wrapfs-5.10.y.git
4 years agonvme-rdma: fix possible hang when failing to set io queues
Sagi Grimberg [Mon, 15 Mar 2021 21:04:27 +0000 (14:04 -0700)]
nvme-rdma: fix possible hang when failing to set io queues

[ Upstream commit c4c6df5fc84659690d4391d1fba155cd94185295 ]

We only setup io queues for nvme controllers, and it makes absolutely no
sense to allow a controller (re)connect without any I/O queues.  If we
happen to fail setting the queue count for any reason, we should not allow
this to be a successful reconnect as I/O has no chance in going through.
Instead just fail and schedule another reconnect.

Reported-by: Chao Leng <lengchao@huawei.com>
Fixes: 711023071960 ("nvme-rdma: add a NVMe over Fabrics RDMA host driver")
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chao Leng <lengchao@huawei.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agocounter: stm32-timer-cnt: Report count function when SLAVE_MODE_DISABLED
William Breathitt Gray [Fri, 26 Feb 2021 01:29:31 +0000 (10:29 +0900)]
counter: stm32-timer-cnt: Report count function when SLAVE_MODE_DISABLED

[ Upstream commit fae6f62e6a580b663ecf42c2120a0898deae9137 ]

When in SLAVE_MODE_DISABLED mode, the count still increases if the
counter is enabled because an internal clock is used. This patch fixes
the stm32_count_function_get() and stm32_count_function_set() functions
to properly handle this behavior.

Fixes: ad29937e206f ("counter: Add STM32 Timer quadrature encoder")
Cc: Fabrice Gasnier <fabrice.gasnier@st.com>
Cc: Maxime Coquelin <mcoquelin.stm32@gmail.com>
Cc: Alexandre Torgue <alexandre.torgue@st.com>
Signed-off-by: William Breathitt Gray <vilhelm.gray@gmail.com>
Reviewed-by: Fabrice Gasnier <fabrice.gasnier@foss.st.com>
Link: https://lore.kernel.org/r/20210226012931.161429-1-vilhelm.gray@gmail.com
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoscsi: myrs: Fix a double free in myrs_cleanup()
Lv Yunlong [Thu, 11 Mar 2021 06:30:05 +0000 (22:30 -0800)]
scsi: myrs: Fix a double free in myrs_cleanup()

commit 2bb817712e2f77486d6ee17e7efaf91997a685f8 upstream.

In myrs_cleanup(), cs->mmio_base will be freed twice by iounmap().

Link: https://lore.kernel.org/r/20210311063005.9963-1-lyl2019@mail.ustc.edu.cn
Fixes: 77266186397c ("scsi: myrs: Add Mylex RAID controller (SCSI interface)")
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Lv Yunlong <lyl2019@mail.ustc.edu.cn>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoscsi: lpfc: Fix some error codes in debugfs
Dan Carpenter [Fri, 12 Mar 2021 07:42:11 +0000 (10:42 +0300)]
scsi: lpfc: Fix some error codes in debugfs

commit 19f1bc7edf0f97186810e13a88f5b62069d89097 upstream.

If copy_from_user() or kstrtoull() fail then the correct behavior is to
return a negative error code.

Link: https://lore.kernel.org/r/YEsbU/UxYypVrC7/@mwanda
Fixes: f9bb2da11db8 ("[SCSI] lpfc 8.3.27: T10 additions for SLI4")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoriscv: Correct SPARSEMEM configuration
Kefeng Wang [Mon, 15 Mar 2021 12:03:07 +0000 (20:03 +0800)]
riscv: Correct SPARSEMEM configuration

commit a5406a7ff56e63376c210b06072aa0ef23473366 upstream.

There are two issues for RV32,
1) if use FLATMEM, it is useless to enable SPARSEMEM_STATIC.
2) if use SPARSMEM, both SPARSEMEM_VMEMMAP and SPARSEMEM_STATIC is enabled.

Fixes: d95f1a542c3d ("RISC-V: Implement sparsemem")
Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com>
Cc: stable@vger.kernel.org
Signed-off-by: Palmer Dabbelt <palmerdabbelt@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agokbuild: Fix <linux/version.h> for empty SUBLEVEL or PATCHLEVEL again
Masahiro Yamada [Sat, 27 Feb 2021 14:20:23 +0000 (23:20 +0900)]
kbuild: Fix <linux/version.h> for empty SUBLEVEL or PATCHLEVEL again

commit 207da4c82ade9a6d59f7e794d737ba0748613fa2 upstream.

Commit 78d3bb4483ba ("kbuild: Fix <linux/version.h> for empty SUBLEVEL
or PATCHLEVEL") fixed the build error for empty SUBLEVEL or PATCHLEVEL
by prepending a zero.

Commit 9b82f13e7ef3 ("kbuild: clamp SUBLEVEL to 255") re-introduced
this issue.

This time, we cannot take the same approach because we have C code:

  #define LINUX_VERSION_PATCHLEVEL $(PATCHLEVEL)
  #define LINUX_VERSION_SUBLEVEL $(SUBLEVEL)

Replace empty SUBLEVEL/PATCHLEVEL with a zero.

Fixes: 9b82f13e7ef3 ("kbuild: clamp SUBLEVEL to 255")
Reported-by: Christian Zigotzky <chzigotzky@xenosoft.de>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
Reviewed-and-tested-by: Sasha Levin <sashal@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonet/qrtr: fix __netdev_alloc_skb call
Pavel Skripkin [Sun, 28 Feb 2021 23:22:40 +0000 (02:22 +0300)]
net/qrtr: fix __netdev_alloc_skb call

commit 093b036aa94e01a0bea31a38d7f0ee28a2749023 upstream.

syzbot found WARNING in __alloc_pages_nodemask()[1] when order >= MAX_ORDER.
It was caused by a huge length value passed from userspace to qrtr_tun_write_iter(),
which tries to allocate skb. Since the value comes from the untrusted source
there is no need to raise a warning in __alloc_pages_nodemask().

[1] WARNING in __alloc_pages_nodemask+0x5f8/0x730 mm/page_alloc.c:5014
Call Trace:
 __alloc_pages include/linux/gfp.h:511 [inline]
 __alloc_pages_node include/linux/gfp.h:524 [inline]
 alloc_pages_node include/linux/gfp.h:538 [inline]
 kmalloc_large_node+0x60/0x110 mm/slub.c:3999
 __kmalloc_node_track_caller+0x319/0x3f0 mm/slub.c:4496
 __kmalloc_reserve net/core/skbuff.c:150 [inline]
 __alloc_skb+0x4e4/0x5a0 net/core/skbuff.c:210
 __netdev_alloc_skb+0x70/0x400 net/core/skbuff.c:446
 netdev_alloc_skb include/linux/skbuff.h:2832 [inline]
 qrtr_endpoint_post+0x84/0x11b0 net/qrtr/qrtr.c:442
 qrtr_tun_write_iter+0x11f/0x1a0 net/qrtr/tun.c:98
 call_write_iter include/linux/fs.h:1901 [inline]
 new_sync_write+0x426/0x650 fs/read_write.c:518
 vfs_write+0x791/0xa30 fs/read_write.c:605
 ksys_write+0x12d/0x250 fs/read_write.c:658
 do_syscall_64+0x2d/0x70 arch/x86/entry/common.c:46
 entry_SYSCALL_64_after_hwframe+0x44/0xa9

Reported-by: syzbot+80dccaee7c6630fa9dcf@syzkaller.appspotmail.com
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Acked-by: Alexander Lobakin <alobakin@pm.me>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agosunrpc: fix refcount leak for rpc auth modules
Daniel Kobras [Fri, 26 Feb 2021 23:04:37 +0000 (00:04 +0100)]
sunrpc: fix refcount leak for rpc auth modules

commit f1442d6349a2e7bb7a6134791bdc26cb776c79af upstream.

If an auth module's accept op returns SVC_CLOSE, svc_process_common()
enters a call path that does not call svc_authorise() before leaving the
function, and thus leaks a reference on the auth module's refcount. Hence,
make sure calls to svc_authenticate() and svc_authorise() are paired for
all call paths, to make sure rpc auth modules can be unloaded.

Signed-off-by: Daniel Kobras <kobras@puzzle-itc.de>
Fixes: 4d712ef1db05 ("svcauth_gss: Close connection when dropping an incoming message")
Link: https://lore.kernel.org/linux-nfs/3F1B347F-B809-478F-A1E9-0BE98E22B0F0@oracle.com/T/#t
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agovfio: IOMMU_API should be selected
Jason Gunthorpe [Tue, 23 Feb 2021 19:17:46 +0000 (15:17 -0400)]
vfio: IOMMU_API should be selected

commit 179209fa12709a3df8888c323b37315da2683c24 upstream.

As IOMMU_API is a kconfig without a description (eg does not show in the
menu) the correct operator is select not 'depends on'. Using 'depends on'
for this kind of symbol means VFIO is not selectable unless some other
random kconfig has already enabled IOMMU_API for it.

Fixes: cba3345cc494 ("vfio: VFIO core")
Signed-off-by: Jason Gunthorpe <jgg@nvidia.com>
Message-Id: <1-v1-df057e0f92c3+91-vfio_arm_compile_test_jgg@nvidia.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agosvcrdma: disable timeouts on rdma backchannel
Timo Rothenpieler [Mon, 22 Feb 2021 23:36:19 +0000 (00:36 +0100)]
svcrdma: disable timeouts on rdma backchannel

commit 6820bf77864d5894ff67b5c00d7dba8f92011e3d upstream.

This brings it in line with the regular tcp backchannel, which also has
all those timeouts disabled.

Prevents the backchannel from timing out, getting some async operations
like server side copying getting stuck indefinitely on the client side.

Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>
Fixes: 5d252f90a800 ("svcrdma: Add class for RDMA backwards direction transport")
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoNFSD: Repair misuse of sv_lock in 5.10.16-rt30.
Joe Korty [Fri, 26 Feb 2021 14:38:20 +0000 (09:38 -0500)]
NFSD: Repair misuse of sv_lock in 5.10.16-rt30.

commit c7de87ff9dac5f396f62d584f3908f80ddc0e07b upstream.

[ This problem is in mainline, but only rt has the chops to be
able to detect it. ]

Lockdep reports a circular lock dependency between serv->sv_lock and
softirq_ctl.lock on system shutdown, when using a kernel built with
CONFIG_PREEMPT_RT=y, and a nfs mount exists.

This is due to the definition of spin_lock_bh on rt:

local_bh_disable();
rt_spin_lock(lock);

which forces a softirq_ctl.lock -> serv->sv_lock dependency.  This is
not a problem as long as _every_ lock of serv->sv_lock is a:

spin_lock_bh(&serv->sv_lock);

but there is one of the form:

spin_lock(&serv->sv_lock);

This is what is causing the circular dependency splat.  The spin_lock()
grabs the lock without first grabbing softirq_ctl.lock via local_bh_disable.
If later on in the critical region,  someone does a local_bh_disable, we
get a serv->sv_lock -> softirq_ctrl.lock dependency established.  Deadlock.

Fix is to make serv->sv_lock be locked with spin_lock_bh everywhere, no
exceptions.

[  OK  ] Stopped target NFS client services.
         Stopping Logout off all iSCSI sessions on shutdown...
         Stopping NFS server and services...
[  109.442380]
[  109.442385] ======================================================
[  109.442386] WARNING: possible circular locking dependency detected
[  109.442387] 5.10.16-rt30 #1 Not tainted
[  109.442389] ------------------------------------------------------
[  109.442390] nfsd/1032 is trying to acquire lock:
[  109.442392] ffff994237617f60 ((softirq_ctrl.lock).lock){+.+.}-{2:2}, at: __local_bh_disable_ip+0xd9/0x270
[  109.442405]
[  109.442405] but task is already holding lock:
[  109.442406] ffff994245cb00b0 (&serv->sv_lock){+.+.}-{0:0}, at: svc_close_list+0x1f/0x90
[  109.442415]
[  109.442415] which lock already depends on the new lock.
[  109.442415]
[  109.442416]
[  109.442416] the existing dependency chain (in reverse order) is:
[  109.442417]
[  109.442417] -> #1 (&serv->sv_lock){+.+.}-{0:0}:
[  109.442421]        rt_spin_lock+0x2b/0xc0
[  109.442428]        svc_add_new_perm_xprt+0x42/0xa0
[  109.442430]        svc_addsock+0x135/0x220
[  109.442434]        write_ports+0x4b3/0x620
[  109.442438]        nfsctl_transaction_write+0x45/0x80
[  109.442440]        vfs_write+0xff/0x420
[  109.442444]        ksys_write+0x4f/0xc0
[  109.442446]        do_syscall_64+0x33/0x40
[  109.442450]        entry_SYSCALL_64_after_hwframe+0x44/0xa9
[  109.442454]
[  109.442454] -> #0 ((softirq_ctrl.lock).lock){+.+.}-{2:2}:
[  109.442457]        __lock_acquire+0x1264/0x20b0
[  109.442463]        lock_acquire+0xc2/0x400
[  109.442466]        rt_spin_lock+0x2b/0xc0
[  109.442469]        __local_bh_disable_ip+0xd9/0x270
[  109.442471]        svc_xprt_do_enqueue+0xc0/0x4d0
[  109.442474]        svc_close_list+0x60/0x90
[  109.442476]        svc_close_net+0x49/0x1a0
[  109.442478]        svc_shutdown_net+0x12/0x40
[  109.442480]        nfsd_destroy+0xc5/0x180
[  109.442482]        nfsd+0x1bc/0x270
[  109.442483]        kthread+0x194/0x1b0
[  109.442487]        ret_from_fork+0x22/0x30
[  109.442492]
[  109.442492] other info that might help us debug this:
[  109.442492]
[  109.442493]  Possible unsafe locking scenario:
[  109.442493]
[  109.442493]        CPU0                    CPU1
[  109.442494]        ----                    ----
[  109.442495]   lock(&serv->sv_lock);
[  109.442496]                                lock((softirq_ctrl.lock).lock);
[  109.442498]                                lock(&serv->sv_lock);
[  109.442499]   lock((softirq_ctrl.lock).lock);
[  109.442501]
[  109.442501]  *** DEADLOCK ***
[  109.442501]
[  109.442501] 3 locks held by nfsd/1032:
[  109.442503]  #0: ffffffff93b49258 (nfsd_mutex){+.+.}-{3:3}, at: nfsd+0x19a/0x270
[  109.442508]  #1: ffff994245cb00b0 (&serv->sv_lock){+.+.}-{0:0}, at: svc_close_list+0x1f/0x90
[  109.442512]  #2: ffffffff93a81b20 (rcu_read_lock){....}-{1:2}, at: rt_spin_lock+0x5/0xc0
[  109.442518]
[  109.442518] stack backtrace:
[  109.442519] CPU: 0 PID: 1032 Comm: nfsd Not tainted 5.10.16-rt30 #1
[  109.442522] Hardware name: Supermicro X9DRL-3F/iF/X9DRL-3F/iF, BIOS 3.2 09/22/2015
[  109.442524] Call Trace:
[  109.442527]  dump_stack+0x77/0x97
[  109.442533]  check_noncircular+0xdc/0xf0
[  109.442546]  __lock_acquire+0x1264/0x20b0
[  109.442553]  lock_acquire+0xc2/0x400
[  109.442564]  rt_spin_lock+0x2b/0xc0
[  109.442570]  __local_bh_disable_ip+0xd9/0x270
[  109.442573]  svc_xprt_do_enqueue+0xc0/0x4d0
[  109.442577]  svc_close_list+0x60/0x90
[  109.442581]  svc_close_net+0x49/0x1a0
[  109.442585]  svc_shutdown_net+0x12/0x40
[  109.442588]  nfsd_destroy+0xc5/0x180
[  109.442590]  nfsd+0x1bc/0x270
[  109.442595]  kthread+0x194/0x1b0
[  109.442600]  ret_from_fork+0x22/0x30
[  109.518225] nfsd: last server has exited, flushing export cache
[  OK  ] Stopped NFSv4 ID-name mapping service.
[  OK  ] Stopped GSSAPI Proxy Daemon.
[  OK  ] Stopped NFS Mount Daemon.
[  OK  ] Stopped NFS status monitor for NFSv2/3 locking..

Fixes: 719f8bcc883e ("svcrpc: fix xpt_list traversal locking on shutdown")
Signed-off-by: Joe Korty <joe.korty@concurrent-rt.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonfsd: Don't keep looking up unhashed files in the nfsd file cache
Trond Myklebust [Fri, 19 Feb 2021 02:02:07 +0000 (21:02 -0500)]
nfsd: Don't keep looking up unhashed files in the nfsd file cache

commit d30881f573e565ebb5dbb50b31ed6106b5c81328 upstream.

If a file is unhashed, then we're going to reject it anyway and retry,
so make sure we skip it when we're doing the RCU lockless lookup.
This avoids a number of unnecessary nfserr_jukebox returns from
nfsd_file_acquire()

Fixes: 65294c1f2c5e ("nfsd: add a new struct file caching facility to nfsd")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonvmet: don't check iosqes,iocqes for discovery controllers
Sagi Grimberg [Mon, 15 Mar 2021 22:34:51 +0000 (15:34 -0700)]
nvmet: don't check iosqes,iocqes for discovery controllers

commit d218a8a3003e84ab136e69a4e30dd4ec7dab2d22 upstream.

From the base spec, Figure 78:

  "Controller Configuration, these fields are defined as parameters to
   configure an "I/O Controller (IOC)" and not to configure a "Discovery
   Controller (DC).

   ...
   If the controller does not support I/O queues, then this field shall
   be read-only with a value of 0h

Just perform this check for I/O controllers.

Fixes: a07b4970f464 ("nvmet: add a generic NVMe target")
Reported-by: Belanger, Martin <Martin.Belanger@dell.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonvme-tcp: fix a NULL deref when receiving a 0-length r2t PDU
Sagi Grimberg [Mon, 15 Mar 2021 21:08:11 +0000 (14:08 -0700)]
nvme-tcp: fix a NULL deref when receiving a 0-length r2t PDU

commit fd0823f405090f9f410fc3e3ff7efb52e7b486fa upstream.

When the controller sends us a 0-length r2t PDU we should not attempt to
try to set up a h2cdata PDU but rather conclude that this is a buggy
controller (forward progress is not possible) and simply fail it
immediately.

Fixes: 3f2304f8c6d6 ("nvme-tcp: add NVMe over TCP host driver")
Reported-by: Belanger, Martin <Martin.Belanger@dell.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonvme-tcp: fix possible hang when failing to set io queues
Sagi Grimberg [Mon, 15 Mar 2021 21:04:26 +0000 (14:04 -0700)]
nvme-tcp: fix possible hang when failing to set io queues

commit 72f572428b83d0bc7028e7c4326d1a5f45205e44 upstream.

We only setup io queues for nvme controllers, and it makes absolutely no
sense to allow a controller (re)connect without any I/O queues.  If we
happen to fail setting the queue count for any reason, we should not
allow this to be a successful reconnect as I/O has no chance in going
through. Instead just fail and schedule another reconnect.

Fixes: 3f2304f8c6d6 ("nvme-tcp: add NVMe over TCP host driver")
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonvme: fix Write Zeroes limitations
Christoph Hellwig [Mon, 15 Mar 2021 09:32:07 +0000 (10:32 +0100)]
nvme: fix Write Zeroes limitations

commit b94e8cd2e6a94fc7563529ddc82726a7e77e04de upstream.

We voluntarily limit the Write Zeroes sizes to the MDTS value provided by
the hardware, but currently get the units wrong, so fix that.

Fixes: 6e02318eaea5 ("nvme: add support for the Write Zeroes command")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Tested-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Klaus Jensen <k.jensen@samsung.com>
Reviewed-by: Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com>
Reviewed-by: Himanshu Madhani <himanshu.madhani@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoafs: Stop listxattr() from listing "afs.*" attributes
David Howells [Tue, 9 Mar 2021 08:27:39 +0000 (08:27 +0000)]
afs: Stop listxattr() from listing "afs.*" attributes

commit a7889c6320b9200e3fe415238f546db677310fa9 upstream.

afs_listxattr() lists all the available special afs xattrs (i.e. those in
the "afs.*" space), no matter what type of server we're dealing with.  But
OpenAFS servers, for example, cannot deal with some of the extra-capable
attributes that AuriStor (YFS) servers provide.  Unfortunately, the
presence of the afs.yfs.* attributes causes errors[1] for anything that
tries to read them if the server is of the wrong type.

Fix the problem by removing afs_listxattr() so that none of the special
xattrs are listed (AFS doesn't support xattrs).  It does mean, however,
that getfattr won't list them, though they can still be accessed with
getxattr() and setxattr().

This can be tested with something like:

getfattr -d -m ".*" /afs/example.com/path/to/file

With this change, none of the afs.* attributes should be visible.

Changes:
ver #2:
 - Hide all of the afs.* xattrs, not just the ACL ones.

Fixes: ae46578b963f ("afs: Get YFS ACLs and information through xattrs")
Reported-by: Gaja Sophie Peters <gaja.peters@math.uni-hamburg.de>
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Gaja Sophie Peters <gaja.peters@math.uni-hamburg.de>
Reviewed-by: Jeffrey Altman <jaltman@auristor.com>
Reviewed-by: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
Link: http://lists.infradead.org/pipermail/linux-afs/2021-March/003502.html
Link: http://lists.infradead.org/pipermail/linux-afs/2021-March/003567.html
Link: http://lists.infradead.org/pipermail/linux-afs/2021-March/003573.html
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoASoC: simple-card-utils: Do not handle device clock
Sameer Pujar [Mon, 15 Mar 2021 17:31:31 +0000 (23:01 +0530)]
ASoC: simple-card-utils: Do not handle device clock

commit 8ca88d53351cc58d535b2bfc7386835378fb0db2 upstream.

This reverts commit 1e30f642cf29 ("ASoC: simple-card-utils: Fix device
module clock"). The original patch ended up breaking following platform,
which depends on set_sysclk() to configure internal PLL on wm8904 codec
and expects simple-card-utils to not update the MCLK rate.
 - "arch/arm64/boot/dts/freescale/fsl-ls1028a-kontron-sl28-var3-ads2.dts"

It would be best if codec takes care of setting MCLK clock via DAI
set_sysclk() callback.

Reported-by: Michael Walle <michael@walle.cc>
Suggested-by: Mark Brown <broonie@kernel.org>
Suggested-by: Michael Walle <michael@walle.cc>
Fixes: 1e30f642cf29 ("ASoC: simple-card-utils: Fix device module clock")
Signed-off-by: Sameer Pujar <spujar@nvidia.com>
Tested-by: Michael Walle <michael@walle.cc>
Link: https://lore.kernel.org/r/1615829492-8972-2-git-send-email-spujar@nvidia.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoASoC: SOF: intel: fix wrong poll bits in dsp power down
Pan Xiuli [Tue, 9 Mar 2021 00:41:27 +0000 (18:41 -0600)]
ASoC: SOF: intel: fix wrong poll bits in dsp power down

commit fd8299181995093948ec6ca75432e797b4a39143 upstream.

The ADSPCS_SPA is Set Power Active bit. To check if DSP is powered
down, we need to check ADSPCS_CPA, the Current Power Active bit.

Fixes: 747503b1813a3 ("ASoC: SOF: Intel: Add Intel specific HDA DSP HW operations")
Reviewed-by: Rander Wang <rander.wang@intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Signed-off-by: Pan Xiuli <xiuli.pan@linux.intel.com>
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Link: https://lore.kernel.org/r/20210309004127.4940-1-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoASoC: SOF: Intel: unregister DMIC device on probe error
Pierre-Louis Bossart [Tue, 2 Mar 2021 00:34:10 +0000 (18:34 -0600)]
ASoC: SOF: Intel: unregister DMIC device on probe error

commit 5bb0ecddb2a7f638d65e457f3da9fa334c967b14 upstream.

We only unregister the platform device during the .remove operation,
but if the probe fails we will never reach this sequence.

Suggested-by: Bard Liao <yung-chuan.liao@linux.intel.com>
Fixes: dd96daca6c83e ("ASoC: SOF: Intel: Add APL/CNL HW DSP support")
Signed-off-by: Pierre-Louis Bossart <pierre-louis.bossart@linux.intel.com>
Reviewed-by: Ranjani Sridharan <ranjani.sridharan@linux.intel.com>
Reviewed-by: Guennadi Liakhovetski <guennadi.liakhovetski@linux.intel.com>
Link: https://lore.kernel.org/r/20210302003410.1178535-1-pierre-louis.bossart@linux.intel.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoASoC: fsl_ssi: Fix TDM slot setup for I2S mode
Alexander Shiyan [Tue, 16 Feb 2021 11:42:21 +0000 (14:42 +0300)]
ASoC: fsl_ssi: Fix TDM slot setup for I2S mode

commit 87263968516fb9507d6215d53f44052627fae8d8 upstream.

When using the driver in I2S TDM mode, the _fsl_ssi_set_dai_fmt()
function rewrites the number of slots previously set by the
fsl_ssi_set_dai_tdm_slot() function to 2 by default.
To fix this, let's use the saved slot count value or, if TDM
is not used and the slot count is not set, proceed as before.

Fixes: 4f14f5c11db1 ("ASoC: fsl_ssi: Fix number of words per frame for I2S-slave mode")
Signed-off-by: Alexander Shiyan <shc_work@mail.ru>
Acked-by: Nicolin Chen <nicoleotsuka@gmail.com>
Link: https://lore.kernel.org/r/20210216114221.26635-1-shc_work@mail.ru
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobtrfs: fix slab cache flags for free space tree bitmap
David Sterba [Mon, 15 Mar 2021 14:18:24 +0000 (15:18 +0100)]
btrfs: fix slab cache flags for free space tree bitmap

commit 34e49994d0dcdb2d31d4d2908d04f4e9ce57e4d7 upstream.

The free space tree bitmap slab cache is created with SLAB_RED_ZONE but
that's a debugging flag and not always enabled. Also the other slabs are
created with at least SLAB_MEM_SPREAD that we want as well to average
the memory placement cost.

Reported-by: Vlastimil Babka <vbabka@suse.cz>
Fixes: 3acd48507dc4 ("btrfs: fix allocation of free space cache v1 bitmap pages")
CC: stable@vger.kernel.org # 5.4+
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobtrfs: fix race when cloning extent buffer during rewind of an old root
Filipe Manana [Thu, 11 Mar 2021 14:31:05 +0000 (14:31 +0000)]
btrfs: fix race when cloning extent buffer during rewind of an old root

commit dbcc7d57bffc0c8cac9dac11bec548597d59a6a5 upstream.

While resolving backreferences, as part of a logical ino ioctl call or
fiemap, we can end up hitting a BUG_ON() when replaying tree mod log
operations of a root, triggering a stack trace like the following:

  ------------[ cut here ]------------
  kernel BUG at fs/btrfs/ctree.c:1210!
  invalid opcode: 0000 [#1] SMP KASAN PTI
  CPU: 1 PID: 19054 Comm: crawl_335 Tainted: G        W         5.11.0-2d11c0084b02-misc-next+ #89
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
  RIP: 0010:__tree_mod_log_rewind+0x3b1/0x3c0
  Code: 05 48 8d 74 10 (...)
  RSP: 0018:ffffc90001eb70b8 EFLAGS: 00010297
  RAX: 0000000000000000 RBX: ffff88812344e400 RCX: ffffffffb28933b6
  RDX: 0000000000000007 RSI: dffffc0000000000 RDI: ffff88812344e42c
  RBP: ffffc90001eb7108 R08: 1ffff11020b60a20 R09: ffffed1020b60a20
  R10: ffff888105b050f9 R11: ffffed1020b60a1f R12: 00000000000000ee
  R13: ffff8880195520c0 R14: ffff8881bc958500 R15: ffff88812344e42c
  FS:  00007fd1955e8700(0000) GS:ffff8881f5600000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00007efdb7928718 CR3: 000000010103a006 CR4: 0000000000170ee0
  Call Trace:
   btrfs_search_old_slot+0x265/0x10d0
   ? lock_acquired+0xbb/0x600
   ? btrfs_search_slot+0x1090/0x1090
   ? free_extent_buffer.part.61+0xd7/0x140
   ? free_extent_buffer+0x13/0x20
   resolve_indirect_refs+0x3e9/0xfc0
   ? lock_downgrade+0x3d0/0x3d0
   ? __kasan_check_read+0x11/0x20
   ? add_prelim_ref.part.11+0x150/0x150
   ? lock_downgrade+0x3d0/0x3d0
   ? __kasan_check_read+0x11/0x20
   ? lock_acquired+0xbb/0x600
   ? __kasan_check_write+0x14/0x20
   ? do_raw_spin_unlock+0xa8/0x140
   ? rb_insert_color+0x30/0x360
   ? prelim_ref_insert+0x12d/0x430
   find_parent_nodes+0x5c3/0x1830
   ? resolve_indirect_refs+0xfc0/0xfc0
   ? lock_release+0xc8/0x620
   ? fs_reclaim_acquire+0x67/0xf0
   ? lock_acquire+0xc7/0x510
   ? lock_downgrade+0x3d0/0x3d0
   ? lockdep_hardirqs_on_prepare+0x160/0x210
   ? lock_release+0xc8/0x620
   ? fs_reclaim_acquire+0x67/0xf0
   ? lock_acquire+0xc7/0x510
   ? poison_range+0x38/0x40
   ? unpoison_range+0x14/0x40
   ? trace_hardirqs_on+0x55/0x120
   btrfs_find_all_roots_safe+0x142/0x1e0
   ? find_parent_nodes+0x1830/0x1830
   ? btrfs_inode_flags_to_xflags+0x50/0x50
   iterate_extent_inodes+0x20e/0x580
   ? tree_backref_for_extent+0x230/0x230
   ? lock_downgrade+0x3d0/0x3d0
   ? read_extent_buffer+0xdd/0x110
   ? lock_downgrade+0x3d0/0x3d0
   ? __kasan_check_read+0x11/0x20
   ? lock_acquired+0xbb/0x600
   ? __kasan_check_write+0x14/0x20
   ? _raw_spin_unlock+0x22/0x30
   ? __kasan_check_write+0x14/0x20
   iterate_inodes_from_logical+0x129/0x170
   ? iterate_inodes_from_logical+0x129/0x170
   ? btrfs_inode_flags_to_xflags+0x50/0x50
   ? iterate_extent_inodes+0x580/0x580
   ? __vmalloc_node+0x92/0xb0
   ? init_data_container+0x34/0xb0
   ? init_data_container+0x34/0xb0
   ? kvmalloc_node+0x60/0x80
   btrfs_ioctl_logical_to_ino+0x158/0x230
   btrfs_ioctl+0x205e/0x4040
   ? __might_sleep+0x71/0xe0
   ? btrfs_ioctl_get_supported_features+0x30/0x30
   ? getrusage+0x4b6/0x9c0
   ? __kasan_check_read+0x11/0x20
   ? lock_release+0xc8/0x620
   ? __might_fault+0x64/0xd0
   ? lock_acquire+0xc7/0x510
   ? lock_downgrade+0x3d0/0x3d0
   ? lockdep_hardirqs_on_prepare+0x210/0x210
   ? lockdep_hardirqs_on_prepare+0x210/0x210
   ? __kasan_check_read+0x11/0x20
   ? do_vfs_ioctl+0xfc/0x9d0
   ? ioctl_file_clone+0xe0/0xe0
   ? lock_downgrade+0x3d0/0x3d0
   ? lockdep_hardirqs_on_prepare+0x210/0x210
   ? __kasan_check_read+0x11/0x20
   ? lock_release+0xc8/0x620
   ? __task_pid_nr_ns+0xd3/0x250
   ? lock_acquire+0xc7/0x510
   ? __fget_files+0x160/0x230
   ? __fget_light+0xf2/0x110
   __x64_sys_ioctl+0xc3/0x100
   do_syscall_64+0x37/0x80
   entry_SYSCALL_64_after_hwframe+0x44/0xa9
  RIP: 0033:0x7fd1976e2427
  Code: 00 00 90 48 8b 05 (...)
  RSP: 002b:00007fd1955e5cf8 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
  RAX: ffffffffffffffda RBX: 00007fd1955e5f40 RCX: 00007fd1976e2427
  RDX: 00007fd1955e5f48 RSI: 00000000c038943b RDI: 0000000000000004
  RBP: 0000000001000000 R08: 0000000000000000 R09: 00007fd1955e6120
  R10: 0000557835366b00 R11: 0000000000000246 R12: 0000000000000004
  R13: 00007fd1955e5f48 R14: 00007fd1955e5f40 R15: 00007fd1955e5ef8
  Modules linked in:
  ---[ end trace ec8931a1c36e57be ]---

  (gdb) l *(__tree_mod_log_rewind+0x3b1)
  0xffffffff81893521 is in __tree_mod_log_rewind (fs/btrfs/ctree.c:1210).
  1205                     * the modification. as we're going backwards, we do the
  1206                     * opposite of each operation here.
  1207                     */
  1208                    switch (tm->op) {
  1209                    case MOD_LOG_KEY_REMOVE_WHILE_FREEING:
  1210                            BUG_ON(tm->slot < n);
  1211                            fallthrough;
  1212                    case MOD_LOG_KEY_REMOVE_WHILE_MOVING:
  1213                    case MOD_LOG_KEY_REMOVE:
  1214                            btrfs_set_node_key(eb, &tm->key, tm->slot);

Here's what happens to hit that BUG_ON():

1) We have one tree mod log user (through fiemap or the logical ino ioctl),
   with a sequence number of 1, so we have fs_info->tree_mod_seq == 1;

2) Another task is at ctree.c:balance_level() and we have eb X currently as
   the root of the tree, and we promote its single child, eb Y, as the new
   root.

   Then, at ctree.c:balance_level(), we call:

      tree_mod_log_insert_root(eb X, eb Y, 1);

3) At tree_mod_log_insert_root() we create tree mod log elements for each
   slot of eb X, of operation type MOD_LOG_KEY_REMOVE_WHILE_FREEING each
   with a ->logical pointing to ebX->start. These are placed in an array
   named tm_list.
   Lets assume there are N elements (N pointers in eb X);

4) Then, still at tree_mod_log_insert_root(), we create a tree mod log
   element of operation type MOD_LOG_ROOT_REPLACE, ->logical set to
   ebY->start, ->old_root.logical set to ebX->start, ->old_root.level set
   to the level of eb X and ->generation set to the generation of eb X;

5) Then tree_mod_log_insert_root() calls tree_mod_log_free_eb() with
   tm_list as argument. After that, tree_mod_log_free_eb() calls
   __tree_mod_log_insert() for each member of tm_list in reverse order,
   from highest slot in eb X, slot N - 1, to slot 0 of eb X;

6) __tree_mod_log_insert() sets the sequence number of each given tree mod
   log operation - it increments fs_info->tree_mod_seq and sets
   fs_info->tree_mod_seq as the sequence number of the given tree mod log
   operation.

   This means that for the tm_list created at tree_mod_log_insert_root(),
   the element corresponding to slot 0 of eb X has the highest sequence
   number (1 + N), and the element corresponding to the last slot has the
   lowest sequence number (2);

7) Then, after inserting tm_list's elements into the tree mod log rbtree,
   the MOD_LOG_ROOT_REPLACE element is inserted, which gets the highest
   sequence number, which is N + 2;

8) Back to ctree.c:balance_level(), we free eb X by calling
   btrfs_free_tree_block() on it. Because eb X was created in the current
   transaction, has no other references and writeback did not happen for
   it, we add it back to the free space cache/tree;

9) Later some other task T allocates the metadata extent from eb X, since
   it is marked as free space in the space cache/tree, and uses it as a
   node for some other btree;

10) The tree mod log user task calls btrfs_search_old_slot(), which calls
    get_old_root(), and finally that calls __tree_mod_log_oldest_root()
    with time_seq == 1 and eb_root == eb Y;

11) First iteration of the while loop finds the tree mod log element with
    sequence number N + 2, for the logical address of eb Y and of type
    MOD_LOG_ROOT_REPLACE;

12) Because the operation type is MOD_LOG_ROOT_REPLACE, we don't break out
    of the loop, and set root_logical to point to tm->old_root.logical
    which corresponds to the logical address of eb X;

13) On the next iteration of the while loop, the call to
    tree_mod_log_search_oldest() returns the smallest tree mod log element
    for the logical address of eb X, which has a sequence number of 2, an
    operation type of MOD_LOG_KEY_REMOVE_WHILE_FREEING and corresponds to
    the old slot N - 1 of eb X (eb X had N items in it before being freed);

14) We then break out of the while loop and return the tree mod log operation
    of type MOD_LOG_ROOT_REPLACE (eb Y), and not the one for slot N - 1 of
    eb X, to get_old_root();

15) At get_old_root(), we process the MOD_LOG_ROOT_REPLACE operation
    and set "logical" to the logical address of eb X, which was the old
    root. We then call tree_mod_log_search() passing it the logical
    address of eb X and time_seq == 1;

16) Then before calling tree_mod_log_search(), task T adds a key to eb X,
    which results in adding a tree mod log operation of type
    MOD_LOG_KEY_ADD to the tree mod log - this is done at
    ctree.c:insert_ptr() - but after adding the tree mod log operation
    and before updating the number of items in eb X from 0 to 1...

17) The task at get_old_root() calls tree_mod_log_search() and gets the
    tree mod log operation of type MOD_LOG_KEY_ADD just added by task T.
    Then it enters the following if branch:

    if (old_root && tm && tm->op != MOD_LOG_KEY_REMOVE_WHILE_FREEING) {
       (...)
    } (...)

    Calls read_tree_block() for eb X, which gets a reference on eb X but
    does not lock it - task T has it locked.
    Then it clones eb X while it has nritems set to 0 in its header, before
    task T sets nritems to 1 in eb X's header. From hereupon we use the
    clone of eb X which no other task has access to;

18) Then we call __tree_mod_log_rewind(), passing it the MOD_LOG_KEY_ADD
    mod log operation we just got from tree_mod_log_search() in the
    previous step and the cloned version of eb X;

19) At __tree_mod_log_rewind(), we set the local variable "n" to the number
    of items set in eb X's clone, which is 0. Then we enter the while loop,
    and in its first iteration we process the MOD_LOG_KEY_ADD operation,
    which just decrements "n" from 0 to (u32)-1, since "n" is declared with
    a type of u32. At the end of this iteration we call rb_next() to find the
    next tree mod log operation for eb X, that gives us the mod log operation
    of type MOD_LOG_KEY_REMOVE_WHILE_FREEING, for slot 0, with a sequence
    number of N + 1 (steps 3 to 6);

20) Then we go back to the top of the while loop and trigger the following
    BUG_ON():

        (...)
        switch (tm->op) {
        case MOD_LOG_KEY_REMOVE_WHILE_FREEING:
                 BUG_ON(tm->slot < n);
                 fallthrough;
        (...)

    Because "n" has a value of (u32)-1 (4294967295) and tm->slot is 0.

Fix this by taking a read lock on the extent buffer before cloning it at
ctree.c:get_old_root(). This should be done regardless of the extent
buffer having been freed and reused, as a concurrent task might be
modifying it (while holding a write lock on it).

Reported-by: Zygo Blaxell <ce3g8jdj@umail.furryterror.org>
Link: https://lore.kernel.org/linux-btrfs/20210227155037.GN28049@hungrycats.org/
Fixes: 834328a8493079 ("Btrfs: tree mod log's old roots could still be part of the tree")
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoARM: 9044/1: vfp: use undef hook for VFP support detection
Ard Biesheuvel [Tue, 16 Mar 2021 16:59:18 +0000 (09:59 -0700)]
ARM: 9044/1: vfp: use undef hook for VFP support detection

commit 3cce9d44321e460e7c88cdec4e4537a6e9ad7c0d upstream.

Commit f77ac2e378be9dd6 ("ARM: 9030/1: entry: omit FP emulation for UND
exceptions taken in kernel mode") failed to take into account that there
is in fact a case where we relied on this code path: during boot, the
VFP detection code issues a read of FPSID, which will trigger an undef
exception on cores that lack VFP support.

So let's reinstate this logic using an undef hook which is registered
only for the duration of the initcall to vpf_init(), and which sets
VFP_arch to a non-zero value - as before - if no VFP support is present.

Fixes: f77ac2e378be9dd6 ("ARM: 9030/1: entry: omit FP emulation for UND ...")
Reported-by: "kernelci.org bot" <bot@kernelci.org>
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoARM: 9030/1: entry: omit FP emulation for UND exceptions taken in kernel mode
Ard Biesheuvel [Mon, 15 Mar 2021 23:19:52 +0000 (16:19 -0700)]
ARM: 9030/1: entry: omit FP emulation for UND exceptions taken in kernel mode

commit f77ac2e378be9dd61eb88728f0840642f045d9d1 upstream.

There are a couple of problems with the exception entry code that deals
with FP exceptions (which are reported as UND exceptions) when building
the kernel in Thumb2 mode:
- the conditional branch to vfp_kmode_exception in vfp_support_entry()
  may be out of range for its target, depending on how the linker decides
  to arrange the sections;
- when the UND exception is taken in kernel mode, the emulation handling
  logic is entered via the 'call_fpe' label, which means we end up using
  the wrong value/mask pairs to match and detect the NEON opcodes.

Since UND exceptions in kernel mode are unlikely to occur on a hot path
(as opposed to the user mode version which is invoked for VFP support
code and lazy restore), we can use the existing undef hook machinery for
any kernel mode instruction emulation that is needed, including calling
the existing vfp_kmode_exception() routine for unexpected cases. So drop
the call to call_fpe, and instead, install an undef hook that will get
called for NEON and VFP instructions that trigger an UND exception in
kernel mode.

While at it, make sure that the PC correction is accurate for the
execution mode where the exception was taken, by checking the PSR
Thumb bit.

[nd: fix conflict in arch/arm/vfp/vfphw.S due to missing
     commit 2cbd1cc3dcd3 ("ARM: 8991/1: use VFP assembler mnemonics if
     available")]

Fixes: eff8728fe698 ("vmlinux.lds.h: Add PGO and AutoFDO input sections")
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Russell King <rmk+kernel@armlinux.org.uk>
Signed-off-by: Nick Desaulniers <ndesaulniers@google.com>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Cc: Dmitry Osipenko <digetx@gmail.com>
Cc: Kees Cook <keescook@chromium.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agos390/vtime: fix increased steal time accounting
Gerald Schaefer [Wed, 10 Mar 2021 13:23:37 +0000 (14:23 +0100)]
s390/vtime: fix increased steal time accounting

commit d54cb7d54877d529bc1e0e1f47a3dd082f73add3 upstream.

Commit 152e9b8676c6e ("s390/vtime: steal time exponential moving average")
inadvertently changed the input value for account_steal_time() from
"cputime_to_nsecs(steal)" to just "steal", resulting in broken increased
steal time accounting.

Fix this by changing it back to "cputime_to_nsecs(steal)".

Fixes: 152e9b8676c6e ("s390/vtime: steal time exponential moving average")
Cc: <stable@vger.kernel.org> # 5.1
Reported-by: Sabine Forkel <sabine.forkel@de.ibm.com>
Reviewed-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Gerald Schaefer <gerald.schaefer@linux.ibm.com>
Signed-off-by: Heiko Carstens <hca@linux.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoRevert "PM: runtime: Update device status before letting suppliers suspend"
Rafael J. Wysocki [Fri, 19 Mar 2021 14:47:25 +0000 (15:47 +0100)]
Revert "PM: runtime: Update device status before letting suppliers suspend"

commit 0cab893f409c53634d0d818fa414641cbcdb0dab upstream.

Revert commit 44cc89f76464 ("PM: runtime: Update device status
before letting suppliers suspend") that introduced a race condition
into __rpm_callback() which allowed a concurrent rpm_resume() to
run and resume the device prematurely after its status had been
changed to RPM_SUSPENDED by __rpm_callback().

Fixes: 44cc89f76464 ("PM: runtime: Update device status before letting suppliers suspend")
Link: https://lore.kernel.org/linux-pm/24dfb6fc-5d54-6ee2-9195-26428b7ecf8a@intel.com/
Reported-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: 4.10+ <stable@vger.kernel.org> # 4.10+
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@intel.com>
Reviewed-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: hda/realtek: Apply headset-mic quirks for Xiaomi Redmibook Air
Xiaoliang Yu [Fri, 12 Mar 2021 23:54:53 +0000 (07:54 +0800)]
ALSA: hda/realtek: Apply headset-mic quirks for Xiaomi Redmibook Air

commit e1c86210fe27428399643861b81b080eccd79f87 upstream.

There is another fix for headset-mic problem on Redmibook (1d72:1602),
it also works on Redmibook Air (1d72:1947), which has the same issue.

Signed-off-by: Xiaoliang Yu <yxl_22@outlook.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/TYBP286MB02856DC016849DEA0F9B6A37EE6F9@TYBP286MB0285.JPNP286.PROD.OUTLOOK.COM
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: hda: generic: Fix the micmute led init state
Hui Wang [Fri, 12 Mar 2021 04:14:08 +0000 (12:14 +0800)]
ALSA: hda: generic: Fix the micmute led init state

commit 2bf44e0ee95f39cc54ea1b942f0a027e0181ca4e upstream.

Recently we found the micmute led init state is not correct after
freshly installing the ubuntu linux on a Lenovo AIO machine. The
internal mic is not muted, but the micmute led is on and led mode is
'follow mute'. If we mute internal mic, the led is keeping on, then
unmute the internal mic, the led is off. And from then on, the
micmute led will work correctly.

So the micmute led init state is not correct. The led is controlled
by codec gpio (ALC233_FIXUP_LENOVO_LINE2_MIC_HOTKEY), in the
patch_realtek, the gpio data is set to 0x4 initially and the led is
on with this data. In the hda_generic, the led_value is set to
0 initially, suppose users set the 'capture switch' to on from
user space and the micmute led should change to be off with this
operation, but the check "if (val == spec->micmute_led.led_value)" in
the call_micmute_led_update() will skip the led setting.

To guarantee the led state will be set by the 1st time of changing
"Capture Switch", set -1 to the init led_value.

Cc: <stable@vger.kernel.org>
Signed-off-by: Hui Wang <hui.wang@canonical.com>
Link: https://lore.kernel.org/r/20210312041408.3776-1-hui.wang@canonical.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: hda/realtek: apply pin quirk for XiaomiNotebook Pro
Xiaoliang Yu [Mon, 15 Mar 2021 16:49:00 +0000 (00:49 +0800)]
ALSA: hda/realtek: apply pin quirk for XiaomiNotebook Pro

commit b95bc12e0412d14d5fc764f0b82631c7bcaf1959 upstream.

Built-in microphone and combojack on Xiaomi Notebook Pro (1d72:1701) needs
to be fixed, the existing quirk for Dell works well on that machine.

Signed-off-by: Xiaoliang Yu <yxl_22@outlook.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/OS0P286MB02749B9E13920E6899902CD8EE6C9@OS0P286MB0274.JPNP286.PROD.OUTLOOK.COM
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoALSA: dice: fix null pointer dereference when node is disconnected
Takashi Sakamoto [Fri, 12 Mar 2021 09:34:07 +0000 (18:34 +0900)]
ALSA: dice: fix null pointer dereference when node is disconnected

commit dd7b836d6bc935df95c826f69ff4d051f5561604 upstream.

When node is removed from IEEE 1394 bus, any transaction fails to the node.
In the case, ALSA dice driver doesn't stop isochronous contexts even if
they are running. As a result, null pointer dereference occurs in callback
from the running context.

This commit fixes the bug to release isochronous contexts always.

Cc: <stable@vger.kernel.org> # v5.4 or later
Fixes: e9f21129b8d8 ("ALSA: dice: support AMDTP domain")
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Link: https://lore.kernel.org/r/20210312093407.23437-1-o-takashi@sakamocchi.jp
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoASoC: ak5558: Add MODULE_DEVICE_TABLE
Shengjiu Wang [Wed, 24 Feb 2021 06:57:52 +0000 (14:57 +0800)]
ASoC: ak5558: Add MODULE_DEVICE_TABLE

commit 80cffd2468ddb850e678f17841fc356930b2304a upstream.

Add missed MODULE_DEVICE_TABLE for the driver can be loaded
automatically at boot.

Fixes: 920884777480 ("ASoC: ak5558: Add support for AK5558 ADC driver")
Cc: <stable@vger.kernel.org>
Signed-off-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Link: https://lore.kernel.org/r/1614149872-25510-2-git-send-email-shengjiu.wang@nxp.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoASoC: ak4458: Add MODULE_DEVICE_TABLE
Shengjiu Wang [Wed, 24 Feb 2021 06:57:51 +0000 (14:57 +0800)]
ASoC: ak4458: Add MODULE_DEVICE_TABLE

commit 4ec5b96775a88dd9b1c3ba1d23c43c478cab95a2 upstream.

Add missed MODULE_DEVICE_TABLE for the driver can be loaded
automatically at boot.

Fixes: 08660086eff9 ("ASoC: ak4458: Add support for AK4458 DAC driver")
Cc: <stable@vger.kernel.org>
Signed-off-by: Shengjiu Wang <shengjiu.wang@nxp.com>
Link: https://lore.kernel.org/r/1614149872-25510-1-git-send-email-shengjiu.wang@nxp.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoLinux 5.4.107 v5.4.107
Greg Kroah-Hartman [Sat, 20 Mar 2021 09:39:47 +0000 (10:39 +0100)]
Linux 5.4.107

Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Jason Self <jason@bluehome.net>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Link: https://lore.kernel.org/r/20210319121745.449875976@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonet: dsa: b53: Support setting learning on port
Florian Fainelli [Mon, 22 Feb 2021 22:30:10 +0000 (14:30 -0800)]
net: dsa: b53: Support setting learning on port

commit f9b3827ee66cfcf297d0acd6ecf33653a5f297ef upstream.

Add support for being able to set the learning attribute on port, and
make sure that the standalone ports start up with learning disabled.

We can remove the code in bcm_sf2 that configured the ports learning
attribute because we want the standalone ports to have learning disabled
by default and port 7 cannot be bridged, so its learning attribute will
not change past its initial configuration.

Signed-off-by: Florian Fainelli <f.fainelli@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonet: dsa: tag_mtk: fix 802.1ad VLAN egress
DENG Qingfang [Mon, 1 Mar 2021 16:01:59 +0000 (00:01 +0800)]
net: dsa: tag_mtk: fix 802.1ad VLAN egress

commit 9200f515c41f4cbaeffd8fdd1d8b6373a18b1b67 upstream.

A different TPID bit is used for 802.1ad VLAN frames.

Reported-by: Ilario Gelmetti <iochesonome@gmail.com>
Fixes: f0af34317f4b ("net: dsa: mediatek: combine MediaTek tag with VLAN tag")
Signed-off-by: DENG Qingfang <dqfext@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agocrypto: x86/aes-ni-xts - use direct calls to and 4-way stride
Ard Biesheuvel [Thu, 31 Dec 2020 16:41:54 +0000 (17:41 +0100)]
crypto: x86/aes-ni-xts - use direct calls to and 4-way stride

commit 86ad60a65f29dd862a11c22bb4b5be28d6c5cef1 upstream.

The XTS asm helper arrangement is a bit odd: the 8-way stride helper
consists of back-to-back calls to the 4-way core transforms, which
are called indirectly, based on a boolean that indicates whether we
are performing encryption or decryption.

Given how costly indirect calls are on x86, let's switch to direct
calls, and given how the 8-way stride doesn't really add anything
substantial, use a 4-way stride instead, and make the asm core
routine deal with any multiple of 4 blocks. Since 512 byte sectors
or 4 KB blocks are the typical quantities XTS operates on, increase
the stride exported to the glue helper to 512 bytes as well.

As a result, the number of indirect calls is reduced from 3 per 64 bytes
of in/output to 1 per 512 bytes of in/output, which produces a 65% speedup
when operating on 1 KB blocks (measured on a Intel(R) Core(TM) i7-8650U CPU)

Fixes: 9697fa39efd3f ("x86/retpoline/crypto: Convert crypto assembler indirect jumps")
Tested-by: Eric Biggers <ebiggers@google.com> # x86_64
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
[ardb: rebase onto stable/linux-5.4.y]
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agocrypto: aesni - Use TEST %reg,%reg instead of CMP $0,%reg
Uros Bizjak [Fri, 27 Nov 2020 09:44:52 +0000 (10:44 +0100)]
crypto: aesni - Use TEST %reg,%reg instead of CMP $0,%reg

commit 032d049ea0f45b45c21f3f02b542aa18bc6b6428 upstream.

CMP $0,%reg can't set overflow flag, so we can use shorter TEST %reg,%reg
instruction when only zero and sign flags are checked (E,L,LE,G,GE conditions).

Signed-off-by: Uros Bizjak <ubizjak@gmail.com>
Cc: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Borislav Petkov <bp@alien8.de>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Ard Biesheuvel <ardb@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agocrypto: x86 - Regularize glue function prototypes
Kees Cook [Wed, 27 Nov 2019 06:08:02 +0000 (22:08 -0800)]
crypto: x86 - Regularize glue function prototypes

commit 9c1e8836edbbaf3656bc07437b59c04be034ac4e upstream.

The crypto glue performed function prototype casting via macros to make
indirect calls to assembly routines. Instead of performing casts at the
call sites (which trips Control Flow Integrity prototype checking), switch
each prototype to a common standard set of arguments which allows the
removal of the existing macros. In order to keep pointer math unchanged,
internal casting between u128 pointers and u8 pointers is added.

Co-developed-by: João Moreira <joao.moreira@intel.com>
Signed-off-by: João Moreira <joao.moreira@intel.com>
Signed-off-by: Kees Cook <keescook@chromium.org>
Reviewed-by: Eric Biggers <ebiggers@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Cc: Ard Biesheuvel <ardb@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agofuse: fix live lock in fuse_iget()
Amir Goldstein [Thu, 4 Mar 2021 09:09:12 +0000 (11:09 +0200)]
fuse: fix live lock in fuse_iget()

commit 775c5033a0d164622d9d10dd0f0a5531639ed3ed upstream.

Commit 5d069dbe8aaf ("fuse: fix bad inode") replaced make_bad_inode()
in fuse_iget() with a private implementation fuse_make_bad().

The private implementation fails to remove the bad inode from inode
cache, so the retry loop with iget5_locked() finds the same bad inode
and marks it bad forever.

kmsg snip:

[ ] rcu: INFO: rcu_sched self-detected stall on CPU
...
[ ]  ? bit_wait_io+0x50/0x50
[ ]  ? fuse_init_file_inode+0x70/0x70
[ ]  ? find_inode.isra.32+0x60/0xb0
[ ]  ? fuse_init_file_inode+0x70/0x70
[ ]  ilookup5_nowait+0x65/0x90
[ ]  ? fuse_init_file_inode+0x70/0x70
[ ]  ilookup5.part.36+0x2e/0x80
[ ]  ? fuse_init_file_inode+0x70/0x70
[ ]  ? fuse_inode_eq+0x20/0x20
[ ]  iget5_locked+0x21/0x80
[ ]  ? fuse_inode_eq+0x20/0x20
[ ]  fuse_iget+0x96/0x1b0

Fixes: 5d069dbe8aaf ("fuse: fix bad inode")
Cc: stable@vger.kernel.org # 5.10+
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agodrm/i915/gvt: Fix vfio_edid issue for BXT/APL
Colin Xu [Wed, 17 Mar 2021 02:55:04 +0000 (10:55 +0800)]
drm/i915/gvt: Fix vfio_edid issue for BXT/APL

commit 4ceb06e7c336f4a8d3f3b6ac9a4fea2e9c97dc07 upstream

BXT/APL has different isr/irr/hpd regs compared with other GEN9. If not
setting these regs bits correctly according to the emulated monitor
(currently a DP on PORT_B), although gvt still triggers a virtual HPD
event, the guest driver won't detect a valid HPD pulse thus no full
display detection will be executed to read the updated EDID.

With this patch, the vfio_edid is enabled again on BXT/APL, which is
previously disabled.

Fixes: 642403e3599e ("drm/i915/gvt: Temporarily disable vfio_edid for BXT/APL")
Signed-off-by: Colin Xu <colin.xu@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20201201060329.142375-1-colin.xu@intel.com
Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
(cherry picked from commit 4ceb06e7c336f4a8d3f3b6ac9a4fea2e9c97dc07)
Signed-off-by: Colin Xu <colin.xu@intel.com>
Cc: <stable@vger.kernel.org> # 5.4.y
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agodrm/i915/gvt: Fix port number for BDW on EDID region setup
Colin Xu [Wed, 17 Mar 2021 02:55:03 +0000 (10:55 +0800)]
drm/i915/gvt: Fix port number for BDW on EDID region setup

From: Zhenyu Wang <zhenyuw@linux.intel.com>

commit 28284943ac94014767ecc2f7b3c5747c4a5617a0 upstream

Current BDW virtual display port is initialized as PORT_B, so need
to use same port for VFIO EDID region, otherwise invalid EDID blob
pointer is assigned which caused kernel null pointer reference. We
might evaluate actual display hotplug for BDW to make this function
work as expected, anyway this is always required to be fixed first.

Reported-by: Alejandro Sior <aho@sior.be>
Cc: Alejandro Sior <aho@sior.be>
Fixes: 0178f4ce3c3b ("drm/i915/gvt: Enable vfio edid for all GVT supported platform")
Reviewed-by: Hang Yuan <hang.yuan@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20200914030302.2775505-1-zhenyuw@linux.intel.com
(cherry picked from commit 28284943ac94014767ecc2f7b3c5747c4a5617a0)
Signed-off-by: Colin Xu <colin.xu@intel.com>
Cc: <stable@vger.kernel.org> # 5.4.y
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agodrm/i915/gvt: Fix virtual display setup for BXT/APL
Colin Xu [Wed, 17 Mar 2021 02:55:02 +0000 (10:55 +0800)]
drm/i915/gvt: Fix virtual display setup for BXT/APL

commit a5a8ef937cfa79167f4b2a5602092b8d14fd6b9a upstream

Program display related vregs to proper value at initialization, setup
virtual monitor and hotplug.

vGPU virtual display vregs inherit the value from pregs. The virtual DP
monitor is always setup on PORT_B for BXT/APL. However the host may
connect monitor on other PORT or without any monitor connected. Without
properly setup PIPE/DDI/PLL related vregs, guest driver may not setup
the virutal display as expected, and the guest desktop may not be
created.
Since only one virtual display is supported, enable PIPE_A only. And
enable transcoder/DDI/PLL based on which port is setup for BXT/APL.

V2:
Revise commit message.

V3:
set_edid should on PORT_B for BXT.
Inject hpd event for BXT.

V4:
Temporarily disable vfio edid on BXT/APL until issue fixed.

V5:
Rebase to use new HPD define GEN8_DE_PORT_HOTPLUG for BXT.
Put vfio edid disabling on BXT/APL to a separate patch.

Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Colin Xu <colin.xu@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20201109073922.757759-1-colin.xu@intel.com
(cherry picked from commit a5a8ef937cfa79167f4b2a5602092b8d14fd6b9a)
Signed-off-by: Colin Xu <colin.xu@intel.com>
Cc: <stable@vger.kernel.org> # 5.4.y
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agodrm/i915/gvt: Fix mmio handler break on BXT/APL.
Colin Xu [Wed, 17 Mar 2021 02:55:01 +0000 (10:55 +0800)]
drm/i915/gvt: Fix mmio handler break on BXT/APL.

commit 92010a97098c4c9fd777408cc98064d26b32695b upstream

- Remove dup mmio handler for BXT/APL. Otherwise mmio handler will fail
  to init.
- Add engine GPR with F_CMD_ACCESS since BXT/APL will load them via
  LRI. Otherwise, guest will enter failsafe mode.

V2:
Use RCS/BCS GPR macros instead of offset.
Revise commit message.

V3:
Use GEN8_RING_CS_GPR macros on ring base.

Reviewed-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Colin Xu <colin.xu@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20201016052913.209248-1-colin.xu@intel.com
(cherry picked from commit 92010a97098c4c9fd777408cc98064d26b32695b)
Signed-off-by: Colin Xu <colin.xu@intel.com>
Cc: <stable@vger.kernel.org> # 5.4.y
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agodrm/i915/gvt: Set SNOOP for PAT3 on BXT/APL to workaround GPU BB hang
Colin Xu [Wed, 17 Mar 2021 02:55:00 +0000 (10:55 +0800)]
drm/i915/gvt: Set SNOOP for PAT3 on BXT/APL to workaround GPU BB hang

commit 8fe105679765700378eb328495fcfe1566cdbbd0 upstream

If guest fills non-priv bb on ApolloLake/Broxton as Mesa i965 does in:
717e7539124d (i965: Use a WC map and memcpy for the batch instead of pw-)
Due to the missing flush of bb filled by VM vCPU, host GPU hangs on
executing these MI_BATCH_BUFFER.

Temporarily workaround this by setting SNOOP bit for PAT3 used by PPGTT
PML4 PTE: PAT(0) PCD(1) PWT(1).

The performance is still expected to be low, will need further improvement.

Acked-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Signed-off-by: Colin Xu <colin.xu@intel.com>
Signed-off-by: Zhenyu Wang <zhenyuw@linux.intel.com>
Link: http://patchwork.freedesktop.org/patch/msgid/20201012045231.226748-1-colin.xu@intel.com
(cherry picked from commit 8fe105679765700378eb328495fcfe1566cdbbd0)
Signed-off-by: Colin Xu <colin.xu@intel.com>
Cc: <stable@vger.kernel.org> # 5.4.y
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobtrfs: scrub: Don't check free space before marking a block group RO
Qu Wenruo [Fri, 15 Nov 2019 02:09:00 +0000 (10:09 +0800)]
btrfs: scrub: Don't check free space before marking a block group RO

commit b12de52896c0e8213f70e3a168fde9e6eee95909 upstream.

[BUG]
When running btrfs/072 with only one online CPU, it has a pretty high
chance to fail:

#  btrfs/072 12s ... _check_dmesg: something found in dmesg (see xfstests-dev/results//btrfs/072.dmesg)
#  - output mismatch (see xfstests-dev/results//btrfs/072.out.bad)
#      --- tests/btrfs/072.out     2019-10-22 15:18:14.008965340 +0800
#      +++ /xfstests-dev/results//btrfs/072.out.bad      2019-11-14 15:56:45.877152240 +0800
#      @@ -1,2 +1,3 @@
#       QA output created by 072
#       Silence is golden
#      +Scrub find errors in "-m dup -d single" test
#      ...

And with the following call trace:

  BTRFS info (device dm-5): scrub: started on devid 1
  ------------[ cut here ]------------
  BTRFS: Transaction aborted (error -27)
  WARNING: CPU: 0 PID: 55087 at fs/btrfs/block-group.c:1890 btrfs_create_pending_block_groups+0x3e6/0x470 [btrfs]
  CPU: 0 PID: 55087 Comm: btrfs Tainted: G        W  O      5.4.0-rc1-custom+ #13
  Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS 0.0.0 02/06/2015
  RIP: 0010:btrfs_create_pending_block_groups+0x3e6/0x470 [btrfs]
  Call Trace:
   __btrfs_end_transaction+0xdb/0x310 [btrfs]
   btrfs_end_transaction+0x10/0x20 [btrfs]
   btrfs_inc_block_group_ro+0x1c9/0x210 [btrfs]
   scrub_enumerate_chunks+0x264/0x940 [btrfs]
   btrfs_scrub_dev+0x45c/0x8f0 [btrfs]
   btrfs_ioctl+0x31a1/0x3fb0 [btrfs]
   do_vfs_ioctl+0x636/0xaa0
   ksys_ioctl+0x67/0x90
   __x64_sys_ioctl+0x43/0x50
   do_syscall_64+0x79/0xe0
   entry_SYSCALL_64_after_hwframe+0x49/0xbe
  ---[ end trace 166c865cec7688e7 ]---

[CAUSE]
The error number -27 is -EFBIG, returned from the following call chain:
btrfs_end_transaction()
|- __btrfs_end_transaction()
   |- btrfs_create_pending_block_groups()
      |- btrfs_finish_chunk_alloc()
         |- btrfs_add_system_chunk()

This happens because we have used up all space of
btrfs_super_block::sys_chunk_array.

The root cause is, we have the following bad loop of creating tons of
system chunks:

1. The only SYSTEM chunk is being scrubbed
   It's very common to have only one SYSTEM chunk.
2. New SYSTEM bg will be allocated
   As btrfs_inc_block_group_ro() will check if we have enough space
   after marking current bg RO. If not, then allocate a new chunk.
3. New SYSTEM bg is still empty, will be reclaimed
   During the reclaim, we will mark it RO again.
4. That newly allocated empty SYSTEM bg get scrubbed
   We go back to step 2, as the bg is already mark RO but still not
   cleaned up yet.

If the cleaner kthread doesn't get executed fast enough (e.g. only one
CPU), then we will get more and more empty SYSTEM chunks, using up all
the space of btrfs_super_block::sys_chunk_array.

[FIX]
Since scrub/dev-replace doesn't always need to allocate new extent,
especially chunk tree extent, so we don't really need to do chunk
pre-allocation.

To break above spiral, here we introduce a new parameter to
btrfs_inc_block_group(), @do_chunk_alloc, which indicates whether we
need extra chunk pre-allocation.

For relocation, we pass @do_chunk_alloc=true, while for scrub, we pass
@do_chunk_alloc=false.
This should keep unnecessary empty chunks from popping up for scrub.

Also, since there are two parameters for btrfs_inc_block_group_ro(),
add more comment for it.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobpf, selftests: Fix up some test_verifier cases for unprivileged
Piotr Krysiuk [Tue, 16 Mar 2021 10:44:42 +0000 (11:44 +0100)]
bpf, selftests: Fix up some test_verifier cases for unprivileged

commit 0a13e3537ea67452d549a6a80da3776d6b7dedb3 upstream.

Fix up test_verifier error messages for the case where the original error
message changed, or for the case where pointer alu errors differ between
privileged and unprivileged tests. Also, add alternative tests for keeping
coverage of the original verifier rejection error message (fp alu), and
newly reject map_ptr += rX where rX == 0 given we now forbid alu on these
types for unprivileged. All test_verifier cases pass after the change. The
test case fixups were kept separate to ease backporting of core changes.

Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobpf: Add sanity check for upper ptr_limit
Piotr Krysiuk [Tue, 16 Mar 2021 08:47:02 +0000 (09:47 +0100)]
bpf: Add sanity check for upper ptr_limit

commit 1b1597e64e1a610c7a96710fc4717158e98a08b3 upstream.

Given we know the max possible value of ptr_limit at the time of retrieving
the latter, add basic assertions, so that the verifier can bail out if
anything looks odd and reject the program. Nothing triggered this so far,
but it also does not hurt to have these.

Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobpf: Simplify alu_limit masking for pointer arithmetic
Piotr Krysiuk [Tue, 16 Mar 2021 07:26:25 +0000 (08:26 +0100)]
bpf: Simplify alu_limit masking for pointer arithmetic

commit b5871dca250cd391885218b99cc015aca1a51aea upstream.

Instead of having the mov32 with aux->alu_limit - 1 immediate, move this
operation to retrieve_ptr_limit() instead to simplify the logic and to
allow for subsequent sanity boundary checks inside retrieve_ptr_limit().
This avoids in future that at the time of the verifier masking rewrite
we'd run into an underflow which would not sign extend due to the nature
of mov32 instruction.

Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobpf: Fix off-by-one for area size in creating mask to left
Piotr Krysiuk [Tue, 16 Mar 2021 07:20:16 +0000 (08:20 +0100)]
bpf: Fix off-by-one for area size in creating mask to left

commit 10d2bb2e6b1d8c4576c56a748f697dbeb8388899 upstream.

retrieve_ptr_limit() computes the ptr_limit for registers with stack and
map_value type. ptr_limit is the size of the memory area that is still
valid / in-bounds from the point of the current position and direction
of the operation (add / sub). This size will later be used for masking
the operation such that attempting out-of-bounds access in the speculative
domain is redirected to remain within the bounds of the current map value.

When masking to the right the size is correct, however, when masking to
the left, the size is off-by-one which would lead to an incorrect mask
and thus incorrect arithmetic operation in the non-speculative domain.
Piotr found that if the resulting alu_limit value is zero, then the
BPF_MOV32_IMM() from the fixup_bpf_calls() rewrite will end up loading
0xffffffff into AX instead of sign-extending to the full 64 bit range,
and as a result, this allows abuse for executing speculatively out-of-
bounds loads against 4GB window of address space and thus extracting the
contents of kernel memory via side-channel.

Fixes: 979d63d50c0c ("bpf: prevent out of bounds speculation on pointer arithmetic")
Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobpf: Prohibit alu ops for pointer types not defining ptr_limit
Piotr Krysiuk [Tue, 16 Mar 2021 08:47:02 +0000 (09:47 +0100)]
bpf: Prohibit alu ops for pointer types not defining ptr_limit

commit f232326f6966cf2a1d1db7bc917a4ce5f9f55f76 upstream.

The purpose of this patch is to streamline error propagation and in particular
to propagate retrieve_ptr_limit() errors for pointer types that are not defining
a ptr_limit such that register-based alu ops against these types can be rejected.

The main rationale is that a gap has been identified by Piotr in the existing
protection against speculatively out-of-bounds loads, for example, in case of
ctx pointers, unprivileged programs can still perform pointer arithmetic. This
can be abused to execute speculatively out-of-bounds loads without restrictions
and thus extract contents of kernel memory.

Fix this by rejecting unprivileged programs that attempt any pointer arithmetic
on unprotected pointer types. The two affected ones are pointer to ctx as well
as pointer to map. Field access to a modified ctx' pointer is rejected at a
later point in time in the verifier, and 7c6967326267 ("bpf: Permit map_ptr
arithmetic with opcode add and offset 0") only relevant for root-only use cases.
Risk of unprivileged program breakage is considered very low.

Fixes: 7c6967326267 ("bpf: Permit map_ptr arithmetic with opcode add and offset 0")
Fixes: b2157399cc98 ("bpf: prevent out-of-bounds speculation")
Signed-off-by: Piotr Krysiuk <piotras@gmail.com>
Co-developed-by: Daniel Borkmann <daniel@iogearbox.net>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Alexei Starovoitov <ast@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: arm64: nvhe: Save the SPE context early
Suzuki K Poulose [Tue, 16 Mar 2021 18:33:53 +0000 (18:33 +0000)]
KVM: arm64: nvhe: Save the SPE context early

commit b96b0c5de685df82019e16826a282d53d86d112c upstream

The nVHE KVM hyp drains and disables the SPE buffer, before
entering the guest, as the EL1&0 translation regime
is going to be loaded with that of the guest.

But this operation is performed way too late, because :
 - The owning translation regime of the SPE buffer
   is transferred to EL2. (MDCR_EL2_E2PB == 0)
 - The guest Stage1 is loaded.

Thus the flush could use the host EL1 virtual address,
but use the EL2 translations instead of host EL1, for writing
out any cached data.

Fix this by moving the SPE buffer handling early enough.
The restore path is doing the right thing.

Cc: stable@vger.kernel.org # v5.4-
Cc: Christoffer Dall <christoffer.dall@arm.com>
Cc: Marc Zyngier <maz@kernel.org>
Cc: Will Deacon <will@kernel.org>
Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Alexandru Elisei <alexandru.elisei@arm.com>
Signed-off-by: Suzuki K Poulose <suzuki.poulose@arm.com>
Acked-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoLinux 5.4.106 v5.4.106
Greg Kroah-Hartman [Wed, 17 Mar 2021 16:03:58 +0000 (17:03 +0100)]
Linux 5.4.106

Tested-by: Jon Hunter <jonathanh@nvidia.com>
Tested-by: Florian Fainelli <f.fainelli@gmail.com>
Tested-by: Jason Self <jason@bluehome.net>
Tested-by: Linux Kernel Functional Testing <lkft@linaro.org>
Tested-by: Guenter Roeck <linux@roeck-us.net>
Tested-by: Hulk Robot <hulkrobot@huawei.com>
Tested-by: Ross Schmidt <ross.schm.dev@gmail.com>
Link: https://lore.kernel.org/r/20210315135550.333963635@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxen/events: avoid handling the same event on two cpus at the same time
Juergen Gross [Mon, 15 Mar 2021 08:22:38 +0000 (09:22 +0100)]
xen/events: avoid handling the same event on two cpus at the same time

commit b6622798bc50b625a1e62f82c7190df40c1f5b21 upstream.

When changing the cpu affinity of an event it can happen today that
(with some unlucky timing) the same event will be handled on the old
and the new cpu at the same time.

Avoid that by adding an "event active" flag to the per-event data and
call the handler only if this flag isn't set.

Cc: stable@vger.kernel.org
Reported-by: Julien Grall <julien@xen.org>
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Julien Grall <jgrall@amazon.com>
Link: https://lore.kernel.org/r/20210306161833.4552-4-jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxen/events: don't unmask an event channel when an eoi is pending
Juergen Gross [Mon, 15 Mar 2021 08:22:37 +0000 (09:22 +0100)]
xen/events: don't unmask an event channel when an eoi is pending

commit 25da4618af240fbec6112401498301a6f2bc9702 upstream.

An event channel should be kept masked when an eoi is pending for it.
When being migrated to another cpu it might be unmasked, though.

In order to avoid this keep three different flags for each event channel
to be able to distinguish "normal" masking/unmasking from eoi related
masking/unmasking and temporary masking. The event channel should only
be able to generate an interrupt if all flags are cleared.

Cc: stable@vger.kernel.org
Fixes: 54c9de89895e ("xen/events: add a new "late EOI" evtchn framework")
Reported-by: Julien Grall <julien@xen.org>
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Julien Grall <jgrall@amazon.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Tested-by: Ross Lagerwall <ross.lagerwall@citrix.com>
Link: https://lore.kernel.org/r/20210306161833.4552-3-jgross@suse.com
[boris -- corrected Fixed tag format]

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoxen/events: reset affinity of 2-level event when tearing it down
Juergen Gross [Mon, 15 Mar 2021 08:22:37 +0000 (09:22 +0100)]
xen/events: reset affinity of 2-level event when tearing it down

commit 9e77d96b8e2724ed00380189f7b0ded61113b39f upstream.

When creating a new event channel with 2-level events the affinity
needs to be reset initially in order to avoid using an old affinity
from earlier usage of the event channel port. So when tearing an event
channel down reset all affinity bits.

The same applies to the affinity when onlining a vcpu: all old
affinity settings for this vcpu must be reset. As percpu events get
initialized before the percpu event channel hook is called,
resetting of the affinities happens after offlining a vcpu (this is
working, as initial percpu memory is zeroed out).

Cc: stable@vger.kernel.org
Reported-by: Julien Grall <julien@xen.org>
Signed-off-by: Juergen Gross <jgross@suse.com>
Reviewed-by: Julien Grall <jgrall@amazon.com>
Link: https://lore.kernel.org/r/20210306161833.4552-2-jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: arm64: Reject VM creation when the default IPA size is unsupported
Marc Zyngier [Mon, 15 Mar 2021 11:09:52 +0000 (11:09 +0000)]
KVM: arm64: Reject VM creation when the default IPA size is unsupported

Commit 7d717558dd5ef10d28866750d5c24ff892ea3778 upstream.

KVM/arm64 has forever used a 40bit default IPA space, partially
due to its 32bit heritage (where the only choice is 40bit).

However, there are implementations in the wild that have a *cough*
much smaller *cough* IPA space, which leads to a misprogramming of
VTCR_EL2, and a guest that is stuck on its first memory access
if userspace dares to ask for the default IPA setting (which most
VMMs do).

Instead, blundly reject the creation of such VM, as we can't
satisfy the requirements from userspace (with a one-off warning).
Also clarify the boot warning, and document that the VM creation
will fail when an unsupported IPA size is provided.

Although this is an ABI change, it doesn't really change much
for userspace:

- the guest couldn't run before this change, but no error was
  returned. At least userspace knows what is happening.

- a memory slot that was accepted because it did fit the default
  IPA space now doesn't even get a chance to be registered.

The other thing that is left doing is to convince userspace to
actually use the IPA space setting instead of relying on the
antiquated default.

Fixes: 233a7cb23531 ("kvm: arm64: Allow tuning the physical address size for VM")
Signed-off-by: Marc Zyngier <maz@kernel.org>
Cc: stable@vger.kernel.org
Reviewed-by: Andrew Jones <drjones@redhat.com>
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Link: https://lore.kernel.org/r/20210311100016.3830038-2-maz@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: arm64: Ensure I-cache isolation between vcpus of a same VM
Marc Zyngier [Mon, 15 Mar 2021 11:10:00 +0000 (11:10 +0000)]
KVM: arm64: Ensure I-cache isolation between vcpus of a same VM

Commit 01dc9262ff5797b675c32c0c6bc682777d23de05 upstream.

It recently became apparent that the ARMv8 architecture has interesting
rules regarding attributes being used when fetching instructions
if the MMU is off at Stage-1.

In this situation, the CPU is allowed to fetch from the PoC and
allocate into the I-cache (unless the memory is mapped with
the XN attribute at Stage-2).

If we transpose this to vcpus sharing a single physical CPU,
it is possible for a vcpu running with its MMU off to influence
another vcpu running with its MMU on, as the latter is expected to
fetch from the PoU (and self-patching code doesn't flush below that
level).

In order to solve this, reuse the vcpu-private TLB invalidation
code to apply the same policy to the I-cache, nuking it every time
the vcpu runs on a physical CPU that ran another vcpu of the same
VM in the past.

This involve renaming __kvm_tlb_flush_local_vmid() to
__kvm_flush_cpu_context(), and inserting a local i-cache invalidation
there.

Cc: stable@vger.kernel.org
Signed-off-by: Marc Zyngier <maz@kernel.org>
Acked-by: Will Deacon <will@kernel.org>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/20210303164505.68492-1-maz@kernel.org
[maz: added 32bit ARM support]
Signed-off-by: Marc Zyngier <maz@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonvme: release namespace head reference on error
Keith Busch [Thu, 9 Apr 2020 16:09:00 +0000 (09:09 -0700)]
nvme: release namespace head reference on error

commit ac262508daa88fb12c5dc53cf30bde163f9f26c9 upstream.

If a namespace identification does not match the subsystem's head for
that NSID, release the reference that was taken when the matching head
was initially found.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agonvme: unlink head after removing last namespace
Keith Busch [Thu, 9 Apr 2020 16:08:59 +0000 (09:08 -0700)]
nvme: unlink head after removing last namespace

commit d567572906d986dedb78b37f111c44eba033f3ef upstream.

The driver had been unlinking the namespace head from the subsystem's
list only after the last reference was released, and outside of the
list's subsys->lock protection.

There is no reason to track an empty head, so unlink the entry from the
subsystem's list when the last namespace using that head is removed and
with the mutex lock protecting the list update. The next namespace to
attach reusing the previous NSID will allocate a new head rather than
find the old head with mismatched identifiers.

Signed-off-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoKVM: arm64: Fix exclusive limit for IPA size
Marc Zyngier [Thu, 11 Mar 2021 10:00:16 +0000 (10:00 +0000)]
KVM: arm64: Fix exclusive limit for IPA size

commit 262b003d059c6671601a19057e9fe1a5e7f23722 upstream.

When registering a memslot, we check the size and location of that
memslot against the IPA size to ensure that we can provide guest
access to the whole of the memory.

Unfortunately, this check rejects memslot that end-up at the exact
limit of the addressing capability for a given IPA size. For example,
it refuses the creation of a 2GB memslot at 0x8000000 with a 32bit
IPA space.

Fix it by relaxing the check to accept a memslot reaching the
limit of the IPA space.

Fixes: c3058d5da222 ("arm/arm64: KVM: Ensure memslots are within KVM_PHYS_SIZE")
Reviewed-by: Eric Auger <eric.auger@redhat.com>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Cc: stable@vger.kernel.org
Reviewed-by: Andrew Jones <drjones@redhat.com>
Link: https://lore.kernel.org/r/20210311100016.3830038-3-maz@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agox86/unwind/orc: Disable KASAN checking in the ORC unwinder, part 2
Josh Poimboeuf [Fri, 5 Feb 2021 14:24:02 +0000 (08:24 -0600)]
x86/unwind/orc: Disable KASAN checking in the ORC unwinder, part 2

commit e504e74cc3a2c092b05577ce3e8e013fae7d94e6 upstream.

KASAN reserves "redzone" areas between stack frames in order to detect
stack overruns.  A read or write to such an area triggers a KASAN
"stack-out-of-bounds" BUG.

Normally, the ORC unwinder stays in-bounds and doesn't access the
redzone.  But sometimes it can't find ORC metadata for a given
instruction.  This can happen for code which is missing ORC metadata, or
for generated code.  In such cases, the unwinder attempts to fall back
to frame pointers, as a best-effort type thing.

This fallback often works, but when it doesn't, the unwinder can get
confused and go off into the weeds into the KASAN redzone, triggering
the aforementioned KASAN BUG.

But in this case, the unwinder's confusion is actually harmless and
working as designed.  It already has checks in place to prevent
off-stack accesses, but those checks get short-circuited by the KASAN
BUG.  And a BUG is a lot more disruptive than a harmless unwinder
warning.

Disable the KASAN checks by using READ_ONCE_NOCHECK() for all stack
accesses.  This finishes the job started by commit 881125bfe65b
("x86/unwind: Disable KASAN checking in the ORC unwinder"), which only
partially fixed the issue.

Fixes: ee9f8fce9964 ("x86/unwind: Add the ORC unwinder")
Reported-by: Ivan Babrou <ivan@cloudflare.com>
Signed-off-by: Josh Poimboeuf <jpoimboe@redhat.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Tested-by: Ivan Babrou <ivan@cloudflare.com>
Cc: stable@kernel.org
Link: https://lkml.kernel.org/r/9583327904ebbbeda399eca9c56d6c7085ac20fe.1612534649.git.jpoimboe@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agobinfmt_misc: fix possible deadlock in bm_register_write
Lior Ribak [Sat, 13 Mar 2021 05:07:41 +0000 (21:07 -0800)]
binfmt_misc: fix possible deadlock in bm_register_write

commit e7850f4d844e0acfac7e570af611d89deade3146 upstream.

There is a deadlock in bm_register_write:

First, in the begining of the function, a lock is taken on the binfmt_misc
root inode with inode_lock(d_inode(root)).

Then, if the user used the MISC_FMT_OPEN_FILE flag, the function will call
open_exec on the user-provided interpreter.

open_exec will call a path lookup, and if the path lookup process includes
the root of binfmt_misc, it will try to take a shared lock on its inode
again, but it is already locked, and the code will get stuck in a deadlock

To reproduce the bug:
$ echo ":iiiii:E::ii::/proc/sys/fs/binfmt_misc/bla:F" > /proc/sys/fs/binfmt_misc/register

backtrace of where the lock occurs (#5):
0  schedule () at ./arch/x86/include/asm/current.h:15
1  0xffffffff81b51237 in rwsem_down_read_slowpath (sem=0xffff888003b202e0, count=<optimized out>, state=state@entry=2) at kernel/locking/rwsem.c:992
2  0xffffffff81b5150a in __down_read_common (state=2, sem=<optimized out>) at kernel/locking/rwsem.c:1213
3  __down_read (sem=<optimized out>) at kernel/locking/rwsem.c:1222
4  down_read (sem=<optimized out>) at kernel/locking/rwsem.c:1355
5  0xffffffff811ee22a in inode_lock_shared (inode=<optimized out>) at ./include/linux/fs.h:783
6  open_last_lookups (op=0xffffc9000022fe34, file=0xffff888004098600, nd=0xffffc9000022fd10) at fs/namei.c:3177
7  path_openat (nd=nd@entry=0xffffc9000022fd10, op=op@entry=0xffffc9000022fe34, flags=flags@entry=65) at fs/namei.c:3366
8  0xffffffff811efe1c in do_filp_open (dfd=<optimized out>, pathname=pathname@entry=0xffff8880031b9000, op=op@entry=0xffffc9000022fe34) at fs/namei.c:3396
9  0xffffffff811e493f in do_open_execat (fd=fd@entry=-100, name=name@entry=0xffff8880031b9000, flags=<optimized out>, flags@entry=0) at fs/exec.c:913
10 0xffffffff811e4a92 in open_exec (name=<optimized out>) at fs/exec.c:948
11 0xffffffff8124aa84 in bm_register_write (file=<optimized out>, buffer=<optimized out>, count=19, ppos=<optimized out>) at fs/binfmt_misc.c:682
12 0xffffffff811decd2 in vfs_write (file=file@entry=0xffff888004098500, buf=buf@entry=0xa758d0 ":iiiii:E::ii::i:CF
", count=count@entry=19, pos=pos@entry=0xffffc9000022ff10) at fs/read_write.c:603
13 0xffffffff811defda in ksys_write (fd=<optimized out>, buf=0xa758d0 ":iiiii:E::ii::i:CF
", count=19) at fs/read_write.c:658
14 0xffffffff81b49813 in do_syscall_64 (nr=<optimized out>, regs=0xffffc9000022ff58) at arch/x86/entry/common.c:46
15 0xffffffff81c0007c in entry_SYSCALL_64 () at arch/x86/entry/entry_64.S:120

To solve the issue, the open_exec call is moved to before the write
lock is taken by bm_register_write

Link: https://lkml.kernel.org/r/20210228224414.95962-1-liorribak@gmail.com
Fixes: 948b701a607f1 ("binfmt_misc: add persistent opened binary handler for containers")
Signed-off-by: Lior Ribak <liorribak@gmail.com>
Acked-by: Helge Deller <deller@gmx.de>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agopowerpc/64s: Fix instruction encoding for lis in ppc_function_entry()
Naveen N. Rao [Thu, 4 Mar 2021 02:04:11 +0000 (07:34 +0530)]
powerpc/64s: Fix instruction encoding for lis in ppc_function_entry()

commit cea15316ceee2d4a51dfdecd79e08a438135416c upstream.

'lis r2,N' is 'addis r2,0,N' and the instruction encoding in the macro
LIS_R2 is incorrect (it currently maps to 'addis r0,r2,N'). Fix the
same.

Fixes: c71b7eff426f ("powerpc: Add ABIv2 support to ppc_function_entry")
Cc: stable@vger.kernel.org # v3.16+
Reported-by: Jiri Olsa <jolsa@redhat.com>
Signed-off-by: Naveen N. Rao <naveen.n.rao@linux.vnet.ibm.com>
Acked-by: Segher Boessenkool <segher@kernel.crashing.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20210304020411.16796-1-naveen.n.rao@linux.vnet.ibm.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agosched/membarrier: fix missing local execution of ipi_sync_rq_state()
Mathieu Desnoyers [Wed, 17 Feb 2021 16:56:51 +0000 (11:56 -0500)]
sched/membarrier: fix missing local execution of ipi_sync_rq_state()

commit ce29ddc47b91f97e7f69a0fb7cbb5845f52a9825 upstream.

The function sync_runqueues_membarrier_state() should copy the
membarrier state from the @mm received as parameter to each runqueue
currently running tasks using that mm.

However, the use of smp_call_function_many() skips the current runqueue,
which is unintended. Replace by a call to on_each_cpu_mask().

Fixes: 227a4aadc75b ("sched/membarrier: Fix p->mm->membarrier_state racy load")
Reported-by: Nadav Amit <nadav.amit@gmail.com>
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Cc: stable@vger.kernel.org # 5.4.x+
Link: https://lore.kernel.org/r/74F1E842-4A84-47BF-B6C2-5407DFDD4A4A@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agozram: fix return value on writeback_store
Minchan Kim [Sat, 13 Mar 2021 05:08:38 +0000 (21:08 -0800)]
zram: fix return value on writeback_store

commit 57e0076e6575a7b7cef620a0bd2ee2549ef77818 upstream.

writeback_store's return value is overwritten by submit_bio_wait's return
value.  Thus, writeback_store will return zero since there was no IO
error.  In the end, write syscall from userspace will see the zero as
return value, which could make the process stall to keep trying the write
until it will succeed.

Link: https://lkml.kernel.org/r/20210312173949.2197662-1-minchan@kernel.org
Fixes: 3b82a051c101("drivers/block/zram/zram_drv.c: fix error return codes not being returned in writeback_store")
Signed-off-by: Minchan Kim <minchan@kernel.org>
Cc: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Cc: Colin Ian King <colin.king@canonical.com>
Cc: John Dias <joaodias@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agoinclude/linux/sched/mm.h: use rcu_dereference in in_vfork()
Matthew Wilcox (Oracle) [Sat, 13 Mar 2021 05:08:03 +0000 (21:08 -0800)]
include/linux/sched/mm.h: use rcu_dereference in in_vfork()

[ Upstream commit 149fc787353f65b7e72e05e7b75d34863266c3e2 ]

Fix a sparse warning by using rcu_dereference().  Technically this is a
bug and a sufficiently aggressive compiler could reload the `real_parent'
pointer outside the protection of the rcu lock (and access freed memory),
but I think it's pretty unlikely to happen.

Link: https://lkml.kernel.org/r/20210221194207.1351703-1-willy@infradead.org
Fixes: b18dc5f291c0 ("mm, oom: skip vforked tasks from being selected")
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
Reviewed-by: Miaohe Lin <linmiaohe@huawei.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agostop_machine: mark helpers __always_inline
Arnd Bergmann [Sat, 13 Mar 2021 05:07:04 +0000 (21:07 -0800)]
stop_machine: mark helpers __always_inline

[ Upstream commit cbf78d85079cee662c45749ef4f744d41be85d48 ]

With clang-13, some functions only get partially inlined, with a
specialized version referring to a global variable.  This triggers a
harmless build-time check for the intel-rng driver:

WARNING: modpost: drivers/char/hw_random/intel-rng.o(.text+0xe): Section mismatch in reference from the function stop_machine() to the function .init.text:intel_rng_hw_init()
The function stop_machine() references
the function __init intel_rng_hw_init().
This is often because stop_machine lacks a __init
annotation or the annotation of intel_rng_hw_init is wrong.

In this instance, an easy workaround is to force the stop_machine()
function to be inline, along with related interfaces that did not show the
same behavior at the moment, but theoretically could.

The combination of the two patches listed below triggers the behavior in
clang-13, but individually these commits are correct.

Link: https://lkml.kernel.org/r/20210225130153.1956990-1-arnd@kernel.org
Fixes: fe5595c07400 ("stop_machine: Provide stop_machine_cpuslocked()")
Fixes: ee527cd3a20c ("Use stop_machine_run in the Intel RNG driver")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Nathan Chancellor <nathan@kernel.org>
Cc: Nick Desaulniers <ndesaulniers@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Cc: "Paul E. McKenney" <paulmck@kernel.org>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Prarit Bhargava <prarit@redhat.com>
Cc: Daniel Bristot de Oliveira <bristot@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Valentin Schneider <valentin.schneider@arm.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agohrtimer: Update softirq_expires_next correctly after __hrtimer_get_next_event()
Anna-Maria Behnsen [Tue, 23 Feb 2021 16:02:40 +0000 (17:02 +0100)]
hrtimer: Update softirq_expires_next correctly after __hrtimer_get_next_event()

[ Upstream commit 46eb1701c046cc18c032fa68f3c8ccbf24483ee4 ]

hrtimer_force_reprogram() and hrtimer_interrupt() invokes
__hrtimer_get_next_event() to find the earliest expiry time of hrtimer
bases. __hrtimer_get_next_event() does not update
cpu_base::[softirq_]_expires_next to preserve reprogramming logic. That
needs to be done at the callsites.

hrtimer_force_reprogram() updates cpu_base::softirq_expires_next only when
the first expiring timer is a softirq timer and the soft interrupt is not
activated. That's wrong because cpu_base::softirq_expires_next is left
stale when the first expiring timer of all bases is a timer which expires
in hard interrupt context. hrtimer_interrupt() does never update
cpu_base::softirq_expires_next which is wrong too.

That becomes a problem when clock_settime() sets CLOCK_REALTIME forward and
the first soft expiring timer is in the CLOCK_REALTIME_SOFT base. Setting
CLOCK_REALTIME forward moves the clock MONOTONIC based expiry time of that
timer before the stale cpu_base::softirq_expires_next.

cpu_base::softirq_expires_next is cached to make the check for raising the
soft interrupt fast. In the above case the soft interrupt won't be raised
until clock monotonic reaches the stale cpu_base::softirq_expires_next
value. That's incorrect, but what's worse it that if the softirq timer
becomes the first expiring timer of all clock bases after the hard expiry
timer has been handled the reprogramming of the clockevent from
hrtimer_interrupt() will result in an interrupt storm. That happens because
the reprogramming does not use cpu_base::softirq_expires_next, it uses
__hrtimer_get_next_event() which returns the actual expiry time. Once clock
MONOTONIC reaches cpu_base::softirq_expires_next the soft interrupt is
raised and the storm subsides.

Change the logic in hrtimer_force_reprogram() to evaluate the soft and hard
bases seperately, update softirq_expires_next and handle the case when a
soft expiring timer is the first of all bases by comparing the expiry times
and updating the required cpu base fields. Split this functionality into a
separate function to be able to use it in hrtimer_interrupt() as well
without copy paste.

Fixes: 5da70160462e ("hrtimer: Implement support for softirq based hrtimers")
Reported-by: Mikael Beckius <mikael.beckius@windriver.com>
Suggested-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Mikael Beckius <mikael.beckius@windriver.com>
Signed-off-by: Anna-Maria Behnsen <anna-maria@linutronix.de>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Link: https://lore.kernel.org/r/20210223160240.27518-1-anna-maria@linutronix.de
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoarm64: mm: use a 48-bit ID map when possible on 52-bit VA builds
Ard Biesheuvel [Wed, 10 Mar 2021 17:15:11 +0000 (18:15 +0100)]
arm64: mm: use a 48-bit ID map when possible on 52-bit VA builds

[ Upstream commit 7ba8f2b2d652cd8d8a2ab61f4be66973e70f9f88 ]

52-bit VA kernels can run on hardware that is only 48-bit capable, but
configure the ID map as 52-bit by default. This was not a problem until
recently, because the special T0SZ value for a 52-bit VA space was never
programmed into the TCR register anwyay, and because a 52-bit ID map
happens to use the same number of translation levels as a 48-bit one.

This behavior was changed by commit 1401bef703a4 ("arm64: mm: Always update
TCR_EL1 from __cpu_set_tcr_t0sz()"), which causes the unsupported T0SZ
value for a 52-bit VA to be programmed into TCR_EL1. While some hardware
simply ignores this, Mark reports that Amberwing systems choke on this,
resulting in a broken boot. But even before that commit, the unsupported
idmap_t0sz value was exposed to KVM and used to program TCR_EL2 incorrectly
as well.

Given that we already have to deal with address spaces being either 48-bit
or 52-bit in size, the cleanest approach seems to be to simply default to
a 48-bit VA ID map, and only switch to a 52-bit one if the placement of the
kernel in DRAM requires it. This is guaranteed not to happen unless the
system is actually 52-bit VA capable.

Fixes: 90ec95cda91a ("arm64: mm: Introduce VA_BITS_MIN")
Reported-by: Mark Salter <msalter@redhat.com>
Link: http://lore.kernel.org/r/20210310003216.410037-1-msalter@redhat.com
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Link: https://lore.kernel.org/r/20210310171515.416643-2-ardb@kernel.org
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoconfigfs: fix a use-after-free in __configfs_open_file
Daiyue Zhang [Mon, 1 Mar 2021 06:10:53 +0000 (14:10 +0800)]
configfs: fix a use-after-free in __configfs_open_file

[ Upstream commit 14fbbc8297728e880070f7b077b3301a8c698ef9 ]

Commit b0841eefd969 ("configfs: provide exclusion between IO and removals")
uses ->frag_dead to mark the fragment state, thus no bothering with extra
refcount on config_item when opening a file. The configfs_get_config_item
was removed in __configfs_open_file, but not with config_item_put. So the
refcount on config_item will lost its balance, causing use-after-free
issues in some occasions like this:

Test:
1. Mount configfs on /config with read-only items:
drwxrwx--- 289 root   root            0 2021-04-01 11:55 /config
drwxr-xr-x   2 root   root            0 2021-04-01 11:54 /config/a
--w--w--w-   1 root   root         4096 2021-04-01 11:53 /config/a/1.txt
......

2. Then run:
for file in /config
do
echo $file
grep -R 'key' $file
done

3. __configfs_open_file will be called in parallel, the first one
got called will do:
if (file->f_mode & FMODE_READ) {
if (!(inode->i_mode & S_IRUGO))
goto out_put_module;
config_item_put(buffer->item);
kref_put()
package_details_release()
kfree()

the other one will run into use-after-free issues like this:
BUG: KASAN: use-after-free in __configfs_open_file+0x1bc/0x3b0
Read of size 8 at addr fffffff155f02480 by task grep/13096
CPU: 0 PID: 13096 Comm: grep VIP: 00 Tainted: G        W       4.14.116-kasan #1
TGID: 13096 Comm: grep
Call trace:
dump_stack+0x118/0x160
kasan_report+0x22c/0x294
__asan_load8+0x80/0x88
__configfs_open_file+0x1bc/0x3b0
configfs_open_file+0x28/0x34
do_dentry_open+0x2cc/0x5c0
vfs_open+0x80/0xe0
path_openat+0xd8c/0x2988
do_filp_open+0x1c4/0x2fc
do_sys_open+0x23c/0x404
SyS_openat+0x38/0x48

Allocated by task 2138:
kasan_kmalloc+0xe0/0x1ac
kmem_cache_alloc_trace+0x334/0x394
packages_make_item+0x4c/0x180
configfs_mkdir+0x358/0x740
vfs_mkdir2+0x1bc/0x2e8
SyS_mkdirat+0x154/0x23c
el0_svc_naked+0x34/0x38

Freed by task 13096:
kasan_slab_free+0xb8/0x194
kfree+0x13c/0x910
package_details_release+0x524/0x56c
kref_put+0xc4/0x104
config_item_put+0x24/0x34
__configfs_open_file+0x35c/0x3b0
configfs_open_file+0x28/0x34
do_dentry_open+0x2cc/0x5c0
vfs_open+0x80/0xe0
path_openat+0xd8c/0x2988
do_filp_open+0x1c4/0x2fc
do_sys_open+0x23c/0x404
SyS_openat+0x38/0x48
el0_svc_naked+0x34/0x38

To fix this issue, remove the config_item_put in
__configfs_open_file to balance the refcount of config_item.

Fixes: b0841eefd969 ("configfs: provide exclusion between IO and removals")
Signed-off-by: Daiyue Zhang <zhangdaiyue1@huawei.com>
Signed-off-by: Yi Chen <chenyi77@huawei.com>
Signed-off-by: Ge Qiu <qiuge@huawei.com>
Reviewed-by: Chao Yu <yuchao0@huawei.com>
Acked-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoblock: rsxx: fix error return code of rsxx_pci_probe()
Jia-Ju Bai [Wed, 10 Mar 2021 03:30:17 +0000 (19:30 -0800)]
block: rsxx: fix error return code of rsxx_pci_probe()

[ Upstream commit df66617bfe87487190a60783d26175b65d2502ce ]

When create_singlethread_workqueue returns NULL to card->event_wq, no
error return code of rsxx_pci_probe() is assigned.

To fix this bug, st is assigned with -ENOMEM in this case.

Fixes: 8722ff8cdbfa ("block: IBM RamSan 70/80 device driver")
Reported-by: TOTE Robot <oslab@tsinghua.edu.cn>
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Link: https://lore.kernel.org/r/20210310033017.4023-1-baijiaju1990@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoNFSv4.2: fix return value of _nfs4_get_security_label()
Ondrej Mosnacek [Fri, 15 Jan 2021 17:43:56 +0000 (18:43 +0100)]
NFSv4.2: fix return value of _nfs4_get_security_label()

[ Upstream commit 53cb245454df5b13d7063162afd7a785aed6ebf2 ]

An xattr 'get' handler is expected to return the length of the value on
success, yet _nfs4_get_security_label() (and consequently also
nfs4_xattr_get_nfs4_label(), which is used as an xattr handler) returns
just 0 on success.

Fix this by returning label.len instead, which contains the length of
the result.

Fixes: aa9c2669626c ("NFS: Client implementation of Labeled-NFS")
Signed-off-by: Ondrej Mosnacek <omosnace@redhat.com>
Reviewed-by: James Morris <jamorris@linux.microsoft.com>
Reviewed-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoNFS: Don't gratuitously clear the inode cache when lookup failed
Trond Myklebust [Mon, 8 Mar 2021 19:42:52 +0000 (14:42 -0500)]
NFS: Don't gratuitously clear the inode cache when lookup failed

[ Upstream commit 47397915ede0192235474b145ebcd81b37b03624 ]

The fact that the lookup revalidation failed, does not mean that the
inode contents have changed.

Fixes: 5ceb9d7fdaaf ("NFS: Refactor nfs_lookup_revalidate()")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoNFS: Don't revalidate the directory permissions on a lookup failure
Trond Myklebust [Mon, 8 Mar 2021 19:42:51 +0000 (14:42 -0500)]
NFS: Don't revalidate the directory permissions on a lookup failure

[ Upstream commit 82e7ca1334ab16e2e04fafded1cab9dfcdc11b40 ]

There should be no reason to expect the directory permissions to change
just because the directory contents changed or a negative lookup timed
out. So let's avoid doing a full call to nfs_mark_for_revalidate() in
that case.
Furthermore, if this is a negative dentry, and we haven't actually done
a new lookup, then we have no reason yet to believe the directory has
changed at all. So let's remove the gratuitous directory inode
invalidation altogether when called from
nfs_lookup_revalidate_negative().

Reported-by: Geert Jansen <gerardu@amazon.com>
Fixes: 5ceb9d7fdaaf ("NFS: Refactor nfs_lookup_revalidate()")
Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoSUNRPC: Set memalloc_nofs_save() for sync tasks
Benjamin Coddington [Wed, 3 Mar 2021 13:47:16 +0000 (08:47 -0500)]
SUNRPC: Set memalloc_nofs_save() for sync tasks

[ Upstream commit f0940f4b3284a00f38a5d42e6067c2aaa20e1f2e ]

We could recurse into NFS doing memory reclaim while sending a sync task,
which might result in a deadlock.  Set memalloc_nofs_save for sync task
execution.

Fixes: a1231fda7e94 ("SUNRPC: Set memalloc_nofs_save() on all rpciod/xprtiod jobs")
Signed-off-by: Benjamin Coddington <bcodding@redhat.com>
Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agoarm64/mm: Fix pfn_valid() for ZONE_DEVICE based memory
Anshuman Khandual [Fri, 5 Mar 2021 05:24:57 +0000 (10:54 +0530)]
arm64/mm: Fix pfn_valid() for ZONE_DEVICE based memory

[ Upstream commit eeb0753ba27b26f609e61f9950b14f1b934fe429 ]

pfn_valid() validates a pfn but basically it checks for a valid struct page
backing for that pfn. It should always return positive for memory ranges
backed with struct page mapping. But currently pfn_valid() fails for all
ZONE_DEVICE based memory types even though they have struct page mapping.

pfn_valid() asserts that there is a memblock entry for a given pfn without
MEMBLOCK_NOMAP flag being set. The problem with ZONE_DEVICE based memory is
that they do not have memblock entries. Hence memblock_is_map_memory() will
invariably fail via memblock_search() for a ZONE_DEVICE based address. This
eventually fails pfn_valid() which is wrong. memblock_is_map_memory() needs
to be skipped for such memory ranges. As ZONE_DEVICE memory gets hotplugged
into the system via memremap_pages() called from a driver, their respective
memory sections will not have SECTION_IS_EARLY set.

Normal hotplug memory will never have MEMBLOCK_NOMAP set in their memblock
regions. Because the flag MEMBLOCK_NOMAP was specifically designed and set
for firmware reserved memory regions. memblock_is_map_memory() can just be
skipped as its always going to be positive and that will be an optimization
for the normal hotplug memory. Like ZONE_DEVICE based memory, all normal
hotplugged memory too will not have SECTION_IS_EARLY set for their sections

Skipping memblock_is_map_memory() for all non early memory sections would
fix pfn_valid() problem for ZONE_DEVICE based memory and also improve its
performance for normal hotplug memory as well.

Cc: Catalin Marinas <catalin.marinas@arm.com>
Cc: Will Deacon <will@kernel.org>
Cc: Ard Biesheuvel <ardb@kernel.org>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: linux-arm-kernel@lists.infradead.org
Cc: linux-kernel@vger.kernel.org
Acked-by: David Hildenbrand <david@redhat.com>
Fixes: 73b20c84d42d ("arm64: mm: implement pte_devmap support")
Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
Acked-by: Catalin Marinas <catalin.marinas@arm.com>
Link: https://lore.kernel.org/r/1614921898-4099-2-git-send-email-anshuman.khandual@arm.com
Signed-off-by: Will Deacon <will@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agosh_eth: fix TRSCER mask for R7S72100
Sergey Shtylyov [Sun, 28 Feb 2021 20:26:34 +0000 (23:26 +0300)]
sh_eth: fix TRSCER mask for R7S72100

[ Upstream commit 75be7fb7f978202c4c3a1a713af4485afb2ff5f6 ]

According  to  the RZ/A1H Group, RZ/A1M Group User's Manual: Hardware,
Rev. 4.00, the TRSCER register has bit 9 reserved, hence we can't use
the driver's default TRSCER mask.  Add the explicit initializer for
sh_eth_cpu_data::trscer_err_mask for R7S72100.

Fixes: db893473d313 ("sh_eth: Add support for r7s72100")
Signed-off-by: Sergey Shtylyov <s.shtylyov@omprussia.ru>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>
4 years agostaging: comedi: pcl818: Fix endian problem for AI command data
Ian Abbott [Tue, 23 Feb 2021 14:30:50 +0000 (14:30 +0000)]
staging: comedi: pcl818: Fix endian problem for AI command data

commit 148e34fd33d53740642db523724226de14ee5281 upstream.

The analog input subdevice supports Comedi asynchronous commands that
use Comedi's 16-bit sample format.  However, the call to
`comedi_buf_write_samples()` is passing the address of a 32-bit integer
parameter.  On bigendian machines, this will copy 2 bytes from the wrong
end of the 32-bit value.  Fix it by changing the type of the parameter
holding the sample value to `unsigned short`.

[Note: the bug was introduced in commit edf4537bcbf5 ("staging: comedi:
pcl818: use comedi_buf_write_samples()") but the patch applies better to
commit d615416de615 ("staging: comedi: pcl818: introduce
pcl818_ai_write_sample()").]

Fixes: d615416de615 ("staging: comedi: pcl818: introduce pcl818_ai_write_sample()")
Cc: <stable@vger.kernel.org> # 4.0+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Link: https://lore.kernel.org/r/20210223143055.257402-10-abbotti@mev.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: comedi: pcl711: Fix endian problem for AI command data
Ian Abbott [Tue, 23 Feb 2021 14:30:49 +0000 (14:30 +0000)]
staging: comedi: pcl711: Fix endian problem for AI command data

commit a084303a645896e834883f2c5170d044410dfdb3 upstream.

The analog input subdevice supports Comedi asynchronous commands that
use Comedi's 16-bit sample format.  However, the call to
`comedi_buf_write_samples()` is passing the address of a 32-bit integer
variable.  On bigendian machines, this will copy 2 bytes from the wrong
end of the 32-bit value.  Fix it by changing the type of the variable
holding the sample value to `unsigned short`.

Fixes: 1f44c034de2e ("staging: comedi: pcl711: use comedi_buf_write_samples()")
Cc: <stable@vger.kernel.org> # 3.19+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Link: https://lore.kernel.org/r/20210223143055.257402-9-abbotti@mev.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: comedi: me4000: Fix endian problem for AI command data
Ian Abbott [Tue, 23 Feb 2021 14:30:48 +0000 (14:30 +0000)]
staging: comedi: me4000: Fix endian problem for AI command data

commit b39dfcced399d31e7c4b7341693b18e01c8f655e upstream.

The analog input subdevice supports Comedi asynchronous commands that
use Comedi's 16-bit sample format.  However, the calls to
`comedi_buf_write_samples()` are passing the address of a 32-bit integer
variable.  On bigendian machines, this will copy 2 bytes from the wrong
end of the 32-bit value.  Fix it by changing the type of the variable
holding the sample value to `unsigned short`.

Fixes: de88924f67d1 ("staging: comedi: me4000: use comedi_buf_write_samples()")
Cc: <stable@vger.kernel.org> # 3.19+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Link: https://lore.kernel.org/r/20210223143055.257402-8-abbotti@mev.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: comedi: dmm32at: Fix endian problem for AI command data
Ian Abbott [Tue, 23 Feb 2021 14:30:47 +0000 (14:30 +0000)]
staging: comedi: dmm32at: Fix endian problem for AI command data

commit 54999c0d94b3c26625f896f8e3460bc029821578 upstream.

The analog input subdevice supports Comedi asynchronous commands that
use Comedi's 16-bit sample format.  However, the call to
`comedi_buf_write_samples()` is passing the address of a 32-bit integer
variable.  On bigendian machines, this will copy 2 bytes from the wrong
end of the 32-bit value.  Fix it by changing the type of the variable
holding the sample value to `unsigned short`.

[Note: the bug was introduced in commit 1700529b24cc ("staging: comedi:
dmm32at: use comedi_buf_write_samples()") but the patch applies better
to the later (but in the same kernel release) commit 0c0eadadcbe6e
("staging: comedi: dmm32at: introduce dmm32_ai_get_sample()").]

Fixes: 0c0eadadcbe6e ("staging: comedi: dmm32at: introduce dmm32_ai_get_sample()")
Cc: <stable@vger.kernel.org> # 3.19+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Link: https://lore.kernel.org/r/20210223143055.257402-7-abbotti@mev.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: comedi: das800: Fix endian problem for AI command data
Ian Abbott [Tue, 23 Feb 2021 14:30:46 +0000 (14:30 +0000)]
staging: comedi: das800: Fix endian problem for AI command data

commit 459b1e8c8fe97fcba0bd1b623471713dce2c5eaf upstream.

The analog input subdevice supports Comedi asynchronous commands that
use Comedi's 16-bit sample format.  However, the call to
`comedi_buf_write_samples()` is passing the address of a 32-bit integer
variable.  On bigendian machines, this will copy 2 bytes from the wrong
end of the 32-bit value.  Fix it by changing the type of the variable
holding the sample value to `unsigned short`.

Fixes: ad9eb43c93d8 ("staging: comedi: das800: use comedi_buf_write_samples()")
Cc: <stable@vger.kernel.org> # 3.19+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Link: https://lore.kernel.org/r/20210223143055.257402-6-abbotti@mev.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: comedi: das6402: Fix endian problem for AI command data
Ian Abbott [Tue, 23 Feb 2021 14:30:45 +0000 (14:30 +0000)]
staging: comedi: das6402: Fix endian problem for AI command data

commit 1c0f20b78781b9ca50dc3ecfd396d0db5b141890 upstream.

The analog input subdevice supports Comedi asynchronous commands that
use Comedi's 16-bit sample format.  However, the call to
`comedi_buf_write_samples()` is passing the address of a 32-bit integer
variable.  On bigendian machines, this will copy 2 bytes from the wrong
end of the 32-bit value.  Fix it by changing the type of the variable
holding the sample value to `unsigned short`.

Fixes: d1d24cb65ee3 ("staging: comedi: das6402: read analog input samples in interrupt handler")
Cc: <stable@vger.kernel.org> # 3.19+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Link: https://lore.kernel.org/r/20210223143055.257402-5-abbotti@mev.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: comedi: adv_pci1710: Fix endian problem for AI command data
Ian Abbott [Tue, 23 Feb 2021 14:30:44 +0000 (14:30 +0000)]
staging: comedi: adv_pci1710: Fix endian problem for AI command data

commit b2e78630f733a76508b53ba680528ca39c890e82 upstream.

The analog input subdevice supports Comedi asynchronous commands that
use Comedi's 16-bit sample format.  However, the calls to
`comedi_buf_write_samples()` are passing the address of a 32-bit integer
variable.  On bigendian machines, this will copy 2 bytes from the wrong
end of the 32-bit value.  Fix it by changing the type of the variables
holding the sample value to `unsigned short`.  The type of the `val`
parameter of `pci1710_ai_read_sample()` is changed to `unsigned short *`
accordingly.  The type of the `val` variable in `pci1710_ai_insn_read()`
is also changed to `unsigned short` since its address is passed to
`pci1710_ai_read_sample()`.

Fixes: a9c3a015c12f ("staging: comedi: adv_pci1710: use comedi_buf_write_samples()")
Cc: <stable@vger.kernel.org> # 4.0+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Link: https://lore.kernel.org/r/20210223143055.257402-4-abbotti@mev.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: comedi: addi_apci_1500: Fix endian problem for command sample
Ian Abbott [Tue, 23 Feb 2021 14:30:43 +0000 (14:30 +0000)]
staging: comedi: addi_apci_1500: Fix endian problem for command sample

commit ac0bbf55ed3be75fde1f8907e91ecd2fd589bde3 upstream.

The digital input subdevice supports Comedi asynchronous commands that
read interrupt status information.  This uses 16-bit Comedi samples (of
which only the bottom 8 bits contain status information).  However, the
interrupt handler is calling `comedi_buf_write_samples()` with the
address of a 32-bit variable `unsigned int status`.  On a bigendian
machine, this will copy 2 bytes from the wrong end of the variable.  Fix
it by changing the type of the variable to `unsigned short`.

Fixes: a8c66b684efa ("staging: comedi: addi_apci_1500: rewrite the subdevice support functions")
Cc: <stable@vger.kernel.org> #4.0+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Link: https://lore.kernel.org/r/20210223143055.257402-3-abbotti@mev.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: comedi: addi_apci_1032: Fix endian problem for COS sample
Ian Abbott [Tue, 23 Feb 2021 14:30:42 +0000 (14:30 +0000)]
staging: comedi: addi_apci_1032: Fix endian problem for COS sample

commit 25317f428a78fde71b2bf3f24d05850f08a73a52 upstream.

The Change-Of-State (COS) subdevice supports Comedi asynchronous
commands to read 16-bit change-of-state values.  However, the interrupt
handler is calling `comedi_buf_write_samples()` with the address of a
32-bit integer `&s->state`.  On bigendian architectures, it will copy 2
bytes from the wrong end of the 32-bit integer.  Fix it by transferring
the value via a 16-bit integer.

Fixes: 6bb45f2b0c86 ("staging: comedi: addi_apci_1032: use comedi_buf_write_samples()")
Cc: <stable@vger.kernel.org> # 3.19+
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Link: https://lore.kernel.org/r/20210223143055.257402-2-abbotti@mev.co.uk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: rtl8192e: Fix possible buffer overflow in _rtl92e_wx_set_scan
Lee Gibson [Fri, 26 Feb 2021 14:51:57 +0000 (14:51 +0000)]
staging: rtl8192e: Fix possible buffer overflow in _rtl92e_wx_set_scan

commit 8687bf9ef9551bcf93897e33364d121667b1aadf upstream.

Function _rtl92e_wx_set_scan calls memcpy without checking the length.
A user could control that length and trigger a buffer overflow.
Fix by checking the length is within the maximum allowed size.

Reviewed-by: Dan Carpenter <dan.carpenter@oracle.com>
Signed-off-by: Lee Gibson <leegib@gmail.com>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20210226145157.424065-1-leegib@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: rtl8712: Fix possible buffer overflow in r8712_sitesurvey_cmd
Lee Gibson [Mon, 1 Mar 2021 13:26:48 +0000 (13:26 +0000)]
staging: rtl8712: Fix possible buffer overflow in r8712_sitesurvey_cmd

commit b93c1e3981af19527beee1c10a2bef67a228c48c upstream.

Function r8712_sitesurvey_cmd calls memcpy without checking the length.
A user could control that length and trigger a buffer overflow.
Fix by checking the length is within the maximum allowed size.

Signed-off-by: Lee Gibson <leegib@gmail.com>
Link: https://lore.kernel.org/r/20210301132648.420296-1-leegib@gmail.com
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: ks7010: prevent buffer overflow in ks_wlan_set_scan()
Dan Carpenter [Tue, 2 Mar 2021 11:19:39 +0000 (14:19 +0300)]
staging: ks7010: prevent buffer overflow in ks_wlan_set_scan()

commit e163b9823a0b08c3bb8dc4f5b4b5c221c24ec3e5 upstream.

The user can specify a "req->essid_len" of up to 255 but if it's
over IW_ESSID_MAX_SIZE (32) that can lead to memory corruption.

Fixes: 13a9930d15b4 ("staging: ks7010: add driver from Nanonote extra-repository")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/YD4fS8+HmM/Qmrw6@mwanda
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: rtl8188eu: fix potential memory corruption in rtw_check_beacon_data()
Dan Carpenter [Fri, 5 Mar 2021 08:56:32 +0000 (11:56 +0300)]
staging: rtl8188eu: fix potential memory corruption in rtw_check_beacon_data()

commit d4ac640322b06095128a5c45ba4a1e80929fe7f3 upstream.

The "ie_len" is a value in the 1-255 range that comes from the user.  We
have to cap it to ensure that it's not too large or it could lead to
memory corruption.

Fixes: 9a7fe54ddc3a ("staging: r8188eu: Add source files for new driver - part 1")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/YEHyQCrFZKTXyT7J@mwanda
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: rtl8712: unterminated string leads to read overflow
Dan Carpenter [Wed, 24 Feb 2021 08:45:59 +0000 (11:45 +0300)]
staging: rtl8712: unterminated string leads to read overflow

commit d660f4f42ccea50262c6ee90c8e7ad19a69fb225 upstream.

The memdup_user() function does not necessarily return a NUL terminated
string so this can lead to a read overflow.  Switch from memdup_user()
to strndup_user() to fix this bug.

Fixes: c6dc001f2add ("staging: r8712u: Merging Realtek's latest (v2.6.6). Various fixes.")
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Link: https://lore.kernel.org/r/YDYSR+1rj26NRhvb@mwanda
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: rtl8188eu: prevent ->ssid overflow in rtw_wx_set_scan()
Dan Carpenter [Fri, 5 Mar 2021 08:58:03 +0000 (11:58 +0300)]
staging: rtl8188eu: prevent ->ssid overflow in rtw_wx_set_scan()

commit 74b6b20df8cfe90ada777d621b54c32e69e27cd7 upstream.

This code has a check to prevent read overflow but it needs another
check to prevent writing beyond the end of the ->ssid[] array.

Fixes: a2c60d42d97c ("staging: r8188eu: Add files for new driver - part 16")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/YEHymwsnHewzoam7@mwanda
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agostaging: rtl8192u: fix ->ssid overflow in r8192_wx_set_scan()
Dan Carpenter [Fri, 5 Mar 2021 08:12:49 +0000 (11:12 +0300)]
staging: rtl8192u: fix ->ssid overflow in r8192_wx_set_scan()

commit 87107518d7a93fec6cdb2559588862afeee800fb upstream.

We need to cap len at IW_ESSID_MAX_SIZE (32) to avoid memory corruption.
This can be controlled by the user via the ioctl.

Fixes: 5f53d8ca3d5d ("Staging: add rtl8192SU wireless usb driver")
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Cc: stable <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/YEHoAWMOSZBUw91F@mwanda
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agomisc: fastrpc: restrict user apps from sending kernel RPC messages
Dmitry Baryshkov [Fri, 12 Feb 2021 19:26:58 +0000 (22:26 +0300)]
misc: fastrpc: restrict user apps from sending kernel RPC messages

commit 20c40794eb85ea29852d7bc37c55713802a543d6 upstream.

Verify that user applications are not using the kernel RPC message
handle to restrict them from directly attaching to guest OS on the
remote subsystem. This is a port of CVE-2019-2308 fix.

Fixes: c68cfb718c8f ("misc: fastrpc: Add support for context Invoke method")
Cc: Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
Cc: Jonathan Marek <jonathan@marek.ca>
Cc: stable@vger.kernel.org
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Link: https://lore.kernel.org/r/20210212192658.3476137-1-dmitry.baryshkov@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agomisc/pvpanic: Export module FDT device table
Shile Zhang [Thu, 18 Feb 2021 12:31:16 +0000 (20:31 +0800)]
misc/pvpanic: Export module FDT device table

commit 65527a51c66f4edfa28602643d7dd4fa366eb826 upstream.

Export the module FDT device table to ensure the FDT compatible strings
are listed in the module alias. This help the pvpanic driver can be
loaded on boot automatically not only the ACPI device, but also the FDT
device.

Fixes: 46f934c9a12fc ("misc/pvpanic: add support to get pvpanic device info FDT")
Signed-off-by: Shile Zhang <shile.zhang@linux.alibaba.com>
Link: https://lore.kernel.org/r/20210218123116.207751-1-shile.zhang@linux.alibaba.com
Cc: stable <stable@vger.kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agousbip: fix vudc usbip_sockfd_store races leading to gpf
Shuah Khan [Mon, 8 Mar 2021 03:53:31 +0000 (20:53 -0700)]
usbip: fix vudc usbip_sockfd_store races leading to gpf

commit 46613c9dfa964c0c60b5385dbdf5aaa18be52a9c upstream.

usbip_sockfd_store() is invoked when user requests attach (import)
detach (unimport) usb gadget device from usbip host. vhci_hcd sends
import request and usbip_sockfd_store() exports the device if it is
free for export.

Export and unexport are governed by local state and shared state
- Shared state (usbip device status, sockfd) - sockfd and Device
  status are used to determine if stub should be brought up or shut
  down. Device status is shared between host and client.
- Local state (tcp_socket, rx and tx thread task_struct ptrs)
  A valid tcp_socket controls rx and tx thread operations while the
  device is in exported state.
- While the device is exported, device status is marked used and socket,
  sockfd, and thread pointers are valid.

Export sequence (stub-up) includes validating the socket and creating
receive (rx) and transmit (tx) threads to talk to the client to provide
access to the exported device. rx and tx threads depends on local and
shared state to be correct and in sync.

Unexport (stub-down) sequence shuts the socket down and stops the rx and
tx threads. Stub-down sequence relies on local and shared states to be
in sync.

There are races in updating the local and shared status in the current
stub-up sequence resulting in crashes. These stem from starting rx and
tx threads before local and global state is updated correctly to be in
sync.

1. Doesn't handle kthread_create() error and saves invalid ptr in local
   state that drives rx and tx threads.
2. Updates tcp_socket and sockfd,  starts stub_rx and stub_tx threads
   before updating usbip_device status to SDEV_ST_USED. This opens up a
   race condition between the threads and usbip_sockfd_store() stub up
   and down handling.

Fix the above problems:
- Stop using kthread_get_run() macro to create/start threads.
- Create threads and get task struct reference.
- Add kthread_create() failure handling and bail out.
- Hold usbip_device lock to update local and shared states after
  creating rx and tx threads.
- Update usbip_device status to SDEV_ST_USED.
- Update usbip_device tcp_socket, sockfd, tcp_rx, and tcp_tx
- Start threads after usbip_device (tcp_socket, sockfd, tcp_rx, tcp_tx,
  and status) is complete.

Credit goes to syzbot and Tetsuo Handa for finding and root-causing the
kthread_get_run() improper error handling problem and others. This is a
hard problem to find and debug since the races aren't seen in a normal
case. Fuzzing forces the race window to be small enough for the
kthread_get_run() error path bug and starting threads before updating the
local and shared state bug in the stub-up sequence.

Fixes: 9720b4bc76a83807 ("staging/usbip: convert to kthread")
Cc: stable@vger.kernel.org
Reported-by: syzbot <syzbot+a93fba6d384346a761e3@syzkaller.appspotmail.com>
Reported-by: syzbot <syzbot+bf1a360e305ee719e364@syzkaller.appspotmail.com>
Reported-by: syzbot <syzbot+95ce4b142579611ef0a9@syzkaller.appspotmail.com>
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Link: https://lore.kernel.org/r/b1c08b983ffa185449c9f0f7d1021dc8c8454b60.1615171203.git.skhan@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agousbip: fix vhci_hcd attach_store() races leading to gpf
Shuah Khan [Mon, 8 Mar 2021 03:53:30 +0000 (20:53 -0700)]
usbip: fix vhci_hcd attach_store() races leading to gpf

commit 718ad9693e3656120064b715fe931f43a6201e67 upstream.

attach_store() is invoked when user requests import (attach) a device
from usbip host.

Attach and detach are governed by local state and shared state
- Shared state (usbip device status) - Device status is used to manage
  the attach and detach operations on import-able devices.
- Local state (tcp_socket, rx and tx thread task_struct ptrs)
  A valid tcp_socket controls rx and tx thread operations while the
  device is in exported state.
- Device has to be in the right state to be attached and detached.

Attach sequence includes validating the socket and creating receive (rx)
and transmit (tx) threads to talk to the host to get access to the
imported device. rx and tx threads depends on local and shared state to
be correct and in sync.

Detach sequence shuts the socket down and stops the rx and tx threads.
Detach sequence relies on local and shared states to be in sync.

There are races in updating the local and shared status in the current
attach sequence resulting in crashes. These stem from starting rx and
tx threads before local and global state is updated correctly to be in
sync.

1. Doesn't handle kthread_create() error and saves invalid ptr in local
   state that drives rx and tx threads.
2. Updates tcp_socket and sockfd,  starts stub_rx and stub_tx threads
   before updating usbip_device status to VDEV_ST_NOTASSIGNED. This opens
   up a race condition between the threads, port connect, and detach
   handling.

Fix the above problems:
- Stop using kthread_get_run() macro to create/start threads.
- Create threads and get task struct reference.
- Add kthread_create() failure handling and bail out.
- Hold vhci and usbip_device locks to update local and shared states after
  creating rx and tx threads.
- Update usbip_device status to VDEV_ST_NOTASSIGNED.
- Update usbip_device tcp_socket, sockfd, tcp_rx, and tcp_tx
- Start threads after usbip_device (tcp_socket, sockfd, tcp_rx, tcp_tx,
  and status) is complete.

Credit goes to syzbot and Tetsuo Handa for finding and root-causing the
kthread_get_run() improper error handling problem and others. This is
hard problem to find and debug since the races aren't seen in a normal
case. Fuzzing forces the race window to be small enough for the
kthread_get_run() error path bug and starting threads before updating the
local and shared state bug in the attach sequence.
- Update usbip_device tcp_rx and tcp_tx pointers holding vhci and
  usbip_device locks.

Tested with syzbot reproducer:
- https://syzkaller.appspot.com/text?tag=ReproC&x=14801034d00000

Fixes: 9720b4bc76a83807 ("staging/usbip: convert to kthread")
Cc: stable@vger.kernel.org
Reported-by: syzbot <syzbot+a93fba6d384346a761e3@syzkaller.appspotmail.com>
Reported-by: syzbot <syzbot+bf1a360e305ee719e364@syzkaller.appspotmail.com>
Reported-by: syzbot <syzbot+95ce4b142579611ef0a9@syzkaller.appspotmail.com>
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Link: https://lore.kernel.org/r/bb434bd5d7a64fbec38b5ecfb838a6baef6eb12b.1615171203.git.skhan@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agousbip: fix stub_dev usbip_sockfd_store() races leading to gpf
Shuah Khan [Mon, 8 Mar 2021 03:53:29 +0000 (20:53 -0700)]
usbip: fix stub_dev usbip_sockfd_store() races leading to gpf

commit 9380afd6df70e24eacbdbde33afc6a3950965d22 upstream.

usbip_sockfd_store() is invoked when user requests attach (import)
detach (unimport) usb device from usbip host. vhci_hcd sends import
request and usbip_sockfd_store() exports the device if it is free
for export.

Export and unexport are governed by local state and shared state
- Shared state (usbip device status, sockfd) - sockfd and Device
  status are used to determine if stub should be brought up or shut
  down.
- Local state (tcp_socket, rx and tx thread task_struct ptrs)
  A valid tcp_socket controls rx and tx thread operations while the
  device is in exported state.
- While the device is exported, device status is marked used and socket,
  sockfd, and thread pointers are valid.

Export sequence (stub-up) includes validating the socket and creating
receive (rx) and transmit (tx) threads to talk to the client to provide
access to the exported device. rx and tx threads depends on local and
shared state to be correct and in sync.

Unexport (stub-down) sequence shuts the socket down and stops the rx and
tx threads. Stub-down sequence relies on local and shared states to be
in sync.

There are races in updating the local and shared status in the current
stub-up sequence resulting in crashes. These stem from starting rx and
tx threads before local and global state is updated correctly to be in
sync.

1. Doesn't handle kthread_create() error and saves invalid ptr in local
   state that drives rx and tx threads.
2. Updates tcp_socket and sockfd,  starts stub_rx and stub_tx threads
   before updating usbip_device status to SDEV_ST_USED. This opens up a
   race condition between the threads and usbip_sockfd_store() stub up
   and down handling.

Fix the above problems:
- Stop using kthread_get_run() macro to create/start threads.
- Create threads and get task struct reference.
- Add kthread_create() failure handling and bail out.
- Hold usbip_device lock to update local and shared states after
  creating rx and tx threads.
- Update usbip_device status to SDEV_ST_USED.
- Update usbip_device tcp_socket, sockfd, tcp_rx, and tcp_tx
- Start threads after usbip_device (tcp_socket, sockfd, tcp_rx, tcp_tx,
  and status) is complete.

Credit goes to syzbot and Tetsuo Handa for finding and root-causing the
kthread_get_run() improper error handling problem and others. This is a
hard problem to find and debug since the races aren't seen in a normal
case. Fuzzing forces the race window to be small enough for the
kthread_get_run() error path bug and starting threads before updating the
local and shared state bug in the stub-up sequence.

Tested with syzbot reproducer:
- https://syzkaller.appspot.com/text?tag=ReproC&x=14801034d00000

Fixes: 9720b4bc76a83807 ("staging/usbip: convert to kthread")
Cc: stable@vger.kernel.org
Reported-by: syzbot <syzbot+a93fba6d384346a761e3@syzkaller.appspotmail.com>
Reported-by: syzbot <syzbot+bf1a360e305ee719e364@syzkaller.appspotmail.com>
Reported-by: syzbot <syzbot+95ce4b142579611ef0a9@syzkaller.appspotmail.com>
Reported-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Link: https://lore.kernel.org/r/268a0668144d5ff36ec7d87fdfa90faf583b7ccc.1615171203.git.skhan@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
4 years agousbip: fix vudc to check for stream socket
Shuah Khan [Mon, 8 Mar 2021 03:53:28 +0000 (20:53 -0700)]
usbip: fix vudc to check for stream socket

commit 6801854be94fe8819b3894979875ea31482f5658 upstream.

Fix usbip_sockfd_store() to validate the passed in file descriptor is
a stream socket. If the file descriptor passed was a SOCK_DGRAM socket,
sock_recvmsg() can't detect end of stream.

Cc: stable@vger.kernel.org
Suggested-by: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Signed-off-by: Shuah Khan <skhan@linuxfoundation.org>
Link: https://lore.kernel.org/r/387a670316002324113ac7ea1e8b53f4085d0c95.1615171203.git.skhan@linuxfoundation.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>