj@kernel.org>
Cc: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Reviewed-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: stable@vger.kernel.org # 4.18
5 years agoipmi: fix use-after-free of user->release_barrier.rda
Yang Yingliang [Wed, 16 Jan 2019 05:33:22 +0000 (13:33 +0800)]
ipmi: fix use-after-free of user->release_barrier.rda

When we do the following test, we got oops in ipmi_msghandler driver
while((1))
do
service ipmievd restart & service ipmievd restart
done

---------------------------------------------------------------
[  294.230186] Unable to handle kernel paging request at virtual address 0000803fea6ea008
[  294.230188] Mem abort info:
[  294.230190]   ESR = 0x96000004
[  294.230191]   Exception class = DABT (current EL), IL = 32 bits
[  294.230193]   SET = 0, FnV = 0
[  294.230194]   EA = 0, S1PTW = 0
[  294.230195] Data abort info:
[  294.230196]   ISV = 0, ISS = 0x00000004
[  294.230197]   CM = 0, WnR = 0
[  294.230199] user pgtable: 4k pages, 48-bit VAs, pgdp = 00000000a1c1b75a
[  294.230201] [0000803fea6ea008] pgd=0000000000000000
[  294.230204] Internal error: Oops: 96000004 [#1] SMP
[  294.235211] Modules linked in: nls_utf8 isofs rpcrdma ib_iser ib_srpt target_core_mod ib_srp scsi_transport_srp ib_ipoib rdma_ucm ib_umad rdma_cm ib_cm iw_cm dm_mirror dm_region_hash dm_log dm_mod aes_ce_blk crypto_simd cryptd aes_ce_cipher ghash_ce sha2_ce ses sha256_arm64 sha1_ce hibmc_drm hisi_sas_v2_hw enclosure sg hisi_sas_main sbsa_gwdt ip_tables mlx5_ib ib_uverbs marvell ib_core mlx5_core ixgbe ipmi_si mdio hns_dsaf ipmi_devintf ipmi_msghandler hns_enet_drv hns_mdio
[  294.277745] CPU: 3 PID: 0 Comm: swapper/3 Kdump: loaded Not tainted 5.0.0-rc2+ #113
[  294.285511] Hardware name: Huawei TaiShan 2280 /BC11SPCD, BIOS 1.37 11/21/2017
[  294.292835] pstate: 80000005 (Nzcv daif -PAN -UAO)
[  294.297695] pc : __srcu_read_lock+0x38/0x58
[  294.301940] lr : acquire_ipmi_user+0x2c/0x70 [ipmi_msghandler]
[  294.307853] sp : ffff00001001bc80
[  294.311208] x29: ffff00001001bc80 x28: ffff0000117e5000
[  294.316594] x27: 0000000000000000 x26: dead000000000100
[  294.321980] x25: dead000000000200 x24: ffff803f6bd06800
[  294.327366] x23: 0000000000000000 x22: 0000000000000000
[  294.332752] x21: ffff00001001bd04 x20: ffff80df33d19018
[  294.338137] x19: ffff80df33d19018 x18: 0000000000000000
[  294.343523] x17: 0000000000000000 x16: 0000000000000000
[  294.348908] x15: 0000000000000000 x14: 0000000000000002
[  294.354293] x13: 0000000000000000 x12: 0000000000000000
[  294.359679] x11: 0000000000000000 x10: 0000000000100000
[  294.365065] x9 : 0000000000000000 x8 : 0000000000000004
[  294.370451] x7 : 0000000000000000 x6 : ffff80df34558678
[  294.375836] x5 : 000000000000000c x4 : 0000000000000000
[  294.381221] x3 : 0000000000000001 x2 : 0000803fea6ea000
[  294.386607] x1 : 0000803fea6ea008 x0 : 0000000000000001
[  294.391994] Process swapper/3 (pid: 0, stack limit = 0x0000000083087293)
[  294.398791] Call trace:
[  294.401266]  __srcu_read_lock+0x38/0x58
[  294.405154]  acquire_ipmi_user+0x2c/0x70 [ipmi_msghandler]
[  294.410716]  deliver_response+0x80/0xf8 [ipmi_msghandler]
[  294.416189]  deliver_local_response+0x28/0x68 [ipmi_msghandler]
[  294.422193]  handle_one_recv_msg+0x158/0xcf8 [ipmi_msghandler]
[  294.432050]  handle_new_recv_msgs+0xc0/0x210 [ipmi_msghandler]
[  294.441984]  smi_recv_tasklet+0x8c/0x158 [ipmi_msghandler]
[  294.451618]  tasklet_action_common.isra.5+0x88/0x138
[  294.460661]  tasklet_action+0x2c/0x38
[  294.468191]  __do_softirq+0x120/0x2f8
[  294.475561]  irq_exit+0x134/0x140
[  294.482445]  __handle_domain_irq+0x6c/0xc0
[  294.489954]  gic_handle_irq+0xb8/0x178
[  294.497037]  el1_irq+0xb0/0x140
[  294.503381]  arch_cpu_idle+0x34/0x1a8
[  294.510096]  do_idle+0x1d4/0x290
[  294.516322]  cpu_startup_entry+0x28/0x30
[  294.523230]  secondary_start_kernel+0x184/0x1d0
[  294.530657] Code: d538d082 d2800023 8b010c81 8b020021 (c85f7c25)
[  294.539746] ---[ end trace 8a7a880dee570b29 ]---
[  294.547341] Kernel panic - not syncing: Fatal exception in interrupt
[  294.556837] SMP: stopping secondary CPUs
[  294.563996] Kernel Offset: disabled
[  294.570515] CPU features: 0x002,21006008
[  294.577638] Memory Limit: none
[  294.587178] Starting crashdump kernel...
[  294.594314] Bye!

Because the user->release_barrier.rda is freed in ipmi_destroy_user(), but
the refcount is not zero, when acquire_ipmi_user() uses user->release_barrier.rda
in __srcu_read_lock(), it causes oops.
Fix this by calling cleanup_srcu_struct() when the refcount is zero.

Fixes: e86ee2d44b44 ("ipmi: Rework locking and shutdown for hot remove")
Cc: stable@vger.kernel.org # 4.18
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
Signed-off-by: Corey Minyard <cminyard@mvista.com>
5 years agoipmi: Prevent use-after-free in deliver_response
Fred Klassen [Sat, 19 Jan 2019 22:28:18 +0000 (14:28 -0800)]
ipmi: Prevent use-after-free in deliver_response

Some IPMI modules (e.g. ibmpex_msg_handler()) will have ipmi_usr_hdlr
handlers that call ipmi_free_recv_msg() directly. This will essentially
kfree(msg), leading to use-after-free.

This does not happen in the ipmi_devintf module, which will queue the
message and run ipmi_free_recv_msg() later.

BUG: KASAN: use-after-free in deliver_response+0x12f/0x1b0
Read of size 8 at addr ffff888a7bf20018 by task ksoftirqd/3/27
CPU: 3 PID: 27 Comm: ksoftirqd/3 Tainted: G           O      4.19.11-amd64-ani99-debug #12.0.1.601133+pv
Hardware name: AppNeta r1000/X11SPW-TF, BIOS 2.1a-AP 09/17/2018
Call Trace:
dump_stack+0x92/0xeb
print_address_description+0x73/0x290
kasan_report+0x258/0x380
deliver_response+0x12f/0x1b0
? ipmi_free_recv_msg+0x50/0x50
deliver_local_response+0xe/0x50
handle_one_recv_msg+0x37a/0x21d0
handle_new_recv_msgs+0x1ce/0x440
...

Allocated by task 9885:
kasan_kmalloc+0xa0/0xd0
kmem_cache_alloc_trace+0x116/0x290
ipmi_alloc_recv_msg+0x28/0x70
i_ipmi_request+0xb4a/0x1640
ipmi_request_settime+0x1b8/0x1e0
...

Freed by task 27:
__kasan_slab_free+0x12e/0x180
kfree+0xe9/0x280
deliver_response+0x122/0x1b0
deliver_local_response+0xe/0x50
handle_one_recv_msg+0x37a/0x21d0
handle_new_recv_msgs+0x1ce/0x440
tasklet_action_common.isra.19+0xc4/0x250
__do_softirq+0x11f/0x51f

Fixes: e86ee2d44b44 ("ipmi: Rework locking and shutdown for hot remove")
Cc: stable@vger.kernel.org # 4.18
Signed-off-by: Fred Klassen <fklassen@appneta.com>
Signed-off-by: Corey Minyard <cminyard@mvista.com>
5 years agoipmi: msghandler: Fix potential Spectre v1 vulnerabilities
Gustavo A. R. Silva [Wed, 9 Jan 2019 23:39:06 +0000 (17:39 -0600)]
ipmi: msghandler: Fix potential Spectre v1 vulnerabilities

channel and addr->channel are indirectly controlled by user-space,
hence leading to a potential exploitation of the Spectre variant 1
vulnerability.

These issues were detected with the help of Smatch:

drivers/char/ipmi/ipmi_msghandler.c:1381 ipmi_set_my_address() warn: potential spectre issue 'user->intf->addrinfo' [w] (local cap)
drivers/char/ipmi/ipmi_msghandler.c:1401 ipmi_get_my_address() warn: potential spectre issue 'user->intf->addrinfo' [r] (local cap)
drivers/char/ipmi/ipmi_msghandler.c:1421 ipmi_set_my_LUN() warn: potential spectre issue 'user->intf->addrinfo' [w] (local cap)
drivers/char/ipmi/ipmi_msghandler.c:1441 ipmi_get_my_LUN() warn: potential spectre issue 'user->intf->addrinfo' [r] (local cap)
drivers/char/ipmi/ipmi_msghandler.c:2260 check_addr() warn: potential spectre issue 'intf->addrinfo' [r] (local cap)

Fix this by sanitizing channel and addr->channel before using them to
index user->intf->addrinfo and intf->addrinfo, correspondingly.

Notice that given that speculation windows are large, the policy is
to kill the speculation on the first load and not worry if it can be
completed with a dependent load/store [1].

[1] https://lore.kernel.org/lkml/20180423164740.GY17484@dhcp22.suse.cz/

Cc: stable@vger.kernel.org
Signed-off-by: Gustavo A. R. Silva <gustavo@embeddedor.com>
Signed-off-by: Corey Minyard <cminyard@mvista.com>
5 years agoipmi:ssif: Fix handling of multi-part return messages
Corey Minyard [Fri, 16 Nov 2018 15:59:21 +0000 (09:59 -0600)]
ipmi:ssif: Fix handling of multi-part return messages

The block number was not being compared right, it was off by one
when checking the response.

Some statistics wouldn't be incremented properly in some cases.

Check to see if that middle-part messages always have 31 bytes of
data.

Signed-off-by: Corey Minyard <cminyard@mvista.com>
Cc: stable@vger.kernel.org # 4.4
5 years agovfio-pci/nvlink2: Fix ancient gcc warnings
Alexey Kardashevskiy [Wed, 23 Jan 2019 04:07:11 +0000 (15:07 +1100)]
vfio-pci/nvlink2: Fix ancient gcc warnings

Using the {0} construct as a generic initializer is perfectly fine in C,
however due to a bug in old gcc there is a warning:

  + /kisskb/src/drivers/vfio/pci/vfio_pci_nvlink2.c: warning: (near
initialization for 'cap.header') [-Wmissing-braces]:  => 181:9

Since for whatever reason we still want to compile the modern kernel
with such an old gcc without warnings, this changes the capabilities
initialization.

The gcc bugzilla: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=53119

Fixes: 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")
Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
5 years agodrm/i915/gvt: release shadow batch buffer and wa_ctx before destroy one workload
Weinan Li [Tue, 22 Jan 2019 05:46:27 +0000 (13:46 +0800)]
drm/i915/gvt: release shadow batch buffer and wa_ctx before destroy one workload

GVT-g will shadow the privilege batch buffer and the indirect context
during command scan, move the release process into
intel_vgpu_destroy_workload() to ensure the resources are recycled
properly.

Fixes: 0cce2823ed37 ("drm/i915/gvt/kvmgt:Refine error handling for prepare_execlist_workload")
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Weinan Li <weinan.z.li@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
5 years agonet: phy: Fixup GPLv2+ SPDX tags based on license text
Andrew Lunn [Mon, 21 Jan 2019 18:08:49 +0000 (19:08 +0100)]
net: phy: Fixup GPLv2+ SPDX tags based on license text

A few PHY drivers have the GPLv2+ license text. They then either have
a MODULE_LICENSE() of GPLv2 only, or an SPDX tag of GPLv2 only.

Since the license text is much easier to understand than either the
SPDX tag or the MODULE_LICENSE, use it as the definitive source of the
licence, and fixup the others when there are contradictions.

Cc: David Wu <david.wu@rock-chips.com>
Cc: Dongpo Li <lidongpo@hisilicon.com>
Cc: Michael Schmitz <schmitzmic@gmail.com>
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Acked-by: Michael Schmitz <schmitzmic@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: fec: get regulator optional
Stefan Agner [Mon, 21 Jan 2019 14:58:47 +0000 (15:58 +0100)]
net: fec: get regulator optional

According to the device tree binding the phy-supply property is
optional. Use the regulator_get_optional API accordingly. The
code already handles NULL just fine.

This gets rid of the following warning:
  fec 2188000.ethernet: 2188000.ethernet supply phy not found, using dummy regulator

Signed-off-by: Stefan Agner <stefan@agner.ch>
Reviewed-by: Marcel Ziswiler <marcel.ziswiler@toradex.com>
Acked-by: Fugang Duan <fugang.duan@nxp.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/ipv6: lower the level of "link is not ready" messages
Lubomir Rintel [Mon, 21 Jan 2019 13:54:20 +0000 (14:54 +0100)]
net/ipv6: lower the level of "link is not ready" messages

This message gets logged far too often for how interesting is it.

Most distributions nowadays configure NetworkManager to use randomly
generated MAC addresses for Wi-Fi network scans. The interfaces end up
being periodically brought down for the address change. When they're
subsequently brought back up, the message is logged, eventually flooding
the log.

Perhaps the message is not all that helpful: it seems to be more
interesting to hear when the addrconf actually start, not when it does
not. Let's lower its level.

Signed-off-by: Lubomir Rintel <lkundrak@v3.sk>
Acked-By: Thomas Haller <thaller@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'for-rc' of git://git.kernel.org/pub/scm/linux/kernel/git/rzhang/linux
Linus Torvalds [Wed, 23 Jan 2019 03:23:41 +0000 (16:23 +1300)]
Merge branch 'for-rc' of git://git./linux/kernel/git/rzhang/linux

Pull thermal management fixes from Zhang Rui:

 - Fix a race condition that sysfs could be accessed before necessary
   initialization in int340x thermal driver. (Aaron Hill)

 - Fix a NULL vs IS_ERR() check in int340x thermal driver. (Dan
   Carpenter)

* 'for-rc' of git://git.kernel.org/pub/scm/linux/kernel/git/rzhang/linux:
  drivers: thermal: int340x_thermal: Fix sysfs race condition
  thermal: int340x_thermal: Fix a NULL vs IS_ERR() check

5 years agoscsi: ufs: Use explicit access size in ufshcd_dump_regs
Marc Gonzalez [Tue, 22 Jan 2019 17:29:22 +0000 (18:29 +0100)]
scsi: ufs: Use explicit access size in ufshcd_dump_regs

memcpy_fromio() doesn't provide any control over access size.  For example,
on arm64, it is implemented using readb and readq.  This may trigger a
synchronous external abort:

[    3.729943] Internal error: synchronous external abort: 96000210 [#1] PREEMPT SMP
[    3.737000] Modules linked in:
[    3.744371] CPU: 2 PID: 1 Comm: swapper/0 Tainted: G S                4.20.0-rc4 #16
[    3.747413] Hardware name: Qualcomm Technologies, Inc. MSM8998 v1 MTP (DT)
[    3.755295] pstate: 00000005 (nzcv daif -PAN -UAO)
[    3.761978] pc : __memcpy_fromio+0x68/0x80
[    3.766718] lr : ufshcd_dump_regs+0x50/0xb0
[    3.770767] sp : ffff00000807ba00
[    3.774830] x29: ffff00000807ba00 x28: 00000000fffffffb
[    3.778344] x27: ffff0000089db068 x26: ffff8000f6e58000
[    3.783728] x25: 000000000000000e x24: 0000000000000800
[    3.789023] x23: ffff8000f6e587c8 x22: 0000000000000800
[    3.794319] x21: ffff000008908368 x20: ffff8000f6e1ab80
[    3.799615] x19: 000000000000006c x18: ffffffffffffffff
[    3.804910] x17: 0000000000000000 x16: 0000000000000000
[    3.810206] x15: ffff000009199648 x14: ffff000089244187
[    3.815502] x13: ffff000009244195 x12: ffff0000091ab000
[    3.820797] x11: 0000000005f5e0ff x10: ffff0000091998a0
[    3.826093] x9 : 0000000000000000 x8 : ffff8000f6e1ac00
[    3.831389] x7 : 0000000000000000 x6 : 0000000000000068
[    3.836676] x5 : ffff8000f6e1abe8 x4 : 0000000000000000
[    3.841971] x3 : ffff00000928c868 x2 : ffff8000f6e1abec
[    3.847267] x1 : ffff00000928c868 x0 : ffff8000f6e1abe8
[    3.852567] Process swapper/0 (pid: 1, stack limit = 0x(____ptrval____))
[    3.857900] Call trace:
[    3.864473]  __memcpy_fromio+0x68/0x80
[    3.866683]  ufs_qcom_dump_dbg_regs+0x1c0/0x370
[    3.870522]  ufshcd_print_host_regs+0x168/0x190
[    3.874946]  ufshcd_init+0xd4c/0xde0
[    3.879459]  ufshcd_pltfrm_init+0x3c8/0x550
[    3.883264]  ufs_qcom_probe+0x24/0x60
[    3.887188]  platform_drv_probe+0x50/0xa0

Assuming aligned 32-bit registers, let's use readl, after making sure
that 'offset' and 'len' are indeed multiples of 4.

Fixes: ba80917d9932d ("scsi: ufs: ufshcd_dump_regs to use memcpy_fromio")
Cc: <stable@vger.kernel.org>
Signed-off-by: Marc Gonzalez <marc.w.gonzalez@free.fr>
Acked-by: Tomas Winkler <tomas.winkler@intel.com>
Reviewed-by: Jeffrey Hugo <jhugo@codeaurora.org>
Reviewed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Tested-by: Evan Green <evgreen@chromium.org>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: tcmu: fix use after free
Xiubo Li [Tue, 22 Jan 2019 10:10:51 +0000 (18:10 +0800)]
scsi: tcmu: fix use after free

Fixes: a94a2572b977 ("scsi: tcmu: avoid cmd/qfull timers updated whenever a new cmd comes")
Reported-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: Mike Christie <mchristi@redhat.com>
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Mike Christie <mchristi@redhat.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agonet: altera_tse: fix connect_local_phy error path
Atsushi Nemoto [Mon, 21 Jan 2019 08:26:41 +0000 (17:26 +0900)]
net: altera_tse: fix connect_local_phy error path

The connect_local_phy should return NULL (not negative errno) on
error, since its caller expects it.

Signed-off-by: Atsushi Nemoto <atsushi.nemoto@sord.co.jp>
Acked-by: Thor Thayer <thor.thayer@linux.intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoscsi: csiostor: fix NULL pointer dereference in csio_vport_set_state()
Varun Prakash [Sat, 12 Jan 2019 16:44:30 +0000 (22:14 +0530)]
scsi: csiostor: fix NULL pointer dereference in csio_vport_set_state()

Assign fc_vport to ln->fc_vport before calling csio_fcoe_alloc_vnp() to
avoid a NULL pointer dereference in csio_vport_set_state().

ln->fc_vport is dereferenced in csio_vport_set_state().

Signed-off-by: Varun Prakash <varun@chelsio.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: lpfc: nvmet: avoid hang / use-after-free when destroying targetport
Ewan D. Milne [Thu, 17 Jan 2019 16:14:45 +0000 (11:14 -0500)]
scsi: lpfc: nvmet: avoid hang / use-after-free when destroying targetport

We cannot wait on a completion object in the lpfc_nvme_targetport structure
in the _destroy_targetport() code path because the NVMe/fc transport will
free that structure immediately after the .targetport_delete() callback.
This results in a use-after-free, and a hang if slub_debug=FZPU is enabled.

Fix this by putting the completion on the stack.

Signed-off-by: Ewan D. Milne <emilne@redhat.com>
Acked-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: lpfc: nvme: avoid hang / use-after-free when destroying localport
Ewan D. Milne [Thu, 17 Jan 2019 16:14:44 +0000 (11:14 -0500)]
scsi: lpfc: nvme: avoid hang / use-after-free when destroying localport

We cannot wait on a completion object in the lpfc_nvme_lport structure in
the _destroy_localport() code path because the NVMe/fc transport will free
that structure immediately after the .localport_delete() callback.  This
results in a use-after-free, and a hang if slub_debug=FZPU is enabled.

Fix this by putting the completion on the stack.

Signed-off-by: Ewan D. Milne <emilne@redhat.com>
Acked-by: James Smart <james.smart@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agoscsi: communicate max segment size to the DMA mapping code
Christoph Hellwig [Wed, 16 Jan 2019 16:12:15 +0000 (17:12 +0100)]
scsi: communicate max segment size to the DMA mapping code

When a host driver sets a maximum segment size we should not only propagate
that setting to the block layer, which can merge segments, but also to the
DMA mapping layer which can merge segments as well.

Fixes: 50c2e9107f ("scsi: introduce a max_segment_size host_template parameters")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
5 years agonet: dpaa2: improve PTP Kconfig option
Yangbo Lu [Mon, 21 Jan 2019 06:26:37 +0000 (14:26 +0800)]
net: dpaa2: improve PTP Kconfig option

Converted to use "imply" instead of "select" for PTP_1588_CLOCK
driver selecting. This could break the hard dependency between
the PTP clock subsystem and ethernet drivers.
This patch also set "default y" for dpaa2 ptp driver building to
provide user an available ptp clock in default.

Signed-off-by: Yangbo Lu <yangbo.lu@nxp.com>
Acked-by: Richard Cochran <richardcochran@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge branch 'qed-Error-recovery-process'
David S. Miller [Wed, 23 Jan 2019 01:30:39 +0000 (17:30 -0800)]
Merge branch 'qed-Error-recovery-process'

Michal Kalderon says:

====================
qed*: Error recovery process

Parity errors might happen in the device's memories due to momentary bit
flips which are caused by radiation.
Errors that are not correctable initiate a process kill event, which blocks
the device access towards the host and the network, and a recovery process
is started in the management FW and in the driver.

This series adds the support of this process in the qed core module and in
the qede driver (patches 2 & 3).
Patch 1 in the series revises the load sequence, to avoid PCI errors that
might be observed during a recovery process.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoqede: Error recovery process
Tomer Tayar [Sun, 20 Jan 2019 09:36:39 +0000 (11:36 +0200)]
qede: Error recovery process

This patch adds the error recovery process in the qede driver.
The process includes a partial/customized driver unload and load, which
allows it to look like a short suspend period to the kernel while
preserving the net devices' state.

Signed-off-by: Tomer Tayar <tomer.tayar@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: Michal Kalderon <michal.kalderon@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoqed: Add infrastructure for error detection and recovery
Tomer Tayar [Sun, 20 Jan 2019 09:36:38 +0000 (11:36 +0200)]
qed: Add infrastructure for error detection and recovery

This patch adds the detection and handling of a parity error ("process kill
event"), including the update of the protocol drivers, and the prevention
of any HW access that will lead to device access towards the host while
recovery is in progress.
It also provides the means for the protocol drivers to trigger a recovery
process on their decision.

Signed-off-by: Tomer Tayar <tomer.tayar@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: Michal Kalderon <michal.kalderon@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoqed: Revise load sequence to avoid PCI errors
Tomer Tayar [Sun, 20 Jan 2019 09:36:37 +0000 (11:36 +0200)]
qed: Revise load sequence to avoid PCI errors

Initiating final cleanup after an ungraceful driver unload can lead to bad
PCI accesses towards the host.
This patch revises the load sequence so final cleanup is sent while the
internal master enable is cleared, to prevent the host accesses, and clears
the internal error indications just before enabling the internal master
enable.

Signed-off-by: Tomer Tayar <tomer.tayar@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: Michal Kalderon <michal.kalderon@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet/ipv6: don't return positive numbers when nothing was dumped
Jakub Kicinski [Tue, 22 Jan 2019 22:47:19 +0000 (14:47 -0800)]
net/ipv6: don't return positive numbers when nothing was dumped

in6_dump_addrs() returns a positive 1 if there was nothing to dump.
This return value can not be passed as return from inet6_dump_addr()
as is, because it will confuse rtnetlink, resulting in NLMSG_DONE
never getting set:

$ ip addr list dev lo
EOF on netlink
Dump terminated

v2: flip condition to avoid a new goto (DaveA)

Fixes: 7c1e8a3817c5 ("netlink: fixup regression in RTM_GETADDR")
Reported-by: Brendan Galloway <brendan.galloway@netronome.com>
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Tested-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agoMerge tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Wed, 23 Jan 2019 01:04:22 +0000 (14:04 +1300)]
Merge tag 'clk-fixes-for-linus' of git://git./linux/kernel/git/clk/linux

Pull clk fixes from Stephen Boyd:
 "This is a sort of random collection of clk fixes that have come in
  since the merge window:

   - Handful of memory allocation and potentially bad pointer usage
     fixes

   - JSON format was incorrect for clk_dump because it missed a comma

   - Two Kconfig fixes, one duplicate and one missing select line

   - Compiler warning fix for the VC5 clk driver

   - Name and rate fixes for PLLs in the stratix10 driver so it can
     properly detect PLL rates and parents"

* tag 'clk-fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/clk/linux:
  clk: socfpga: stratix10: fix naming convention for the fixed-clocks
  clk: socfpga: stratix10: fix rate calculation for pll clocks
  clk: qcom: Select QCOM_GDSC with MSM_GCC_8998
  clk: vc5: Abort clock configuration without upstream clock
  clk: sysfs: fix invalid JSON in clk_dump
  clk: imx: Remove Kconfig duplicate include
  clk: zynqmp: Fix memory allocation in zynqmp_clk_setup
  clk: tegra: dfll: Fix a potential Oop in remove()
  clk: imx: fix potential NULL dereference in imx8qxp_lpcg_clk_probe()

5 years agoMerge tag 'linux-kselftest-5.0-rc4' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Wed, 23 Jan 2019 01:02:14 +0000 (14:02 +1300)]
Merge tag 'linux-kselftest-5.0-rc4' of git://git./linux/kernel/git/shuah/linux-kselftest

Pull kselftest fixes from Shuah Khan:
 "Fixes to rtc, seccomp and other tests"

* tag 'linux-kselftest-5.0-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/shuah/linux-kselftest:
  selftests/seccomp: Abort without user notification support
  selftests: gpio-mockup-chardev: Check asprintf() for error
  selftests: seccomp: use LDLIBS instead of LDFLAGS
  selftests/vm/gup_benchmark.c: match gup struct to kernel
  tools/testing/selftests/x86/unwind_vdso.c: Remove duplicate header
  x86/mpx/selftests: fix spelling mistake "succeded" -> "succeeded"
  selftests: rtc: rtctest: add alarm test on minute boundary
  selftests: rtc: rtctest: fix alarm tests

5 years agowriteback: synchronize sync(2) against cgroup writeback membership switches
Tejun Heo [Tue, 12 Dec 2017 16:38:30 +0000 (08:38 -0800)]
writeback: synchronize sync(2) against cgroup writeback membership switches

sync_inodes_sb() can race against cgwb (cgroup writeback) membership
switches and fail to writeback some inodes.  For example, if an inode
switches to another wb while sync_inodes_sb() is in progress, the new
wb might not be visible to bdi_split_work_to_wbs() at all or the inode
might jump from a wb which hasn't issued writebacks yet to one which
already has.

This patch adds backing_dev_info->wb_switch_rwsem to synchronize cgwb
switch path against sync_inodes_sb() so that sync_inodes_sb() is
guaranteed to see all the target wbs and inodes can't jump wbs to
escape syncing.

v2: Fixed misplaced rwsem init.  Spotted by Jiufei.

Signed-off-by: Tejun Heo <tj@kernel.org>
Reported-by: Jiufei Xue <xuejiufei@gmail.com>
Link: http://lkml.kernel.org/r/dc694ae2-f07f-61e1-7097-7c8411cee12d@gmail.com
Acked-by: Jan Kara <jack@suse.cz>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agonet: ip_gre: use erspan key field for tunnel lookup
Lorenzo Bianconi [Fri, 18 Jan 2019 11:05:39 +0000 (12:05 +0100)]
net: ip_gre: use erspan key field for tunnel lookup

Use ERSPAN key header field as tunnel key in gre_parse_header routine
since ERSPAN protocol sets the key field of the external GRE header to
0 resulting in a tunnel lookup fail in ip6gre_err.
In addition remove key field parsing and pskb_may_pull check in
erspan_rcv and ip6erspan_rcv

Fixes: 5a963eb61b7c ("ip6_gre: Add ERSPAN native tunnel support")
Signed-off-by: Lorenzo Bianconi <lorenzo.bianconi@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agonet: sun: cassini: Cleanup license conflict
Thomas Gleixner [Fri, 18 Jan 2019 10:49:58 +0000 (11:49 +0100)]
net: sun: cassini: Cleanup license conflict

The recent addition of SPDX license identifiers to the files in
drivers/net/ethernet/sun created a licensing conflict.

The cassini driver files contain a proper license notice:

  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public License as
  * published by the Free Software Foundation; either version 2 of the
  * License, or (at your option) any later version.

but the SPDX change added:

   SPDX-License-Identifier: GPL-2.0

So the file got tagged GPL v2 only while in fact it is licensed under GPL
v2 or later.

It's nice that people care about the SPDX tags, but they need to be more
careful about it. Not everything under (the) sun belongs to ...

Fix up the SPDX identifier and remove the boiler plate text as it is
redundant.

Fixes: c861ef83d771 ("sun: Add SPDX license tags to Sun network drivers")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Shannon Nelson <shannon.nelson@oracle.com>
Cc: Zhu Yanjun <yanjun.zhu@oracle.com>
Cc: David S. Miller <davem@davemloft.net>
Cc: netdev@vger.kernel.org
Cc: stable@vger.kernel.org
Acked-by: Shannon Nelson <shannon.lee.nelson@gmail.com>
Reviewed-by: Zhu Yanjun <yanjun.zhu@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
5 years agodm: add missing trace_block_split() to __split_and_process_bio()
Mike Snitzer [Fri, 18 Jan 2019 06:21:11 +0000 (01:21 -0500)]
dm: add missing trace_block_split() to __split_and_process_bio()

Provides useful context about bio splits in blktrace.

Signed-off-by: Mike Snitzer <snitzer@redhat.com>
5 years agodm: fix dm_wq_work() to only use __split_and_process_bio() if appropriate
Mike Snitzer [Thu, 17 Jan 2019 19:33:01 +0000 (14:33 -0500)]
dm: fix dm_wq_work() to only use __split_and_process_bio() if appropriate

Otherwise targets that don't support/expect IO splitting could resubmit
bios using code paths with unnecessary IO splitting complexity.

Depends-on: 24113d487843 ("dm: avoid indirect call in __dm_make_request")
Fixes: 978e51ba38e00 ("dm: optimize bio-based NVMe IO submission")
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
5 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid
Linus Torvalds [Tue, 22 Jan 2019 18:16:05 +0000 (07:16 +1300)]
Merge branch 'for-linus' of git://git./linux/kernel/git/hid/hid

Pull HID fixes from Jiri Kosina:

 - descriptor parsing regression fix for devices that have more than 16
   collections, from Peter Hutterer (and followup cleanup from Philipp
   Zabel)

 - quirk for Goodix touchpad

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hid/hid:
  HID: core: simplify active collection tracking
  HID: i2c-hid: Disable runtime PM on Goodix touchpad
  HID: core: replace the collection tree pointers with indices

5 years agovfio/pci: Cleanup license mess
Thomas Gleixner [Thu, 17 Jan 2019 23:14:25 +0000 (00:14 +0100)]
vfio/pci: Cleanup license mess

The recently added nvlink2 VFIO driver introduced a license conflict in two
files. In both cases the SPDX license identifier is:

  SPDX-License-Identifier: GPL-2.0+

but the files contain also the following license boiler plate text:

  * This program is free software; you can redistribute it and/or modify
  * it under the terms of the GNU General Public License version 2 as
  * published by the Free Software Foundation

The latter is GPL-2.9-only and not GPL-2.0=.

Looking deeper. The nvlink source file is derived from vfio_pci_igd.c which
is also licensed under GPL-2.0-only and it can be assumed that the file was
copied and modified. As the original file is licensed GPL-2.0-only it's not
possible to relicense derivative work to GPL-2.0-or-later.

Fix the SPDX identifier and remove the boiler plate as it is redundant.

Fixes: 7f92891778df ("vfio_pci: Add NVIDIA GV100GL [Tesla V100 SXM2] subdriver")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Alexey Kardashevskiy <aik@ozlabs.ru>
Cc: Alex Williamson <alex.williamson@redhat.com>
Cc: Michael Ellerman <mpe@ellerman.id.au>
Cc: kvm@vger.kernel.org
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
5 years agoblock: cover another queue enter recursion via BIO_QUEUE_ENTERED
Ming Lei [Tue, 22 Jan 2019 08:20:17 +0000 (16:20 +0800)]
block: cover another queue enter recursion via BIO_QUEUE_ENTERED

Except for blk_queue_split(), bio_split() is used for splitting bio too,
then the remained bio is often resubmit to queue via generic_make_request().
So the same queue enter recursion exits in this case too. Unfortunatley
commit cd4a4ae4683dc2 doesn't help this case.

This patch covers the above case by setting BIO_QUEUE_ENTERED before calling
q->make_request_fn.

In theory the per-bio flag is used to simulate one stack variable, it is
just fine to clear it after q->make_request_fn is returned. Especially
the same bio can't be submitted from another context.

Fixes: cd4a4ae4683dc2 ("block: don't use blocking queue entered for recursive bio submits")
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: NeilBrown <neilb@suse.com>
Reviewed-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agox86/Kconfig: Select PCI_LOCKLESS_CONFIG if PCI is enabled
Sinan Kaya [Mon, 21 Jan 2019 23:19:58 +0000 (23:19 +0000)]
x86/Kconfig: Select PCI_LOCKLESS_CONFIG if PCI is enabled

After commit

  5d32a66541c4 ("PCI/ACPI: Allow ACPI to be built without CONFIG_PCI set")

dependencies on CONFIG_PCI that previously were satisfied implicitly
through dependencies on CONFIG_ACPI have to be specified directly.

PCI_LOCKLESS_CONFIG depends on PCI but this dependency has not been
mentioned in the Kconfig so add an explicit dependency here and fix

  WARNING: unmet direct dependencies detected for PCI_LOCKLESS_CONFIG
    Depends on [n]: PCI [=n]
    Selected by [y]:
    - X86 [=y]

Fixes: 5d32a66541c46 ("PCI/ACPI: Allow ACPI to be built without CONFIG_PCI set")
Signed-off-by: Sinan Kaya <okaya@kernel.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-acpi@vger.kernel.org
Cc: x86-ml <x86@kernel.org>
Link: https://lkml.kernel.org/r/20190121231958.28255-2-okaya@kernel.org
5 years agodirect-io: allow direct writes to empty inodes
Ernesto A. Fernández [Mon, 8 Oct 2018 23:58:23 +0000 (20:58 -0300)]
direct-io: allow direct writes to empty inodes

On a DIO_SKIP_HOLES filesystem, the ->get_block() method is currently
not allowed to create blocks for an empty inode.  This confusion comes
from trying to bit shift a negative number, so check the size of the
inode first.

The problem is most visible for hfsplus, because the fallback to
buffered I/O doesn't happen and the write fails with EIO.  This is in
part the fault of the module, because it gives a wrong return value on
->get_block(); that will be fixed in a separate patch.

Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Signed-off-by: Ernesto A. Fernández <ernesto.mnd.fernandez@gmail.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
5 years agoi3c: master: Fix an error checking typo in 'cdns_i3c_master_probe()'
Christophe JAILLET [Sat, 29 Dec 2018 00:05:40 +0000 (01:05 +0100)]
i3c: master: Fix an error checking typo in 'cdns_i3c_master_probe()'

Fix a cut'n'paste typo.
Checking 'master->sysclk' is expected here.

Fixes: 603f2bee2c54 ("i3c: master: Add driver for Cadence IP")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agobinderfs: switch from d_add() to d_instantiate()
Christian Brauner [Mon, 21 Jan 2019 10:48:08 +0000 (11:48 +0100)]
binderfs: switch from d_add() to d_instantiate()

In a previous commit we switched from a d_alloc_name() + d_lookup()
combination to setup a new dentry and find potential duplicates to the more
idiomatic lookup_one_len(). As far as I understand, this also means we need
to switch from d_add() to d_instantiate() since lookup_one_len() will
create a new dentry when it doesn't find an existing one and add the new
dentry to the hash queues. So we only need to call d_instantiate() to
connect the dentry to the inode and turn it into a positive dentry.

If we were to use d_add() we sure see stack traces like the following
indicating that adding the same dentry twice over the same inode:

[  744.441889] CPU: 4 PID: 2849 Comm: landscape-sysin Not tainted 5.0.0-rc1-brauner-binderfs #243
[  744.441889] Hardware name: Dell      DCS XS24-SC2          /XS24-SC2              , BIOS S59_3C20 04/07/2011
[  744.441889] RIP: 0010:__d_lookup_rcu+0x76/0x190
[  744.441889] Code: 89 75 c0 49 c1 e9 20 49 89 fd 45 89 ce 41 83 e6 07 42 8d 04 f5 00 00 00 00 89 45 c8 eb 0c 48 8b 1b 48 85 db 0f 84 81 00 00 00 <44> 8b 63 fc 4c 3b 6b 10 75 ea 48 83 7b 08 00 74 e3 41 83 e4 fe 41
[  744.441889] RSP: 0018:ffffb8c984e27ad0 EFLAGS: 00000282 ORIG_RAX: ffffffffffffff13
[  744.441889] RAX: 0000000000000038 RBX: ffff9407ef770c08 RCX: ffffb8c980011000
[  744.441889] RDX: ffffb8c984e27b54 RSI: ffffb8c984e27ce0 RDI: ffff9407e6689600
[  744.441889] RBP: ffffb8c984e27b28 R08: ffffb8c984e27ba4 R09: 0000000000000007
[  744.441889] R10: ffff9407e5c4f05c R11: 973f3eb9d84a94e5 R12: 0000000000000002
[  744.441889] R13: ffff9407e6689600 R14: 0000000000000007 R15: 00000007bfef7a13
[  744.441889] FS:  00007f0db13bb740(0000) GS:ffff9407f3b00000(0000) knlGS:0000000000000000
[  744.441889] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  744.441889] CR2: 00007f0dacc51024 CR3: 000000032961a000 CR4: 00000000000006e0
[  744.441889] Call Trace:
[  744.441889]  lookup_fast+0x53/0x300
[  744.441889]  walk_component+0x49/0x350
[  744.441889]  ? inode_permission+0x63/0x1a0
[  744.441889]  link_path_walk.part.33+0x1bc/0x5a0
[  744.441889]  ? path_init+0x190/0x310
[  744.441889]  path_lookupat+0x95/0x210
[  744.441889]  filename_lookup+0xb6/0x190
[  744.441889]  ? __check_object_size+0xb8/0x1b0
[  744.441889]  ? strncpy_from_user+0x50/0x1a0
[  744.441889]  user_path_at_empty+0x36/0x40
[  744.441889]  ? user_path_at_empty+0x36/0x40
[  744.441889]  vfs_statx+0x76/0xe0
[  744.441889]  __do_sys_newstat+0x3d/0x70
[  744.441889]  __x64_sys_newstat+0x16/0x20
[  744.441889]  do_syscall_64+0x5a/0x120
[  744.441889]  entry_SYSCALL_64_after_hwframe+0x44/0xa9
[  744.441889] RIP: 0033:0x7f0db0ec2775
[  744.441889] Code: 00 00 00 75 05 48 83 c4 18 c3 e8 26 55 02 00 66 0f 1f 44 00 00 83 ff 01 48 89 f0 77 30 48 89 c7 48 89 d6 b8 04 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 03 f3 c3 90 48 8b 15 e1 b6 2d 00 f7 d8 64 89
[  744.441889] RSP: 002b:00007ffc36bc9388 EFLAGS: 00000246 ORIG_RAX: 0000000000000004
[  744.441889] RAX: ffffffffffffffda RBX: 00007ffc36bc9300 RCX: 00007f0db0ec2775
[  744.441889] RDX: 00007ffc36bc9400 RSI: 00007ffc36bc9400 RDI: 00007f0dad26f050
[  744.441889] RBP: 0000000000c0bc60 R08: 0000000000000000 R09: 0000000000000001
[  744.441889] R10: 0000000000000000 R11: 0000000000000246 R12: 00007ffc36bc9400
[  744.441889] R13: 0000000000000001 R14: 00000000ffffff9c R15: 0000000000c0bc60

Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Christian Brauner <christian@brauner.io>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agobinderfs: drop lock in binderfs_binder_ctl_create
Christian Brauner [Mon, 21 Jan 2019 10:48:07 +0000 (11:48 +0100)]
binderfs: drop lock in binderfs_binder_ctl_create

The binderfs_binder_ctl_create() call is a no-op on subsequent calls and
the first call is done before we unlock the suberblock. Hence, there is no
need to take inode_lock() in there. Let's remove it.

Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Christian Brauner <christian@brauner.io>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agobinderfs: kill_litter_super() before cleanup
Christian Brauner [Mon, 21 Jan 2019 10:48:06 +0000 (11:48 +0100)]
binderfs: kill_litter_super() before cleanup

Al pointed out that first calling kill_litter_super() before cleaning up
info is more correct since destroying info doesn't depend on the state of
the dentries and inodes. That the opposite remains true is not guaranteed.

Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Christian Brauner <christian@brauner.io>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agobinderfs: rework binderfs_binder_device_create()
Christian Brauner [Mon, 21 Jan 2019 10:48:05 +0000 (11:48 +0100)]
binderfs: rework binderfs_binder_device_create()

- switch from d_alloc_name() + d_lookup() to lookup_one_len():
  Instead of using d_alloc_name() and then doing a d_lookup() with the
  allocated dentry to find whether a device with the name we're trying to
  create already exists switch to using lookup_one_len().  The latter will
  either return the existing dentry or a new one.

- switch from kmalloc() + strscpy() to kmemdup():
  Use a more idiomatic way to copy the name for the new dentry that
  userspace gave us.

Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Christian Brauner <christian@brauner.io>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agobinderfs: rework binderfs_fill_super()
Christian Brauner [Mon, 21 Jan 2019 10:48:04 +0000 (11:48 +0100)]
binderfs: rework binderfs_fill_super()

Al pointed out that on binderfs_fill_super() error
deactivate_locked_super() will call binderfs_kill_super() so all of the
freeing and putting we currently do in binderfs_fill_super() is unnecessary
and buggy. Let's simply return errors and let binderfs_fill_super() take
care of cleaning up on error.

Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Christian Brauner <christian@brauner.io>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agobinderfs: prevent renaming the control dentry
Christian Brauner [Mon, 21 Jan 2019 10:48:03 +0000 (11:48 +0100)]
binderfs: prevent renaming the control dentry

- make binderfs control dentry immutable:
  We don't allow to unlink it since it is crucial for binderfs to be
  useable but if we allow to rename it we make the unlink trivial to
  bypass. So prevent renaming too and simply treat the control dentry as
  immutable.

- add is_binderfs_control_device() helper:
  Take the opportunity and turn the check for the control dentry into a
  separate helper is_binderfs_control_device() since it's now used in two
  places.

- simplify binderfs_rename():
  Instead of hand-rolling our custom version of simple_rename() just dumb
  the whole function down to first check whether we're trying to rename the
  control dentry. If we do EPERM the caller and if not call simple_rename().

Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Christian Brauner <christian@brauner.io>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agobinderfs: remove outdated comment
Christian Brauner [Mon, 21 Jan 2019 10:48:02 +0000 (11:48 +0100)]
binderfs: remove outdated comment

The comment stems from an early version of that patchset and is just
confusing now.

Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Christian Brauner <christian@brauner.io>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agobinderfs: use __u32 for device numbers
Christian Brauner [Mon, 21 Jan 2019 11:01:20 +0000 (12:01 +0100)]
binderfs: use __u32 for device numbers

We allow more then 255 binderfs binder devices to be created since there
are workloads that require more than that. If we use __u8 we'll overflow
after 255. So let's use a __u32.
Note that there's no released kernel with binderfs out there so this is
not a regression.

Signed-off-by: Christian Brauner <christian@brauner.io>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agobinderfs: use correct include guards in header
Christian Brauner [Mon, 21 Jan 2019 11:01:19 +0000 (12:01 +0100)]
binderfs: use correct include guards in header

When we switched over from binder_ctl.h to binderfs.h we forgot to change
the include guards. It's minor but it's obviously correct.

Signed-off-by: Christian Brauner <christian@brauner.io>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agostaging: android: ion: Support cpu access during dma_buf_detach
Liam Mark [Fri, 18 Jan 2019 18:37:44 +0000 (10:37 -0800)]
staging: android: ion: Support cpu access during dma_buf_detach

Often userspace doesn't know when the kernel will be calling dma_buf_detach
on the buffer.
If userpace starts its CPU access at the same time as the sg list is being
freed it could end up accessing the sg list after it has been freed.

Thread A Thread B
- DMA_BUF_IOCTL_SYNC IOCT
 - ion_dma_buf_begin_cpu_access
  - list_for_each_entry
- ion_dma_buf_detatch
 - free_duped_table
   - dma_sync_sg_for_cpu

Fix this by getting the ion_buffer lock before freeing the sg table memory.

Fixes: 2a55e7b5e544 ("staging: android: ion: Call dma_map_sg for syncing and mapping")
Signed-off-by: Liam Mark <lmark@codeaurora.org>
Acked-by: Laura Abbott <labbott@redhat.com>
Acked-by: Andrew F. Davis <afd@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agocan: flexcan: fix NULL pointer exception during bringup
Uwe Kleine-König [Fri, 11 Jan 2019 11:20:41 +0000 (12:20 +0100)]
can: flexcan: fix NULL pointer exception during bringup

Commit cbffaf7aa09e ("can: flexcan: Always use last mailbox for TX")
introduced a loop letting i run up to (including) ARRAY_SIZE(regs->mb)
and in the body accessed regs->mb[i] which is an out-of-bounds array
access that then resulted in an access to an reserved register area.

Later this was changed by commit 0517961ccdf1 ("can: flexcan: Add
provision for variable payload size") to iterate a bit differently but
still runs one iteration too much resulting to call

flexcan_get_mb(priv, priv->mb_count)

which results in a WARN_ON and then a NULL pointer exception. This
only affects devices compatible with "fsl,p1010-flexcan",
"fsl,imx53-flexcan", "fsl,imx35-flexcan", "fsl,imx25-flexcan",
"fsl,imx28-flexcan", so newer i.MX SoCs are not affected.

Fixes: cbffaf7aa09e ("can: flexcan: Always use last mailbox for TX")
Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Cc: linux-stable <stable@vger.kernel.org> # >= 4.20
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
5 years agocan: flexcan: fix 'passing zero to ERR_PTR()' warning
YueHaibing [Wed, 12 Dec 2018 09:24:01 +0000 (17:24 +0800)]
can: flexcan: fix 'passing zero to ERR_PTR()' warning

Fix a static code checker warning:
drivers/net/can/flexcan.c:1435 flexcan_setup_stop_mode() warn: passing zero to 'PTR_ERR'

Fixes: de3578c198c6 ("can: flexcan: add self wakeup support")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
5 years agocan: bcm: check timer values before ktime conversion
Oliver Hartkopp [Sun, 13 Jan 2019 18:31:43 +0000 (19:31 +0100)]
can: bcm: check timer values before ktime conversion

Kyungtae Kim detected a potential integer overflow in bcm_[rx|tx]_setup()
when the conversion into ktime multiplies the given value with NSEC_PER_USEC
(1000).

Reference: https://marc.info/?l=linux-can&m=154732118819828&w=2

Add a check for the given tv_usec, so that the value stays below one second.
Additionally limit the tv_sec value to a reasonable value for CAN related
use-cases of 400 days and ensure all values to be positive.

Reported-by: Kyungtae Kim <kt0755@gmail.com>
Tested-by: Oliver Hartkopp <socketcan@hartkopp.net>
Signed-off-by: Oliver Hartkopp <socketcan@hartkopp.net>
Cc: linux-stable <stable@vger.kernel.org> # >= 2.6.26
Tested-by: Kyungtae Kim <kt0755@gmail.com>
Acked-by: Andre Naujoks <nautsch2@gmail.com>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
5 years agocan: dev: __can_get_echo_skb(): fix bogous check for non-existing skb by removing it
Manfred Schlaegl [Wed, 19 Dec 2018 18:39:58 +0000 (19:39 +0100)]
can: dev: __can_get_echo_skb(): fix bogous check for non-existing skb by removing it

This patch revert commit 7da11ba5c506
("can: dev: __can_get_echo_skb(): print error message, if trying to echo non existing skb")

After introduction of this change we encountered following new error
message on various i.MX plattforms (flexcan):

| flexcan 53fc8000.can can0: __can_get_echo_skb: BUG! Trying to echo non
| existing skb: can_priv::echo_skb[0]

The introduction of the message was a mistake because
priv->echo_skb[idx] = NULL is a perfectly valid in following case: If
CAN_RAW_LOOPBACK is disabled (setsockopt) in applications, the pkt_type
of the tx skb's given to can_put_echo_skb is set to PACKET_LOOPBACK. In
this case can_put_echo_skb will not set priv->echo_skb[idx]. It is
therefore kept NULL.

As additional argument for revert: The order of check and usage of idx
was changed. idx is used to access an array element before checking it's
boundaries.

Signed-off-by: Manfred Schlaegl <manfred.schlaegl@ginzinger.com>
Fixes: 7da11ba5c506 ("can: dev: __can_get_echo_skb(): print error message, if trying to echo non existing skb")
Cc: linux-stable <stable@vger.kernel.org>
Signed-off-by: Marc Kleine-Budde <mkl@pengutronix.de>
5 years agodrm/sun4i: hdmi: Fix usage of TMDS clock
Priit Laes [Tue, 22 Jan 2019 07:32:32 +0000 (09:32 +0200)]
drm/sun4i: hdmi: Fix usage of TMDS clock

Although TMDS clock is required for HDMI to properly function,
nobody called clk_prepare_enable(). This fixes reference counting
issues and makes sure clock is running when it needs to be running.

Due to TDMS clock being parent clock for DDC clock, TDMS clock
was turned on/off for each EDID probe, causing spurious failures
for certain HDMI/DVI screens.

Fixes: 9c5681011a0c ("drm/sun4i: Add HDMI support")
Signed-off-by: Priit Laes <priit.laes@paf.com>
[Maxime: Moved the TMDS clock enable earlier]
Signed-off-by: Maxime Ripard <maxime.ripard@bootlin.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190122073232.7240-1-plaes@plaes.org
5 years agoserial: fsl_lpuart: fix maximum acceptable baud rate with over-sampling
Tomonori Sakita [Mon, 21 Jan 2019 08:34:16 +0000 (17:34 +0900)]
serial: fsl_lpuart: fix maximum acceptable baud rate with over-sampling

Using over-sampling ratio, lpuart can accept baud rate upto uartclk / 4.

Signed-off-by: Tomonori Sakita <tomonori.sakita@sord.co.jp>
Signed-off-by: Atsushi Nemoto <atsushi.nemoto@sord.co.jp>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agotty: serial: qcom_geni_serial: Allow mctrl when flow control is disabled
Matthias Kaehlcke [Sat, 19 Jan 2019 00:23:05 +0000 (16:23 -0800)]
tty: serial: qcom_geni_serial: Allow mctrl when flow control is disabled

The geni set/get_mctrl() functions currently do nothing unless
hardware flow control is enabled. Remove this arbitrary limitation.

Suggested-by: Johan Hovold <johan@kernel.org>
Fixes: 8a8a66a1a18a ("tty: serial: qcom_geni_serial: Add support for flow control")
Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
Reviewed-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agommc: dw_mmc-bluefield: : Fix the license information
Liming Sun [Fri, 18 Jan 2019 18:12:06 +0000 (13:12 -0500)]
mmc: dw_mmc-bluefield: : Fix the license information

The SPDX license identifier and the boiler plate text are
contradicting. Only the SPDX license identifier is needed. The
other one is removed.

Fixes: 86958dcc5ad7 ("mmc: dw_mmc-bluefield: Add driver extension")
Cc: stable@vger.kernel.org
Reviewed-by: David Woods <dwoods@mellanox.com>
Signed-off-by: Liming Sun <lsun@mellanox.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
5 years agommc: meson-gx: enable signal re-sampling together with tuning
Andreas Fenkart [Fri, 18 Jan 2019 13:32:23 +0000 (14:32 +0100)]
mmc: meson-gx: enable signal re-sampling together with tuning

card detection fails on "BeeLink Mini M8 SII" if enabled too early
mmc1: error -110 whilst initialising MMC card

Fixes: 71645e65729f ("mmc: meson-gx: add signal resampling")
Signed-off-by: Andreas Fenkart <afenkart@gmail.com>
Reviewed-by: Jerome Brunet <jbrunet@baylibre.com>
Tested-by: Jerome Brunet <jbrunet@baylibre.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
5 years agoMerge tag 'xarray-5.0-rc3' of git://git.infradead.org/users/willy/linux-dax
Linus Torvalds [Tue, 22 Jan 2019 04:08:30 +0000 (17:08 +1300)]
Merge tag 'xarray-5.0-rc3' of git://git.infradead.org/users/willy/linux-dax

Pull XArray fixes from Matthew Wilcox:
 "Fix some oversights in the XArray porcelain API:

   - support for m68k's two-byte aligned pointers

   - reserving entries using xa_insert()

   - missing xa_insert_bh() and xa_insert_irq() functions

   - simplify using xa_for_each()

   - use lockdep correctly

   - a few other minor fixes and improvements"

* tag 'xarray-5.0-rc3' of git://git.infradead.org/users/willy/linux-dax:
  XArray: Fix an arithmetic error in xa_is_err
  XArray tests: Check mark 2 gets squashed
  XArray: Fix typo in comment
  XArray: Honour reserved entries in xa_insert
  XArray: Permit storing 2-byte-aligned pointers
  XArray: Change xa_for_each iterator
  XArray: Turn xa_init_flags into a static inline
  XArray tests: Add RCU locking

5 years agoide: fix a typo in the settings proc file name
Christoph Hellwig [Thu, 20 Dec 2018 16:16:53 +0000 (17:16 +0100)]
ide: fix a typo in the settings proc file name

Fixes: ec7d9c9ce8 ("ide: replace ->proc_fops with ->proc_show")
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agodrm/amd/powerplay: OD setting fix on Vega10
Kenneth Feng [Fri, 18 Jan 2019 10:08:19 +0000 (18:08 +0800)]
drm/amd/powerplay: OD setting fix on Vega10

gfxclk for OD setting is limited to 1980M for non-acg
ASICs of Vega10

Signed-off-by: Kenneth Feng <kenneth.feng@amd.com>
Reviewed-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
5 years agoMerge tag 'iommu-fixes-v5.0-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 21 Jan 2019 18:27:17 +0000 (07:27 +1300)]
Merge tag 'iommu-fixes-v5.0-rc3' of git://git./linux/kernel/git/joro/iommu

Pull IOMMU fix from Joerg Roedel:
 "One fix only for now: Fix probe deferral in iommu/of code (broke with
  recent changes to iommu_ops->add_device invocation)"

* tag 'iommu-fixes-v5.0-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu/of: Fix probe-deferral

5 years agoMerge tag 'arc-5.0-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc
Linus Torvalds [Mon, 21 Jan 2019 18:15:58 +0000 (07:15 +1300)]
Merge tag 'arc-5.0-rc3' of git://git./linux/kernel/git/vgupta/arc

Pull ARC architecture updates from Vineet Gupta:

 - Perf support for raw events

 - boot log printing: return stack, action points

 - fix memset to avoid prefetchw bleeding past end of buffer

 - do_page_fault fix for mmap_sem held while returning to userspace

 - other misc fixes

* tag 'arc-5.0-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/vgupta/arc:
  ARCv2: lib: memeset: fix doing prefetchw outside of buffer
  ARC: mm: do_page_fault fixes #1: relinquish mmap_sem if signal arrives while handle_mm_fault
  ARC: show_regs: lockdep: re-enable preemption
  ARC: show_regs: lockdep: avoid page allocator...
  ARC: perf: avoid kernel killing where it is possible
  ARC: perf: move HW events mapping to separate function
  ARC: perf: introduce Kernel PMU events support
  ARC: perf: trivial code cleanup
  ARC: perf: map generic branches to correct hardware condition
  ARC: adjust memblock_reserve of kernel memory
  arc: remove redundant kernel-space generic-y
  ARC: fix __ffs return value to avoid build warnings
  ARC: boot log: print Action point details
  ARCv2: boot log: BPU return stack depth

5 years agoacpi/nfit: Fix command-supported detection
Dan Williams [Sat, 19 Jan 2019 18:55:04 +0000 (10:55 -0800)]
acpi/nfit: Fix command-supported detection

The _DSM function number validation only happens to succeed when the
generic Linux command number translation corresponds with a
DSM-family-specific function number. This breaks NVDIMM-N
implementations that correctly implement _LSR, _LSW, and _LSI, but do
not happen to publish support for DSM function numbers 4, 5, and 6.

Recall that the support for _LS{I,R,W} family of methods results in the
DIMM being marked as supporting those command numbers at
acpi_nfit_register_dimms() time. The DSM function mask is only used for
ND_CMD_CALL support of non-NVDIMM_FAMILY_INTEL devices.

Fixes: 31eca76ba2fc ("nfit, libnvdimm: limited/whitelisted dimm command...")
Cc: <stable@vger.kernel.org>
Link: https://github.com/pmem/ndctl/issues/78
Reported-by: Sujith Pandel <sujith_pandel@dell.com>
Tested-by: Sujith Pandel <sujith_pandel@dell.com>
Reviewed-by: Vishal Verma <vishal.l.verma@intel.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
5 years agoacpi/nfit: Block function zero DSMs
Dan Williams [Mon, 14 Jan 2019 22:07:19 +0000 (14:07 -0800)]
acpi/nfit: Block function zero DSMs

In preparation for using function number 0 as an error value, prevent it
from being considered a valid function value by acpi_nfit_ctl().

Cc: <stable@vger.kernel.org>
Cc: stuart hayes <stuart.w.hayes@gmail.com>
Fixes: e02fb7264d8a ("nfit: add Microsoft NVDIMM DSM command set...")
Reported-by: Jeff Moyer <jmoyer@redhat.com>
Reviewed-by: Jeff Moyer <jmoyer@redhat.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
5 years agolibnvdimm/security: Require nvdimm_security_setup_events() to succeed
Dan Williams [Sat, 19 Jan 2019 16:45:56 +0000 (08:45 -0800)]
libnvdimm/security: Require nvdimm_security_setup_events() to succeed

The following warning:

    ACPI0012:00: security event setup failed: -19

...is meant to capture exceptional failures of sysfs_get_dirent(),
however it will also fail in the common case when security support is
disabled. A few issues:

1/ A dev_warn() report for a common case is too chatty
2/ The setup of this notifier is generic, no need for it to be driven
   from the nfit driver, it can exist completely in the core.
3/ If it fails for any reason besides security support being disabled,
   that's fatal and should abort DIMM activation. Userspace may hang if
   it never gets overwrite notifications.
4/ The dirent needs to be released.

Move the call to the core 'dimm' driver, make it conditional on security
support being active, make it fatal for the exceptional case, add the
missing sysfs_put() at device disable time.

Fixes: 7d988097c546 ("...Add security DSM overwrite support")
Reviewed-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
5 years agonfit_test: fix security state pull for nvdimm security nfit_test
Dave Jiang [Tue, 15 Jan 2019 01:41:04 +0000 (18:41 -0700)]
nfit_test: fix security state pull for nvdimm security nfit_test

The override status function needs to be updated to use the proper
request parameter in order to get the security state.

Fixes: 3c13e2ac747a ("...Add test support for Intel nvdimm security DSMs")
Reported-by: Vishal Verma <vishal.l.verma@intel.com>
Signed-off-by: Dave Jiang <dave.jiang@intel.com>
Signed-off-by: Dan Williams <dan.j.williams@intel.com>
5 years agotty: Handle problem if line discipline does not have receive_buf
Greg Kroah-Hartman [Sun, 20 Jan 2019 09:46:58 +0000 (10:46 +0100)]
tty: Handle problem if line discipline does not have receive_buf

Some tty line disciplines do not have a receive buf callback, so
properly check for that before calling it.  If they do not have this
callback, just eat the character quietly, as we can't fail this call.

Reported-by: Jann Horn <jannh@google.com>
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
5 years agodm: fix redundant IO accounting for bios that need splitting
Mike Snitzer [Thu, 17 Jan 2019 15:48:01 +0000 (10:48 -0500)]
dm: fix redundant IO accounting for bios that need splitting

The risk of redundant IO accounting was not taken into consideration
when commit 18a25da84354 ("dm: ensure bio submission follows a
depth-first tree walk") introduced IO splitting in terms of recursion
via generic_make_request().

Fix this by subtracting the split bio's payload from the IO stats that
were already accounted for by start_io_acct() upon dm_make_request()
entry.  This repeat oscillation of the IO accounting, up then down,
isn't ideal but refactoring DM core's IO splitting to pre-split bios
_before_ they are accounted turned out to be an excessive amount of
change that will need a full development cycle to refine and verify.

Before this fix:

  /dev/mapper/stripe_dev is a 4-way stripe using a 32k chunksize, so
  bios are split on 32k boundaries.

  # fio --name=16M --filename=/dev/mapper/stripe_dev --rw=write --bs=64k --size=16M \
     --iodepth=1 --ioengine=libaio --direct=1 --refill_buffers

  with debugging added:
  [103898.310264] device-mapper: core: start_io_acct: dm-2 WRITE bio->bi_iter.bi_sector=0 len=128
  [103898.318704] device-mapper: core: __split_and_process_bio: recursing for following split bio:
  [103898.329136] device-mapper: core: start_io_acct: dm-2 WRITE bio->bi_iter.bi_sector=64 len=64
  ...

  16M written yet 136M (278528 * 512b) accounted:
  # cat /sys/block/dm-2/stat | awk '{ print $7 }'
  278528

After this fix:

  16M written and 16M (32768 * 512b) accounted:
  # cat /sys/block/dm-2/stat | awk '{ print $7 }'
  32768

Fixes: 18a25da84354 ("dm: ensure bio submission follows a depth-first tree walk")
Cc: stable@vger.kernel.org # 4.16+
Reported-by: Bryan Gurney <bgurney@redhat.com>
Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
5 years agodm: fix clone_bio() to trigger blk_recount_segments()
Mike Snitzer [Wed, 16 Jan 2019 23:53:26 +0000 (18:53 -0500)]
dm: fix clone_bio() to trigger blk_recount_segments()

DM's clone_bio() now benefits from using bio_trim() by fixing the fact
that clone_bio() wasn't clearing BIO_SEG_VALID like bio_trim() does;
which triggers blk_recount_segments() via bio_phys_segments().

Reviewed-by: Ming Lei <ming.lei@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
5 years agoALSA: hda - Add mute LED support for HP ProBook 470 G5
Anthony Wong [Sat, 19 Jan 2019 04:22:31 +0000 (12:22 +0800)]
ALSA: hda - Add mute LED support for HP ProBook 470 G5

Support speaker and mic mute LEDs on HP ProBook 470 G5.

BugLink: https://bugs.launchpad.net/bugs/1811254
Signed-off-by: Anthony Wong <anthony.wong@canonical.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
5 years agoceph: quota: cleanup license mess
Thomas Gleixner [Thu, 17 Jan 2019 23:14:22 +0000 (00:14 +0100)]
ceph: quota: cleanup license mess

Precise and non-ambiguous license information is important. The recently
added quota.c file has a SPDX license identifier, which is nice, but
at the same time it has a contradictionary license boiler plate text.

  SPDX-License-Identifier: GPL-2.0

versus

  * This program is free software; you can redistribute it and/or
  * modify it under the terms of the GNU General Public License
  * as published by the Free Software Foundation; either version 2
  * of the License, or (at your option) any later version.

Oh well.

As the other ceph related files are licensed under the GPL v2 only, it's
assumed that the SPDX id is correct and the boiler plate was randomly
copied into that patch.

Remove the boiler plate as it is wrong and even if correct it is redundant.

Fixes: fb18a57568c2 ("ceph: quota: add initial infrastructure to support cephfs quotas")
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: Luis Henriques <lhenriques@suse.com>
Cc: Jiri Kosina <jkosina@suse.cz>
Cc: "Yan, Zheng" <zyan@redhat.com>
Cc: Sage Weil <sage@redhat.com>
Cc: Ilya Dryomov <idryomov@gmail.com>
Cc: ceph-devel@vger.kernel.org
Acked-by: Luis Henriques <lhenriques@suse.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
5 years agolibceph: avoid KEEPALIVE_PENDING races in ceph_con_keepalive()
Ilya Dryomov [Mon, 14 Jan 2019 20:13:10 +0000 (21:13 +0100)]
libceph: avoid KEEPALIVE_PENDING races in ceph_con_keepalive()

con_fault() can transition the connection into STANDBY right after
ceph_con_keepalive() clears STANDBY in clear_standby():

    libceph user thread               ceph-msgr worker

ceph_con_keepalive()
  mutex_lock(&con->mutex)
  clear_standby(con)
  mutex_unlock(&con->mutex)
                                mutex_lock(&con->mutex)
                                con_fault()
                                  ...
                                  if KEEPALIVE_PENDING isn't set
                                    set state to STANDBY
                                  ...
                                mutex_unlock(&con->mutex)
  set KEEPALIVE_PENDING
  set WRITE_PENDING

This triggers warnings in clear_standby() when either ceph_con_send()
or ceph_con_keepalive() get to clearing STANDBY next time.

I don't see a reason to condition queue_con() call on the previous
value of KEEPALIVE_PENDING, so move the setting of KEEPALIVE_PENDING
into the critical section -- unlike WRITE_PENDING, KEEPALIVE_PENDING
could have been a non-atomic flag.

Reported-by: syzbot+acdeb633f6211ccdf886@syzkaller.appspotmail.com
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Tested-by: Myungho Jung <mhjungk@gmail.com>
5 years agoceph: clear inode pointer when snap realm gets dropped by its inode
Yan, Zheng [Thu, 10 Jan 2019 07:41:09 +0000 (15:41 +0800)]
ceph: clear inode pointer when snap realm gets dropped by its inode

snap realm and corresponding inode have pointers to each other.
The two pointer should get clear at the same time. Otherwise,
snap realm's pointer may reference freed inode.

Cc: stable@vger.kernel.org # 4.17+
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Luis Henriques <lhenriques@suse.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
5 years agolocking/rwsem: Fix (possible) missed wakeup
Xie Yongji [Thu, 29 Nov 2018 12:50:30 +0000 (20:50 +0800)]
locking/rwsem: Fix (possible) missed wakeup

Because wake_q_add() can imply an immediate wakeup (cmpxchg failure
case), we must not rely on the wakeup being delayed. However, commit:

  e38513905eea ("locking/rwsem: Rework zeroing reader waiter->task")

relies on exactly that behaviour in that the wakeup must not happen
until after we clear waiter->task.

[ peterz: Added changelog. ]

Signed-off-by: Xie Yongji <xieyongji@baidu.com>
Signed-off-by: Zhang Yu <zhangyu31@baidu.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: e38513905eea ("locking/rwsem: Rework zeroing reader waiter->task")
Link: https://lkml.kernel.org/r/1543495830-2644-1-git-send-email-xieyongji@baidu.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agofutex: Fix (possible) missed wakeup
Peter Zijlstra [Thu, 29 Nov 2018 13:44:49 +0000 (14:44 +0100)]
futex: Fix (possible) missed wakeup

We must not rely on wake_q_add() to delay the wakeup; in particular
commit:

  1d0dcb3ad9d3 ("futex: Implement lockless wakeups")

moved wake_q_add() before smp_store_release(&q->lock_ptr, NULL), which
could result in futex_wait() waking before observing ->lock_ptr ==
NULL and going back to sleep again.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 1d0dcb3ad9d3 ("futex: Implement lockless wakeups")
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agosched/wake_q: Fix wakeup ordering for wake_q
Peter Zijlstra [Mon, 17 Dec 2018 09:14:53 +0000 (10:14 +0100)]
sched/wake_q: Fix wakeup ordering for wake_q

Notable cmpxchg() does not provide ordering when it fails, however
wake_q_add() requires ordering in this specific case too. Without this
it would be possible for the concurrent wakeup to not observe our
prior state.

Andrea Parri provided:

  C wake_up_q-wake_q_add

  {
int next = 0;
int y = 0;
  }

  P0(int *next, int *y)
  {
int r0;

/* in wake_up_q() */

WRITE_ONCE(*next, 1);   /* node->next = NULL */
smp_mb();               /* implied by wake_up_process() */
r0 = READ_ONCE(*y);
  }

  P1(int *next, int *y)
  {
int r1;

/* in wake_q_add() */

WRITE_ONCE(*y, 1);      /* wake_cond = true */
smp_mb__before_atomic();
r1 = cmpxchg_relaxed(next, 1, 2);
  }

  exists (0:r0=0 /\ 1:r1=0)

  This "exists" clause cannot be satisfied according to the LKMM:

  Test wake_up_q-wake_q_add Allowed
  States 3
  0:r0=0; 1:r1=1;
  0:r0=1; 1:r1=0;
  0:r0=1; 1:r1=1;
  No
  Witnesses
  Positive: 0 Negative: 3
  Condition exists (0:r0=0 /\ 1:r1=0)
  Observation wake_up_q-wake_q_add Never 0 3

Reported-by: Yongji Xie <elohimes@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Waiman Long <longman@redhat.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agosched/wake_q: Document wake_q_add()
Peter Zijlstra [Mon, 17 Dec 2018 09:14:53 +0000 (10:14 +0100)]
sched/wake_q: Document wake_q_add()

The only guarantee provided by wake_q_add() is that a wakeup will
happen after it, it does _NOT_ guarantee the wakeup will be delayed
until the matching wake_up_q().

If wake_q_add() fails the cmpxchg() a concurrent wakeup is pending and
that can happen at any time after the cmpxchg(). This means we should
not rely on the wakeup happening at wake_q_up(), but should be ready
for wake_q_add() to issue the wakeup.

The delay; if provided (most likely); should only result in more efficient
behaviour.

Reported-by: Yongji Xie <elohimes@gmail.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Waiman Long <longman@redhat.com>
Cc: Will Deacon <will.deacon@arm.com>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agosched/wait: Fix rcuwait_wake_up() ordering
Prateek Sood [Fri, 30 Nov 2018 15:10:56 +0000 (20:40 +0530)]
sched/wait: Fix rcuwait_wake_up() ordering

For some peculiar reason rcuwait_wake_up() has the right barrier in
the comment, but not in the code.

This mistake has been observed to cause a deadlock in the following
situation:

    P1 P2

    percpu_up_read() percpu_down_write()
      rcu_sync_is_idle() // false
  rcu_sync_enter()
  ...
      __percpu_up_read()

[S] ,-  __this_cpu_dec(*sem->read_count)
    |   smp_rmb();
[L] |   task = rcu_dereference(w->task) // NULL
    |
    |     [S]     w->task = current
    |     smp_mb();
    |     [L]     readers_active_check() // fail
    `-> <store happens here>

Where the smp_rmb() (obviously) fails to constrain the store.

[ peterz: Added changelog. ]

Signed-off-by: Prateek Sood <prsood@codeaurora.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Reviewed-by: Andrea Parri <andrea.parri@amarulasolutions.com>
Acked-by: Davidlohr Bueso <dbueso@suse.de>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Fixes: 8f95c90ceb54 ("sched/wait, RCU: Introduce rcuwait machinery")
Link: https://lkml.kernel.org/r/1543590656-7157-1-git-send-email-prsood@codeaurora.org
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agoLinux 5.0-rc3 v5.0-rc3
Linus Torvalds [Mon, 21 Jan 2019 00:14:44 +0000 (13:14 +1300)]
Linux 5.0-rc3

5 years agoMerge tag 'pstore-v5.0-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/kees...
Linus Torvalds [Mon, 21 Jan 2019 00:12:03 +0000 (13:12 +1300)]
Merge tag 'pstore-v5.0-rc4' of git://git./linux/kernel/git/kees/linux

Pull pstore fixes from Kees Cook:

 - Fix console ramoops to show the previous boot logs (Sai Prakash
   Ranjan)

 - Avoid allocation and leak of platform data

* tag 'pstore-v5.0-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
  pstore/ram: Avoid allocation and leak of platform data
  pstore/ram: Fix console ramoops to show the previous boot logs

5 years agoMerge tag 'gcc-plugins-v5.0-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Mon, 21 Jan 2019 00:07:03 +0000 (13:07 +1300)]
Merge tag 'gcc-plugins-v5.0-rc4' of git://git./linux/kernel/git/kees/linux

Pull gcc-plugins fixes from Kees Cook:
 "Fix ARM per-task stack protector plugin under GCC 9 (Ard Biesheuvel)"

* tag 'gcc-plugins-v5.0-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/kees/linux:
  gcc-plugins: arm_ssp_per_task_plugin: fix for GCC 9+
  gcc-plugins: arm_ssp_per_task_plugin: sign extend the SP mask

5 years agoMerge git://git.kernel.org/pub/scm/linux/kernel/git/davem/net
Linus Torvalds [Sun, 20 Jan 2019 23:52:31 +0000 (12:52 +1300)]
Merge git://git./linux/kernel/git/davem/net

Pull networking fixes from David Miller:

 1) Fix endless loop in nf_tables, from Phil Sutter.

 2) Fix cross namespace ip6_gre tunnel hash list corruption, from
    Olivier Matz.

 3) Don't be too strict in phy_start_aneg() otherwise we might not allow
    restarting auto negotiation. From Heiner Kallweit.

 4) Fix various KMSAN uninitialized value cases in tipc, from Ying Xue.

 5) Memory leak in act_tunnel_key, from Davide Caratti.

 6) Handle chip errata of mv88e6390 PHY, from Andrew Lunn.

 7) Remove linear SKB assumption in fou/fou6, from Eric Dumazet.

 8) Missing udplite rehash callbacks, from Alexey Kodanev.

 9) Log dirty pages properly in vhost, from Jason Wang.

10) Use consume_skb() in neigh_probe() as this is a normal free not a
    drop, from Yang Wei. Likewise in macvlan_process_broadcast().

11) Missing device_del() in mdiobus_register() error paths, from Thomas
    Petazzoni.

12) Fix checksum handling of short packets in mlx5, from Cong Wang.

* git://git.kernel.org/pub/scm/linux/kernel/git/davem/net: (96 commits)
  bpf: in __bpf_redirect_no_mac pull mac only if present
  virtio_net: bulk free tx skbs
  net: phy: phy driver features are mandatory
  isdn: avm: Fix string plus integer warning from Clang
  net/mlx5e: Fix cb_ident duplicate in indirect block register
  net/mlx5e: Fix wrong (zero) TX drop counter indication for representor
  net/mlx5e: Fix wrong error code return on FEC query failure
  net/mlx5e: Force CHECKSUM_UNNECESSARY for short ethernet frames
  tools: bpftool: Cleanup license mess
  bpf: fix inner map masking to prevent oob under speculation
  bpf: pull in pkt_sched.h header for tooling to fix bpftool build
  selftests: forwarding: Add a test case for externally learned FDB entries
  selftests: mlxsw: Test FDB offload indication
  mlxsw: spectrum_switchdev: Do not treat static FDB entries as sticky
  net: bridge: Mark FDB entries that were added by user as such
  mlxsw: spectrum_fid: Update dummy FID index
  mlxsw: pci: Return error on PCI reset timeout
  mlxsw: pci: Increase PCI SW reset timeout
  mlxsw: pci: Ring CQ's doorbell before RDQ's
  MAINTAINERS: update email addresses of liquidio driver maintainers
  ...

5 years agopstore/ram: Avoid allocation and leak of platform data
Kees Cook [Sun, 20 Jan 2019 22:33:34 +0000 (14:33 -0800)]
pstore/ram: Avoid allocation and leak of platform data

Yue Hu noticed that when parsing device tree the allocated platform data
was never freed. Since it's not used beyond the function scope, this
switches to using a stack variable instead.

Reported-by: Yue Hu <huyue2@yulong.com>
Fixes: 35da60941e44 ("pstore/ram: add Device Tree bindings")
Cc: stable@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
5 years agogcc-plugins: arm_ssp_per_task_plugin: fix for GCC 9+
Ard Biesheuvel [Fri, 18 Jan 2019 10:58:07 +0000 (11:58 +0100)]
gcc-plugins: arm_ssp_per_task_plugin: fix for GCC 9+

GCC 9 reworks the way the references to the stack canary are
emitted, to prevent the value from being spilled to the stack
before the final comparison in the epilogue, defeating the
purpose, given that the spill slot is under control of the
attacker that we are protecting ourselves from.

Since our canary value address is obtained without accessing
memory (as opposed to pre-v7 code that will obtain it from a
literal pool), it is unlikely (although not guaranteed) that
the compiler will spill the canary value in the same way, so
let's just disable this improvement when building with GCC9+.

Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Kees Cook <keescook@chromium.org>
5 years agogcc-plugins: arm_ssp_per_task_plugin: sign extend the SP mask
Ard Biesheuvel [Fri, 18 Jan 2019 10:58:06 +0000 (11:58 +0100)]
gcc-plugins: arm_ssp_per_task_plugin: sign extend the SP mask

The ARM per-task stack protector GCC plugin hits an assert in
the compiler in some case, due to the fact the the SP mask
expression is not sign-extended as it should be. So fix that.

Suggested-by: Kugan Vivekanandarajah <kugan.vivekanandarajah@linaro.org>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-by: Kees Cook <keescook@chromium.org>