usb: xhci: fix TDS for MTK xHCI1.1

commit 72b663a99c074a8d073e7ecdae446cfb024ef551 upstream.

For MTK's xHCI 1.0 or latter, TD size is the number of max
packet sized packets remaining in the TD, not including
this TRB (following spec).

For MTK's xHCI 0.96 and older, TD size is the number of max
packet sized packets remaining in the TD, including this TRB
(not following spec).

Signed-off-by: Chunfeng Yun <chunfeng.yun@mediatek.com>
Signed-off-by: Mathias Nyman <mathias.nyman@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ceph: drop negative child dentries before try pruning inode's alias

commit 040d786032bf59002d374b86d75b04d97624005c upstream.

Negative child dentry holds reference on inode's alias, it makes
d_prune_aliases() do nothing.

Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: core: apply NO_CMD23 quirk to some specific cards

commit 91516a2a4734614d62ee3ed921f8f88acc67c000 upstream.

To get an usdhc Apacer and some ATP SD cards work reliable, CMD23 needs
to be disabled.  This has been tested on i.MX6 (sdhci-esdhc) and rk3288
(dw_mmc-rockchip).

Without this patch on i.MX6 (sdhci-esdhc):

 $ dd if=/dev/urandom of=/mnt/test bs=1M count=10 conv=fsync

    | <mmc0: starting CMD23 arg 00000400 flags 00000015>
    | mmc0: starting CMD25 arg 00a71f00 flags 000000b5
    | mmc0:     blksz 512 blocks 1024 flags 00000100 tsac 3000 ms nsac 0
    | mmc0:     CMD12 arg 00000000 flags 0000049d
    | sdhci [sdhci_irq()]: *** mmc0 got interrupt: 0x00000001
    | mmc0: Timeout waiting for hardware interrupt.

Without this patch on rk3288 (dw_mmc-rockchip):

    | mmc1: Card stuck in programming state! mmcblk1 card_busy_detect
    | dwmmc_rockchip ff0c0000.dwmmc: Busy; trying anyway
    | mmc_host mmc1: Bus speed (slot 0) = 400000Hz (slot req 400000Hz,
    | actual 400000HZ div = 0)
    | mmc1: card never left busy state
    | mmc1: tried to reset card, got error -110
    | blk_update_request: I/O error, dev mmcblk1, sector 139778
    | Buffer I/O error on dev mmcblk1p1, logical block 131586, lost async
    | page write

Signed-off-by: Christoph Fritz <chf.fritz@googlemail.com>
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usbip: fix stub_send_ret_submit() vulnerability to null transfer_buffer

commit be6123df1ea8f01ee2f896a16c2b7be3e4557a5a upstream.

stub_send_ret_submit() handles urb with a potential null transfer_buffer,
when it replays a packet with potential malicious data that could contain
a null buffer. Add a check for the condition when actual_length > 0 and
transfer_buffer is null.

Reported-by: Secunia Research <vuln@secunia.com>
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usbip: prevent vhci_hcd driver from leaking a socket pointer address

commit 2f2d0088eb93db5c649d2a5e34a3800a8a935fc5 upstream.

When a client has a USB device attached over IP, the vhci_hcd driver is
locally leaking a socket pointer address via the

/sys/devices/platform/vhci_hcd/status file (world-readable) and in debug
output when "usbip --debug port" is run.

Fix it to not leak. The socket pointer address is not used at the moment
and it was made visible as a convenient way to find IP address from socket
pointer address by looking up /proc/net/{tcp,tcp6}.

As this opens a security hole, the fix replaces socket pointer address with
sockfd.

Reported-by: Secunia Research <vuln@secunia.com>
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usbip: fix stub_rx: harden CMD_SUBMIT path to handle malicious input

commit c6688ef9f29762e65bce325ef4acd6c675806366 upstream.

Harden CMD_SUBMIT path to handle malicious input that could trigger
large memory allocations. Add checks to validate transfer_buffer_length
and number_of_packets to protect against bad input requesting for
unbounded memory allocations. Validate early in get_pipe() and return
failure.

Reported-by: Secunia Research <vuln@secunia.com>
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usbip: fix stub_rx: get_pipe() to validate endpoint number

commit 635f545a7e8be7596b9b2b6a43cab6bbd5a88e43 upstream.

get_pipe() routine doesn't validate the input endpoint number
and uses to reference ep_in and ep_out arrays. Invalid endpoint
number can trigger BUG(). Range check the epnum and returning
error instead of calling BUG().

Change caller stub_recv_cmd_submit() to handle the get_pipe()
error return.

Reported-by: Secunia Research <vuln@secunia.com>
Signed-off-by: Shuah Khan <shuahkh@osg.samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ovl: update ctx->pos on impure dir iteration

commit b02a16e6413a2f782e542ef60bad9ff6bf212f8a upstream.

This fixes a regression with readdir of impure dir in overlayfs
that is shared to VM via 9p fs.

Reported-by: Miguel Bernal Marin <miguel.bernal.marin@linux.intel.com>
Fixes: 4edb83bb1041 ("ovl: constant d_ino for non-merge dirs")
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Tested-by: Miguel Bernal Marin <miguel.bernal.marin@linux.intel.com>
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ovl: Pass ovl_get_nlink() parameters in right order

commit 08d8f8a5b094b66b29936e8751b4a818b8db1207 upstream.

Right now we seem to be passing index as "lowerdentry" and origin.dentry
as "upperdentry". IIUC, we should pass these parameters in reversed order
and this looks like a bug.

Signed-off-by: Vivek Goyal <vgoyal@redhat.com>
Acked-by: Amir Goldstein <amir73il@gmail.com>
Fixes: caf70cb2ba5d ("ovl: cleanup orphan index entries")
Signed-off-by: Miklos Szeredi <mszeredi@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: core: prevent malicious bNumInterfaces overflow

commit 48a4ff1c7bb5a32d2e396b03132d20d552c0eca7 upstream.

A malicious USB device with crafted descriptors can cause the kernel
to access unallocated memory by setting the bNumInterfaces value too
high in a configuration descriptor.  Although the value is adjusted
during parsing, this adjustment is skipped in one of the error return
paths.

This patch prevents the problem by setting bNumInterfaces to 0
initially.  The existing code already sets it to the proper value
after parsing is complete.

Signed-off-by: Alan Stern <stern@rowland.harvard.edu>
Reported-by: Andrey Konovalov <andreyknvl@google.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

USB: uas and storage: Add US_FL_BROKEN_FUA for another JMicron JMS567 ID

commit 62354454625741f0569c2cbe45b2d192f8fd258e upstream.

There is another JMS567-based USB3 UAS enclosure (152d:0578) that fails
with the following error:

[sda] tag#0 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE
[sda] tag#0 Sense Key : Illegal Request [current]
[sda] tag#0 Add. Sense: Invalid field in cdb

The issue occurs both with UAS (occasionally) and mass storage
(immediately after mounting a FS on a disk in the enclosure).

Enabling US_FL_BROKEN_FUA quirk solves this issue.

This patch adds an UNUSUAL_DEV with US_FL_BROKEN_FUA for the enclosure
for both UAS and mass storage.

Signed-off-by: David Kozub <zub@linux.fjfi.cvut.cz>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tracing: Allocate mask_str buffer dynamically

commit 90e406f96f630c07d631a021fd4af10aac913e77 upstream.

The default NR_CPUS can be very large, but actual possible nr_cpu_ids
usually is very small. For my x86 distribution, the NR_CPUS is 8192 and
nr_cpu_ids is 4. About 2 pages are wasted.

Most machines don't have so many CPUs, so define a array with NR_CPUS
just wastes memory. So let's allocate the buffer dynamically when need.

With this change, the mutext tracing_cpumask_update_lock also can be
removed now, which was used to protect mask_str.

Link: http://lkml.kernel.org/r/1512013183-19107-1-git-send-email-changbin.du@intel.com
Fixes: 36dfe9252bd4c ("ftrace: make use of tracing_cpumask")
Signed-off-by: Changbin Du <changbin.du@intel.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm, oom_reaper: fix memory corruption

commit 4837fe37adff1d159904f0c013471b1ecbcb455e upstream.

David Rientjes has reported the following memory corruption while the
oom reaper tries to unmap the victims address space

  BUG: Bad page map in process oom_reaper  pte:6353826300000000 pmd:00000000
  addr:00007f50cab1d000 vm_flags:08100073 anon_vma:ffff9eea335603f0 mapping:          (null) index:7f50cab1d
  file:          (null) fault:          (null) mmap:          (null) readpage:          (null)
  CPU: 2 PID: 1001 Comm: oom_reaper
  Call Trace:
     unmap_page_range+0x1068/0x1130
     __oom_reap_task_mm+0xd5/0x16b
     oom_reaper+0xff/0x14c
     kthread+0xc1/0xe0

Tetsuo Handa has noticed that the synchronization inside exit_mmap is
insufficient.  We only synchronize with the oom reaper if
tsk_is_oom_victim which is not true if the final __mmput is called from
a different context than the oom victim exit path.  This can trivially
happen from context of any task which has grabbed mm reference (e.g.  to
read /proc/<pid>/ file which requires mm etc.).

The race would look like this

  oom_reaper oom_victim task
mmget_not_zero
do_exit
  mmput
  __oom_reap_task_mm mmput
     __mmput
    exit_mmap
      remove_vma
    unmap_page_range

Fix this issue by providing a new mm_is_oom_victim() helper which
operates on the mm struct rather than a task.  Any context which
operates on a remote mm struct should use this helper in place of
tsk_is_oom_victim.  The flag is set in mark_oom_victim and never cleared
so it is stable in the exit_mmap path.

Debugged by Tetsuo Handa.

Link: http://lkml.kernel.org/r/20171210095130.17110-1-mhocko@kernel.org
Fixes: 212925802454 ("mm: oom: let oom_reap_task and exit_mmap run concurrently")
Signed-off-by: Michal Hocko <mhocko@suse.com>
Reported-by: David Rientjes <rientjes@google.com>
Acked-by: David Rientjes <rientjes@google.com>
Cc: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
Cc: Andrea Argangeli <andrea@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>

kernel: make groups_sort calling a responsibility group_info allocators

commit bdcf0a423ea1c40bbb40e7ee483b50fc8aa3d758 upstream.

In testing, we found that nfsd threads may call set_groups in parallel
for the same entry cached in auth.unix.gid, racing in the call of
groups_sort, corrupting the groups for that entry and leading to
permission denials for the client.

This patch:
 - Make groups_sort globally visible.
 - Move the call to groups_sort to the modifiers of group_info
 - Remove the call to groups_sort from set_groups

Link: http://lkml.kernel.org/r/20171211151420.18655-1-thiago.becker@gmail.com
Signed-off-by: Thiago Rafael Becker <thiago.becker@gmail.com>
Reviewed-by: Matthew Wilcox <mawilcox@microsoft.com>
Reviewed-by: NeilBrown <neilb@suse.com>
Acked-by: "J. Bruce Fields" <bfields@fieldses.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: Martin Schwidefsky <schwidefsky@de.ibm.com>
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>

autofs: fix careless error in recent commit

commit 302ec300ef8a545a7fc7f667e5fd743b091c2eeb upstream.

Commit ecc0c469f277 ("autofs: don't fail mount for transient error") was
meant to replace an 'if' with a 'switch', but instead added the 'switch'
leaving the case in place.

Link: http://lkml.kernel.org/r/87zi6wstmw.fsf@notabene.neil.brown.name
Fixes: ecc0c469f277 ("autofs: don't fail mount for transient error")
Reported-by: Ben Hutchings <ben.hutchings@codethink.co.uk>
Signed-off-by: NeilBrown <neilb@suse.com>
Cc: Ian Kent <raven@themaw.net>
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>

string.h: workaround for increased stack usage

commit 146734b091430c80d80bb96b1139a96fb4bc830e upstream.

The hardened strlen() function causes rather large stack usage in at
least one file in the kernel, in particular when CONFIG_KASAN is
enabled:

  drivers/media/usb/em28xx/em28xx-dvb.c: In function 'em28xx_dvb_init':
  drivers/media/usb/em28xx/em28xx-dvb.c:2062:1: error: the frame size of 3256 bytes is larger than 204 bytes [-Werror=frame-larger-than=]

Analyzing this problem led to the discovery that gcc fails to merge the
stack slots for the i2c_board_info[] structures after we strlcpy() into
them, due to the 'noreturn' attribute on the source string length check.

I reported this as a gcc bug, but it is unlikely to get fixed for gcc-8,
since it is relatively easy to work around, and it gets triggered
rarely.  An earlier workaround I did added an empty inline assembly
statement before the call to fortify_panic(), which works surprisingly
well, but is really ugly and unintuitive.

This is a new approach to the same problem, this time addressing it by
not calling the 'extern __real_strnlen()' function for string constants
where __builtin_strlen() is a compile-time constant and therefore known
to be safe.

We do this by checking if the last character in the string is a
compile-time constant '\0'.  If it is, we can assume that strlen() of
the string is also constant.

As a side-effect, this should also improve the object code output for
any other call of strlen() on a string constant.

[akpm@linux-foundation.org: add comment]
Link: http://lkml.kernel.org/r/20171205215143.3085755-1-arnd@arndb.de
Link: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=82365
Link: https://patchwork.kernel.org/patch/9980413/
Link: https://patchwork.kernel.org/patch/9974047/
Fixes: 6974f0c4555 ("include/linux/string.h: add the option of fortified string.h functions")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Cc: Kees Cook <keescook@chromium.org>
Cc: Mauro Carvalho Chehab <mchehab@kernel.org>
Cc: Dmitry Vyukov <dvyukov@google.com>
Cc: Alexander Potapenko <glider@google.com>
Cc: Andrey Ryabinin <aryabinin@virtuozzo.com>
Cc: Daniel Micay <danielmicay@gmail.com>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Martin Wilck <mwilck@suse.com>
Cc: Dan Williams <dan.j.williams@intel.com>
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>

cifs: fix NULL deref in SMB2_read

commit a821df3f1af72aa6a0d573eea94a7dd2613e9f4e upstream.

Signed-off-by: Ronnie Sahlberg <lsahlber@redhat.com>
Reviewed-by: Pavel Shilovsky <pshilov@microsoft.com>
Signed-off-by: Steve French <smfrench@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

crypto: af_alg - fix NULL pointer dereference in

commit 887207ed9e5812ed9239b6d07185a2d35dda91db upstream.

 af_alg_free_areq_sgls()

If allocating the ->tsgl member of 'struct af_alg_async_req' failed,
during cleanup we dereferenced the NULL ->tsgl pointer in
af_alg_free_areq_sgls(), because ->tsgl_entries was nonzero.

Fix it by only freeing the ->tsgl list if it is non-NULL.

This affected both algif_skcipher and algif_aead.

Fixes: e870456d8e7c ("crypto: algif_skcipher - overhaul memory management")
Fixes: d887c52d6ae4 ("crypto: algif_aead - overhaul memory management")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

crypto: salsa20 - fix blkcipher_walk API usage

commit ecaaab5649781c5a0effdaf298a925063020500e upstream.

When asked to encrypt or decrypt 0 bytes, both the generic and x86
implementations of Salsa20 crash in blkcipher_walk_done(), either when
doing 'kfree(walk->buffer)' or 'free_page((unsigned long)walk->page)',
because walk->buffer and walk->page have not been initialized.

The bug is that Salsa20 is calling blkcipher_walk_done() even when
nothing is in 'walk.nbytes'.  But blkcipher_walk_done() is only meant to
be called when a nonzero number of bytes have been provided.

The broken code is part of an optimization that tries to make only one
call to salsa20_encrypt_bytes() to process inputs that are not evenly
divisible by 64 bytes.  To fix the bug, just remove this "optimization"
and use the blkcipher_walk API the same way all the other users do.

Reproducer:

    #include <linux/if_alg.h>
    #include <sys/socket.h>
    #include <unistd.h>

    int main()
    {
            int algfd, reqfd;
            struct sockaddr_alg addr = {
                    .salg_type = "skcipher",
                    .salg_name = "salsa20",
            };
            char key[16] = { 0 };

            algfd = socket(AF_ALG, SOCK_SEQPACKET, 0);
            bind(algfd, (void *)&addr, sizeof(addr));
            reqfd = accept(algfd, 0, 0);
            setsockopt(algfd, SOL_ALG, ALG_SET_KEY, key, sizeof(key));
            read(reqfd, key, sizeof(key));
    }

Reported-by: syzbot <syzkaller@googlegroups.com>
Fixes: eb6f13eb9f81 ("[CRYPTO] salsa20_generic: Fix multi-page processing")
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

crypto: hmac - require that the underlying hash algorithm is unkeyed

commit af3ff8045bbf3e32f1a448542e73abb4c8ceb6f1 upstream.

Because the HMAC template didn't check that its underlying hash
algorithm is unkeyed, trying to use "hmac(hmac(sha3-512-generic))"
through AF_ALG or through KEYCTL_DH_COMPUTE resulted in the inner HMAC
being used without having been keyed, resulting in sha3_update() being
called without sha3_init(), causing a stack buffer overflow.

This is a very old bug, but it seems to have only started causing real
problems when SHA-3 support was added (requires CONFIG_CRYPTO_SHA3)
because the innermost hash's state is ->import()ed from a zeroed buffer,
and it just so happens that other hash algorithms are fine with that,
but SHA-3 is not.  However, there could be arch or hardware-dependent
hash algorithms also affected; I couldn't test everything.

Fix the bug by introducing a function crypto_shash_alg_has_setkey()
which tests whether a shash algorithm is keyed.  Then update the HMAC
template to require that its underlying hash algorithm is unkeyed.

Here is a reproducer:

    #include <linux/if_alg.h>
    #include <sys/socket.h>

    int main()
    {
        int algfd;
        struct sockaddr_alg addr = {
            .salg_type = "hash",
            .salg_name = "hmac(hmac(sha3-512-generic))",
        };
        char key[4096] = { 0 };

        algfd = socket(AF_ALG, SOCK_SEQPACKET, 0);
        bind(algfd, (const struct sockaddr *)&addr, sizeof(addr));
        setsockopt(algfd, SOL_ALG, ALG_SET_KEY, key, sizeof(key));
    }

Here was the KASAN report from syzbot:

    BUG: KASAN: stack-out-of-bounds in memcpy include/linux/string.h:341  [inline]
    BUG: KASAN: stack-out-of-bounds in sha3_update+0xdf/0x2e0  crypto/sha3_generic.c:161
    Write of size 4096 at addr ffff8801cca07c40 by task syzkaller076574/3044

    CPU: 1 PID: 3044 Comm: syzkaller076574 Not tainted 4.14.0-mm1+ #25
    Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS  Google 01/01/2011
    Call Trace:
      __dump_stack lib/dump_stack.c:17 [inline]
      dump_stack+0x194/0x257 lib/dump_stack.c:53
      print_address_description+0x73/0x250 mm/kasan/report.c:252
      kasan_report_error mm/kasan/report.c:351 [inline]
      kasan_report+0x25b/0x340 mm/kasan/report.c:409
      check_memory_region_inline mm/kasan/kasan.c:260 [inline]
      check_memory_region+0x137/0x190 mm/kasan/kasan.c:267
      memcpy+0x37/0x50 mm/kasan/kasan.c:303
      memcpy include/linux/string.h:341 [inline]
      sha3_update+0xdf/0x2e0 crypto/sha3_generic.c:161
      crypto_shash_update+0xcb/0x220 crypto/shash.c:109
      shash_finup_unaligned+0x2a/0x60 crypto/shash.c:151
      crypto_shash_finup+0xc4/0x120 crypto/shash.c:165
      hmac_finup+0x182/0x330 crypto/hmac.c:152
      crypto_shash_finup+0xc4/0x120 crypto/shash.c:165
      shash_digest_unaligned+0x9e/0xd0 crypto/shash.c:172
      crypto_shash_digest+0xc4/0x120 crypto/shash.c:186
      hmac_setkey+0x36a/0x690 crypto/hmac.c:66
      crypto_shash_setkey+0xad/0x190 crypto/shash.c:64
      shash_async_setkey+0x47/0x60 crypto/shash.c:207
      crypto_ahash_setkey+0xaf/0x180 crypto/ahash.c:200
      hash_setkey+0x40/0x90 crypto/algif_hash.c:446
      alg_setkey crypto/af_alg.c:221 [inline]
      alg_setsockopt+0x2a1/0x350 crypto/af_alg.c:254
      SYSC_setsockopt net/socket.c:1851 [inline]
      SyS_setsockopt+0x189/0x360 net/socket.c:1830
      entry_SYSCALL_64_fastpath+0x1f/0x96

Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

crypto: rsa - fix buffer overread when stripping leading zeroes

commit d2890c3778b164fde587bc16583f3a1c87233ec5 upstream.

In rsa_get_n(), if the buffer contained all 0's and "FIPS mode" is
enabled, we would read one byte past the end of the buffer while
scanning the leading zeroes.  Fix it by checking 'n_sz' before '!*ptr'.

This bug was reachable by adding a specially crafted key of type
"asymmetric" (requires CONFIG_RSA and CONFIG_X509_CERTIFICATE_PARSER).

KASAN report:

    BUG: KASAN: slab-out-of-bounds in rsa_get_n+0x19e/0x1d0 crypto/rsa_helper.c:33
    Read of size 1 at addr ffff88003501a708 by task keyctl/196

    CPU: 1 PID: 196 Comm: keyctl Not tainted 4.14.0-09238-g1d3b78bbc6e9 #26
    Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.11.0-20171110_100015-anatol 04/01/2014
    Call Trace:
     rsa_get_n+0x19e/0x1d0 crypto/rsa_helper.c:33
     asn1_ber_decoder+0x82a/0x1fd0 lib/asn1_decoder.c:328
     rsa_set_pub_key+0xd3/0x320 crypto/rsa.c:278
     crypto_akcipher_set_pub_key ./include/crypto/akcipher.h:364 [inline]
     pkcs1pad_set_pub_key+0xae/0x200 crypto/rsa-pkcs1pad.c:117
     crypto_akcipher_set_pub_key ./include/crypto/akcipher.h:364 [inline]
     public_key_verify_signature+0x270/0x9d0 crypto/asymmetric_keys/public_key.c:106
     x509_check_for_self_signed+0x2ea/0x480 crypto/asymmetric_keys/x509_public_key.c:141
     x509_cert_parse+0x46a/0x620 crypto/asymmetric_keys/x509_cert_parser.c:129
     x509_key_preparse+0x61/0x750 crypto/asymmetric_keys/x509_public_key.c:174
     asymmetric_key_preparse+0xa4/0x150 crypto/asymmetric_keys/asymmetric_type.c:388
     key_create_or_update+0x4d4/0x10a0 security/keys/key.c:850
     SYSC_add_key security/keys/keyctl.c:122 [inline]
     SyS_add_key+0xe8/0x290 security/keys/keyctl.c:62
     entry_SYSCALL_64_fastpath+0x1f/0x96

    Allocated by task 196:
     __do_kmalloc mm/slab.c:3711 [inline]
     __kmalloc_track_caller+0x118/0x2e0 mm/slab.c:3726
     kmemdup+0x17/0x40 mm/util.c:118
     kmemdup ./include/linux/string.h:414 [inline]
     x509_cert_parse+0x2cb/0x620 crypto/asymmetric_keys/x509_cert_parser.c:106
     x509_key_preparse+0x61/0x750 crypto/asymmetric_keys/x509_public_key.c:174
     asymmetric_key_preparse+0xa4/0x150 crypto/asymmetric_keys/asymmetric_type.c:388
     key_create_or_update+0x4d4/0x10a0 security/keys/key.c:850
     SYSC_add_key security/keys/keyctl.c:122 [inline]
     SyS_add_key+0xe8/0x290 security/keys/keyctl.c:62
     entry_SYSCALL_64_fastpath+0x1f/0x96

Fixes: 5a7de97309f5 ("crypto: rsa - return raw integers for the ASN.1 parser")
Cc: Tudor Ambarus <tudor-dan.ambarus@nxp.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: James Morris <james.l.morris@oracle.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

crypto: algif_aead - fix reference counting of null skcipher

commit b32a7dc8aef1882fbf983eb354837488cc9d54dc upstream.

In the AEAD interface for AF_ALG, the reference to the "null skcipher"
held by each tfm was being dropped in the wrong place -- when each
af_alg_ctx was freed instead of when the aead_tfm was freed.  As
discovered by syzkaller, a specially crafted program could use this to
cause the null skcipher to be freed while it is still in use.

Fix it by dropping the reference in the right place.

Fixes: 72548b093ee3 ("crypto: algif_aead - copy AAD from src to dst")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Eric Biggers <ebiggers@google.com>
Reviewed-by: Stephan Mueller <smueller@chronox.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mfd: fsl-imx25: Clean up irq settings during removal

commit 18f77393796848e68909e65d692c1d1436f06e06 upstream.

When fsl-imx25-tsadc is compiled as a module, loading, unloading and
reloading the module will lead to a crash.

Unable to handle kernel paging request at virtual address bf005430
[<c004df6c>] (irq_find_matching_fwspec)
   from [<c028d5ec>] (of_irq_get+0x58/0x74)
[<c028d594>] (of_irq_get)
   from [<c01ff970>] (platform_get_irq+0x48/0xc8)
[<c01ff928>] (platform_get_irq)
   from [<bf00e33c>] (mx25_tsadc_probe+0x220/0x2f4 [fsl_imx25_tsadc])

irq_find_matching_fwspec() loops over all registered irq domains. The
irq domain is still registered from last time the module was loaded but
the pointer to its operations is invalid after the module was unloaded.

Add a removal function which clears the irq handler and removes the irq
domain. With this cleanup in place, it's possible to unload and reload
the module.

Signed-off-by: Martin Kaiser <martin@kaiser.cx>
Reviewed-by: Lucas Stach <l.stach@pengutronix.de>
Signed-off-by: Lee Jones <lee.jones@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 4.14.7

dvb_frontend: don't use-after-free the frontend struct

commit b1cb7372fa822af6c06c8045963571d13ad6348b upstream.

dvb_frontend_invoke_release() may free the frontend struct.
So, the free logic can't update it anymore after calling it.

That's OK, as __dvb_frontend_free() is called only when the
krefs are zeroed, so nobody is using it anymore.

That should fix the following KASAN error:

The KASAN report looks like this (running on kernel 3e0cc09a3a2c40ec1ffb6b4e12da86e98feccb11 (4.14-rc5+)):
==================================================================
BUG: KASAN: use-after-free in __dvb_frontend_free+0x113/0x120
Write of size 8 at addr ffff880067d45a00 by task kworker/0:1/24

CPU: 0 PID: 24 Comm: kworker/0:1 Not tainted 4.14.0-rc5-43687-g06ab8a23e0e6 #545
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
Workqueue: usb_hub_wq hub_event
Call Trace:
 __dump_stack lib/dump_stack.c:16
 dump_stack+0x292/0x395 lib/dump_stack.c:52
 print_address_description+0x78/0x280 mm/kasan/report.c:252
 kasan_report_error mm/kasan/report.c:351
 kasan_report+0x23d/0x350 mm/kasan/report.c:409
 __asan_report_store8_noabort+0x1c/0x20 mm/kasan/report.c:435
 __dvb_frontend_free+0x113/0x120 drivers/media/dvb-core/dvb_frontend.c:156
 dvb_frontend_put+0x59/0x70 drivers/media/dvb-core/dvb_frontend.c:176
 dvb_frontend_detach+0x120/0x150 drivers/media/dvb-core/dvb_frontend.c:2803
 dvb_usb_adapter_frontend_exit+0xd6/0x160 drivers/media/usb/dvb-usb/dvb-usb-dvb.c:340
 dvb_usb_adapter_exit drivers/media/usb/dvb-usb/dvb-usb-init.c:116
 dvb_usb_exit+0x9b/0x200 drivers/media/usb/dvb-usb/dvb-usb-init.c:132
 dvb_usb_device_exit+0xa5/0xf0 drivers/media/usb/dvb-usb/dvb-usb-init.c:295
 usb_unbind_interface+0x21c/0xa90 drivers/usb/core/driver.c:423
 __device_release_driver drivers/base/dd.c:861
 device_release_driver_internal+0x4f1/0x5c0 drivers/base/dd.c:893
 device_release_driver+0x1e/0x30 drivers/base/dd.c:918
 bus_remove_device+0x2f4/0x4b0 drivers/base/bus.c:565
 device_del+0x5c4/0xab0 drivers/base/core.c:1985
 usb_disable_device+0x1e9/0x680 drivers/usb/core/message.c:1170
 usb_disconnect+0x260/0x7a0 drivers/usb/core/hub.c:2124
 hub_port_connect drivers/usb/core/hub.c:4754
 hub_port_connect_change drivers/usb/core/hub.c:5009
 port_event drivers/usb/core/hub.c:5115
 hub_event+0x1318/0x3740 drivers/usb/core/hub.c:5195
 process_one_work+0xc73/0x1d90 kernel/workqueue.c:2119
 worker_thread+0x221/0x1850 kernel/workqueue.c:2253
 kthread+0x363/0x440 kernel/kthread.c:231
 ret_from_fork+0x2a/0x40 arch/x86/entry/entry_64.S:431

Allocated by task 24:
 save_stack_trace+0x1b/0x20 arch/x86/kernel/stacktrace.c:59
 save_stack+0x43/0xd0 mm/kasan/kasan.c:447
 set_track mm/kasan/kasan.c:459
 kasan_kmalloc+0xad/0xe0 mm/kasan/kasan.c:551
 kmem_cache_alloc_trace+0x11e/0x2d0 mm/slub.c:2772
 kmalloc ./include/linux/slab.h:493
 kzalloc ./include/linux/slab.h:666
 dtt200u_fe_attach+0x4c/0x110 drivers/media/usb/dvb-usb/dtt200u-fe.c:212
 dtt200u_frontend_attach+0x35/0x80 drivers/media/usb/dvb-usb/dtt200u.c:136
 dvb_usb_adapter_frontend_init+0x32b/0x660 drivers/media/usb/dvb-usb/dvb-usb-dvb.c:286
 dvb_usb_adapter_init drivers/media/usb/dvb-usb/dvb-usb-init.c:86
 dvb_usb_init drivers/media/usb/dvb-usb/dvb-usb-init.c:162
 dvb_usb_device_init+0xf73/0x17f0 drivers/media/usb/dvb-usb/dvb-usb-init.c:277
 dtt200u_usb_probe+0xa1/0xe0 drivers/media/usb/dvb-usb/dtt200u.c:155
 usb_probe_interface+0x35d/0x8e0 drivers/usb/core/driver.c:361
 really_probe drivers/base/dd.c:413
 driver_probe_device+0x610/0xa00 drivers/base/dd.c:557
 __device_attach_driver+0x230/0x290 drivers/base/dd.c:653
 bus_for_each_drv+0x161/0x210 drivers/base/bus.c:463
 __device_attach+0x26b/0x3c0 drivers/base/dd.c:710
 device_initial_probe+0x1f/0x30 drivers/base/dd.c:757
 bus_probe_device+0x1eb/0x290 drivers/base/bus.c:523
 device_add+0xd0b/0x1660 drivers/base/core.c:1835
 usb_set_configuration+0x104e/0x1870 drivers/usb/core/message.c:1932
 generic_probe+0x73/0xe0 drivers/usb/core/generic.c:174
 usb_probe_device+0xaf/0xe0 drivers/usb/core/driver.c:266
 really_probe drivers/base/dd.c:413
 driver_probe_device+0x610/0xa00 drivers/base/dd.c:557
 __device_attach_driver+0x230/0x290 drivers/base/dd.c:653
 bus_for_each_drv+0x161/0x210 drivers/base/bus.c:463
 __device_attach+0x26b/0x3c0 drivers/base/dd.c:710
 device_initial_probe+0x1f/0x30 drivers/base/dd.c:757
 bus_probe_device+0x1eb/0x290 drivers/base/bus.c:523
 device_add+0xd0b/0x1660 drivers/base/core.c:1835
 usb_new_device+0x7b8/0x1020 drivers/usb/core/hub.c:2457
 hub_port_connect drivers/usb/core/hub.c:4903
 hub_port_connect_change drivers/usb/core/hub.c:5009
 port_event drivers/usb/core/hub.c:5115
 hub_event+0x194d/0x3740 drivers/usb/core/hub.c:5195
 process_one_work+0xc73/0x1d90 kernel/workqueue.c:2119
 worker_thread+0x221/0x1850 kernel/workqueue.c:2253
 kthread+0x363/0x440 kernel/kthread.c:231
 ret_from_fork+0x2a/0x40 arch/x86/entry/entry_64.S:431

Freed by task 24:
 save_stack_trace+0x1b/0x20 arch/x86/kernel/stacktrace.c:59
 save_stack+0x43/0xd0 mm/kasan/kasan.c:447
 set_track mm/kasan/kasan.c:459
 kasan_slab_free+0x72/0xc0 mm/kasan/kasan.c:524
 slab_free_hook mm/slub.c:1390
 slab_free_freelist_hook mm/slub.c:1412
 slab_free mm/slub.c:2988
 kfree+0xf6/0x2f0 mm/slub.c:3919
 dtt200u_fe_release+0x3c/0x50 drivers/media/usb/dvb-usb/dtt200u-fe.c:202
 dvb_frontend_invoke_release.part.13+0x1c/0x30 drivers/media/dvb-core/dvb_frontend.c:2790
 dvb_frontend_invoke_release drivers/media/dvb-core/dvb_frontend.c:2789
 __dvb_frontend_free+0xad/0x120 drivers/media/dvb-core/dvb_frontend.c:153
 dvb_frontend_put+0x59/0x70 drivers/media/dvb-core/dvb_frontend.c:176
 dvb_frontend_detach+0x120/0x150 drivers/media/dvb-core/dvb_frontend.c:2803
 dvb_usb_adapter_frontend_exit+0xd6/0x160 drivers/media/usb/dvb-usb/dvb-usb-dvb.c:340
 dvb_usb_adapter_exit drivers/media/usb/dvb-usb/dvb-usb-init.c:116
 dvb_usb_exit+0x9b/0x200 drivers/media/usb/dvb-usb/dvb-usb-init.c:132
 dvb_usb_device_exit+0xa5/0xf0 drivers/media/usb/dvb-usb/dvb-usb-init.c:295
 usb_unbind_interface+0x21c/0xa90 drivers/usb/core/driver.c:423
 __device_release_driver drivers/base/dd.c:861
 device_release_driver_internal+0x4f1/0x5c0 drivers/base/dd.c:893
 device_release_driver+0x1e/0x30 drivers/base/dd.c:918
 bus_remove_device+0x2f4/0x4b0 drivers/base/bus.c:565
 device_del+0x5c4/0xab0 drivers/base/core.c:1985
 usb_disable_device+0x1e9/0x680 drivers/usb/core/message.c:1170
 usb_disconnect+0x260/0x7a0 drivers/usb/core/hub.c:2124
 hub_port_connect drivers/usb/core/hub.c:4754
 hub_port_connect_change drivers/usb/core/hub.c:5009
 port_event drivers/usb/core/hub.c:5115
 hub_event+0x1318/0x3740 drivers/usb/core/hub.c:5195
 process_one_work+0xc73/0x1d90 kernel/workqueue.c:2119
 worker_thread+0x221/0x1850 kernel/workqueue.c:2253
 kthread+0x363/0x440 kernel/kthread.c:231
 ret_from_fork+0x2a/0x40 arch/x86/entry/entry_64.S:431

The buggy address belongs to the object at ffff880067d45500
 which belongs to the cache kmalloc-2048 of size 2048
The buggy address is located 1280 bytes inside of
 2048-byte region [ffff880067d45500, ffff880067d45d00)
The buggy address belongs to the page:
page:ffffea00019f5000 count:1 mapcount:0 mapping:          (null)
index:0x0 compound_mapcount: 0
flags: 0x100000000008100(slab|head)
raw: 0100000000008100 0000000000000000 0000000000000000 00000001000f000f
raw: dead000000000100 dead000000000200 ffff88006c002d80 0000000000000000
page dumped because: kasan: bad access detected

Memory state around the buggy address:
 ffff880067d45900: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
 ffff880067d45980: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
 ffff880067d45a00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
                   ^
 ffff880067d45a80: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
 ffff880067d45b00: fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb fb
==================================================================

Fixes: ead666000a5f ("media: dvb_frontend: only use kref after initialized")

Reported-by: Andrey Konovalov <andreyknvl@google.com>
Suggested-by: Matthias Schwarzott <zzam@gentoo.org>
Tested-by: Andrey Konovalov <andreyknvl@google.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

media: dvb-core: always call invoke_release() in fe_free()

commit 62229de19ff2b7f3e0ebf4d48ad99061127d0281 upstream.

Follow-up to: ead666000a5f ("media: dvb_frontend: only use kref after initialized")

The aforementioned commit fixed refcount OOPSes when demod driver attaching
succeeded but tuner driver didn't. However, the use count of the attached
demod drivers don't go back to zero and thus couldn't be cleanly unloaded.
Improve on this by calling dvb_frontend_invoke_release() in
__dvb_frontend_free() regardless of fepriv being NULL, instead of returning
when fepriv is NULL. This is safe to do since _invoke_release() will check
for passed pointers being valid before calling the .release() function.

[mchehab@s-opensource.com: changed the logic a little bit to reduce
 conflicts with another bug fix patch under review]
Fixes: ead666000a5f ("media: dvb_frontend: only use kref after initialized")
Signed-off-by: Daniel Scheller <d.scheller@gmx.net>
Signed-off-by: Mauro Carvalho Chehab <mchehab@s-opensource.com>
Cc: Guenter Roeck <linux@roeck-us.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

x86/intel_rdt: Fix potential deadlock during resctrl unmount

[ Upstream commit 36b6f9fcb8928c06b6638a4cf91bc9d69bb49aa2 ]

Lockdep warns about a potential deadlock:

[   66.782842] ======================================================
[   66.782888] WARNING: possible circular locking dependency detected
[   66.782937] 4.14.0-rc2-test-test+ #48 Not tainted
[   66.782983] ------------------------------------------------------
[   66.783052] umount/336 is trying to acquire lock:
[   66.783117]  (cpu_hotplug_lock.rw_sem){++++}, at: [<ffffffff81032395>] rdt_kill_sb+0x215/0x390
[   66.783193]
               but task is already holding lock:
[   66.783244]  (rdtgroup_mutex){+.+.}, at: [<ffffffff810321b6>] rdt_kill_sb+0x36/0x390
[   66.783305]
               which lock already depends on the new lock.

[   66.783364]
               the existing dependency chain (in reverse order) is:
[   66.783419]
               -> #3 (rdtgroup_mutex){+.+.}:
[   66.783467]        __lock_acquire+0x1293/0x13f0
[   66.783509]        lock_acquire+0xaf/0x220
[   66.783543]        __mutex_lock+0x71/0x9b0
[   66.783575]        mutex_lock_nested+0x1b/0x20
[   66.783610]        intel_rdt_online_cpu+0x3b/0x430
[   66.783649]        cpuhp_invoke_callback+0xab/0x8e0
[   66.783687]        cpuhp_thread_fun+0x7a/0x150
[   66.783722]        smpboot_thread_fn+0x1cc/0x270
[   66.783764]        kthread+0x16e/0x190
[   66.783794]        ret_from_fork+0x27/0x40
[   66.783825]
               -> #2 (cpuhp_state){+.+.}:
[   66.783870]        __lock_acquire+0x1293/0x13f0
[   66.783906]        lock_acquire+0xaf/0x220
[   66.783938]        cpuhp_issue_call+0x102/0x170
[   66.783974]        __cpuhp_setup_state_cpuslocked+0x154/0x2a0
[   66.784023]        __cpuhp_setup_state+0xc7/0x170
[   66.784061]        page_writeback_init+0x43/0x67
[   66.784097]        pagecache_init+0x43/0x4a
[   66.784131]        start_kernel+0x3ad/0x3f7
[   66.784165]        x86_64_start_reservations+0x2a/0x2c
[   66.784204]        x86_64_start_kernel+0x72/0x75
[   66.784241]        verify_cpu+0x0/0xfb
[   66.784270]
               -> #1 (cpuhp_state_mutex){+.+.}:
[   66.784319]        __lock_acquire+0x1293/0x13f0
[   66.784355]        lock_acquire+0xaf/0x220
[   66.784387]        __mutex_lock+0x71/0x9b0
[   66.784419]        mutex_lock_nested+0x1b/0x20
[   66.784454]        __cpuhp_setup_state_cpuslocked+0x52/0x2a0
[   66.784497]        __cpuhp_setup_state+0xc7/0x170
[   66.784535]        page_alloc_init+0x28/0x30
[   66.784569]        start_kernel+0x148/0x3f7
[   66.784602]        x86_64_start_reservations+0x2a/0x2c
[   66.784642]        x86_64_start_kernel+0x72/0x75
[   66.784678]        verify_cpu+0x0/0xfb
[   66.784707]
               -> #0 (cpu_hotplug_lock.rw_sem){++++}:
[   66.784759]        check_prev_add+0x32f/0x6e0
[   66.784794]        __lock_acquire+0x1293/0x13f0
[   66.784830]        lock_acquire+0xaf/0x220
[   66.784863]        cpus_read_lock+0x3d/0xb0
[   66.784896]        rdt_kill_sb+0x215/0x390
[   66.784930]        deactivate_locked_super+0x3e/0x70
[   66.784968]        deactivate_super+0x40/0x60
[   66.785003]        cleanup_mnt+0x3f/0x80
[   66.785034]        __cleanup_mnt+0x12/0x20
[   66.785070]        task_work_run+0x8b/0xc0
[   66.785103]        exit_to_usermode_loop+0x94/0xa0
[   66.786804]        syscall_return_slowpath+0xe8/0x150
[   66.788502]        entry_SYSCALL_64_fastpath+0xab/0xad
[   66.790194]
               other info that might help us debug this:

[   66.795139] Chain exists of:
                 cpu_hotplug_lock.rw_sem --> cpuhp_state --> rdtgroup_mutex

[   66.800035]  Possible unsafe locking scenario:

[   66.803267]        CPU0                    CPU1
[   66.804867]        ----                    ----
[   66.806443]   lock(rdtgroup_mutex);
[   66.808002]                                lock(cpuhp_state);
[   66.809565]                                lock(rdtgroup_mutex);
[   66.811110]   lock(cpu_hotplug_lock.rw_sem);
[   66.812608]
                *** DEADLOCK ***

[   66.816983] 2 locks held by umount/336:
[   66.818418]  #0:  (&type->s_umount_key#35){+.+.}, at: [<ffffffff81229738>] deactivate_super+0x38/0x60
[   66.819922]  #1:  (rdtgroup_mutex){+.+.}, at: [<ffffffff810321b6>] rdt_kill_sb+0x36/0x390

When the resctrl filesystem is unmounted the locks should be obtain in the
locks in the same order as was done when the cpus came online:

      cpu_hotplug_lock before rdtgroup_mutex.

This also requires to switch the static_branch_disable() calls to the
_cpulocked variant because now cpu hotplug lock is held already.

[ tglx: Switched to cpus_read_[un]lock ]

Signed-off-by: Reinette Chatre <reinette.chatre@intel.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Tested-by: Sai Praneeth Prakhya <sai.praneeth.prakhya@intel.com>
Acked-by: Vikas Shivappa <vikas.shivappa@linux.intel.com>
Acked-by: Fenghua Yu <fenghua.yu@intel.com>
Acked-by: Tony Luck <tony.luck@intel.com>
Link: https://lkml.kernel.org/r/cc292e76be073f7260604651711c47b09fd0dc81.1508490116.git.reinette.chatre@intel.com
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

RDMA/cxgb4: Annotate r2 and stag as __be32

[ Upstream commit 7d7d065a5eec7e218174d5c64a9f53f99ffdb119 ]

Chelsio cxgb4 HW is big-endian, hence there is need to properly
annotate r2 and stag fields as __be32 and not __u32 to fix the
following sparse warnings.

  drivers/infiniband/hw/cxgb4/qp.c:614:16:
    warning: incorrect type in assignment (different base types)
      expected unsigned int [unsigned] [usertype] r2
      got restricted __be32 [usertype] <noident>
  drivers/infiniband/hw/cxgb4/qp.c:615:18:
    warning: incorrect type in assignment (different base types)
      expected unsigned int [unsigned] [usertype] stag
      got restricted __be32 [usertype] <noident>

Cc: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

md: free unused memory after bitmap resize

[ Upstream commit 0868b99c214a3d55486c700de7c3f770b7243e7c ]

When bitmap is resized, the old kalloced chunks just are not released
once the resized bitmap starts to use new space.

This fixes in particular kmemleak reports like this one:

unreferenced object 0xffff8f4311e9c000 (size 4096):
  comm "lvm", pid 19333, jiffies 4295263268 (age 528.265s)
  hex dump (first 32 bytes):
    02 80 02 80 02 80 02 80 02 80 02 80 02 80 02 80  ................
    02 80 02 80 02 80 02 80 02 80 02 80 02 80 02 80  ................
  backtrace:
    [<ffffffffa69471ca>] kmemleak_alloc+0x4a/0xa0
    [<ffffffffa628c10e>] kmem_cache_alloc_trace+0x14e/0x2e0
    [<ffffffffa676cfec>] bitmap_checkpage+0x7c/0x110
    [<ffffffffa676d0c5>] bitmap_get_counter+0x45/0xd0
    [<ffffffffa676d6b3>] bitmap_set_memory_bits+0x43/0xe0
    [<ffffffffa676e41c>] bitmap_init_from_disk+0x23c/0x530
    [<ffffffffa676f1ae>] bitmap_load+0xbe/0x160
    [<ffffffffc04c47d3>] raid_preresume+0x203/0x2f0 [dm_raid]
    [<ffffffffa677762f>] dm_table_resume_targets+0x4f/0xe0
    [<ffffffffa6774b52>] dm_resume+0x122/0x140
    [<ffffffffa6779b9f>] dev_suspend+0x18f/0x290
    [<ffffffffa677a3a7>] ctl_ioctl+0x287/0x560
    [<ffffffffa677a693>] dm_ctl_ioctl+0x13/0x20
    [<ffffffffa62d6b46>] do_vfs_ioctl+0xa6/0x750
    [<ffffffffa62d7269>] SyS_ioctl+0x79/0x90
    [<ffffffffa6956d41>] entry_SYSCALL_64_fastpath+0x1f/0xc2

Signed-off-by: Zdenek Kabelac <zkabelac@redhat.com>
Signed-off-by: Shaohua Li <shli@fb.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

dm raid: fix panic when attempting to force a raid to sync

[ Upstream commit 233978449074ca7e45d9c959f9ec612d1b852893 ]

Requesting a sync on an active raid device via a table reload
(see 'sync' parameter in Documentation/device-mapper/dm-raid.txt)
skips the super_load() call that defines the superblock size
(rdev->sb_size) -- resulting in an oops if/when super_sync()->memset()
is called.

Fix by moving the initialization of the superblock start and size
out of super_load() to the caller (analyse_superblocks).

Signed-off-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

audit: ensure that 'audit=1' actually enables audit for PID 1

[ Upstream commit 173743dd99a49c956b124a74c8aacb0384739a4c ]

Prior to this patch we enabled audit in audit_init(), which is too
late for PID 1 as the standard initcalls are run after the PID 1 task
is forked.  This means that we never allocate an audit_context (see
audit_alloc()) for PID 1 and therefore miss a lot of audit events
generated by PID 1.

This patch enables audit as early as possible to help ensure that when
PID 1 is forked it can allocate an audit_context if required.

Reviewed-by: Richard Guy Briggs <rgb@redhat.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

audit: Allow auditd to set pid to 0 to end auditing

[ Upstream commit 33e8a907804428109ce1d12301c3365d619cc4df ]

The API to end auditing has historically been for auditd to set the
pid to 0. This patch restores that functionality.

See: https://github.com/linux-audit/audit-kernel/issues/69

Reviewed-by: Richard Guy Briggs <rgb@redhat.com>
Signed-off-by: Steve Grubb <sgrubb@redhat.com>
Signed-off-by: Paul Moore <paul@paul-moore.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

nvmet-rdma: update queue list during ib_device removal

[ Upstream commit 43b92fd27aaef0f529c9321cfebbaec1d7b8f503 ]

A NULL deref happens when nvmet_rdma_remove_one() is called more than once
(e.g. while connected via 2 ports).
The first call frees the queues related to the first ib_device but
doesn't remove them from the queue list.
While calling nvmet_rdma_remove_one() for the second ib_device it goes over
the full queue list again and we get the NULL deref.

Fixes: f1d4ef7d ("nvmet-rdma: register ib_client to not deadlock in device removal")
Signed-off-by: Israel Rukshin <israelr@mellanox.com>
Reviewed-by: Max Gurtovoy <maxg@mellanox.com>
Reviewed-by: Sagi Grimberg <sagi@grmberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

blk-mq: Avoid that request queue removal can trigger list corruption

[ Upstream commit aba7afc5671c23beade64d10caf86e24a9105dab ]

Avoid that removal of a request queue sporadically triggers the
following warning:

list_del corruption. next->prev should be ffff8807d649b970, but was 6b6b6b6b6b6b6b6b
WARNING: CPU: 3 PID: 342 at lib/list_debug.c:56 __list_del_entry_valid+0x92/0xa0
Call Trace:
 process_one_work+0x11b/0x660
 worker_thread+0x3d/0x3b0
 kthread+0x129/0x140
 ret_from_fork+0x27/0x40

Signed-off-by: Bart Van Assche <bart.vanassche@wdc.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Hannes Reinecke <hare@suse.com>
Cc: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ide: ide-atapi: fix compile error with defining macro DEBUG

[ Upstream commit 8dc7a31fbce5e2dbbacd83d910da37105181b054 ]

Compile ide-atapi failed with defining macro "DEBUG"
...
|drivers/ide/ide-atapi.c:285:52: error: 'struct request' has
no member named 'cmd'; did you mean 'csd'?
|  debug_log("%s: rq->cmd[0]: 0x%x\n", __func__, rq->cmd[0]);
...

Since we split the scsi_request out of struct request, it missed
do the same thing on debug_log

Fixes: 82ed4db499b8 ("block: split scsi_request out of struct request")

Signed-off-by: Hongxu Jia <hongxu.jia@windriver.com>
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipvlan: fix ipv6 outbound device

[ Upstream commit ca29fd7cce5a6444d57fb86517589a1a31c759e1 ]

When process the outbound packet of ipv6, we should assign the master
device to output device other than input device.

Signed-off-by: Keefe Liu <liuqifa@huawei.com>
Acked-by: Mahesh Bandewar <maheshb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

powerpc/powernv/idle: Round up latency and residency values

[ Upstream commit 8d4e10e9ed9450e18fbbf6a8872be0eac9fd4999 ]

On PowerNV platforms, firmware provides exit latency and
target residency for each of the idle states in nano
seconds.  Cpuidle framework expects the values in micro
seconds.  Round up to nearest micro seconds to avoid errors
in cases where the values are defined as fractional micro
seconds.

Default idle state of 'snooze' has exit latency of zero.  If
other states have fractional micro second exit latency, they
would get rounded down to zero micro second and make cpuidle
framework choose deeper idle state when snooze loop is the
right choice.

Reported-by: Anton Blanchard <anton@samba.org>
Signed-off-by: Vaidyanathan Srinivasan <svaidy@linux.vnet.ibm.com>
Reviewed-by: Gautham R. Shenoy <ego@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

kbuild: do not call cc-option before KBUILD_CFLAGS initialization

[ Upstream commit 433dc2ebe7d17dd21cba7ad5c362d37323592236 ]

Some $(call cc-option,...) are invoked very early, even before
KBUILD_CFLAGS, etc. are initialized.

The returned string from $(call cc-option,...) depends on
KBUILD_CPPFLAGS, KBUILD_CFLAGS, and GCC_PLUGINS_CFLAGS.

Since they are exported, they are not empty when the top Makefile
is recursively invoked.

The recursion occurs in several places.  For example, the top
Makefile invokes itself for silentoldconfig.  "make tinyconfig",
"make rpm-pkg" are the cases, too.

In those cases, the second call of cc-option from the same line
runs a different shell command due to non-pristine KBUILD_CFLAGS.

To get the same result all the time, KBUILD_* and GCC_PLUGINS_CFLAGS
must be initialized before any call of cc-option.  This avoids
garbage data in the .cache.mk file.

Move all calls of cc-option below the config targets because target
compiler flags are unnecessary for Kconfig.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Reviewed-by: Douglas Anderson <dianders@chromium.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

KVM: arm/arm64: vgic-its: Preserve the revious read from the pending table

commit 64afe6e9eb4841f35317da4393de21a047a883b3 upstream.

The current pending table parsing code assumes that we keep the
previous read of the pending bits, but keep that variable in
the current block, making sure it is discarded on each loop.

We end-up using whatever is on the stack. Who knows, it might
just be the right thing...

Fixes: 33d3bc9556a7d ("KVM: arm64: vgic-its: Read initial LPI pending table")
Reported-by: AKASHI Takahiro <takahiro.akashi@linaro.org>
Reviewed-by: Christoffer Dall <christoffer.dall@linaro.org>
Signed-off-by: Marc Zyngier <marc.zyngier@arm.com>
Signed-off-by: Christoffer Dall <christoffer.dall@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fix kcm_clone()

commit a5739435b5a3b8c449f8844ecd71a3b1e89f0a33 upstream.

1) it's fput() or sock_release(), not both
2) don't do fd_install() until the last failure exit.
3) not a bug per se, but... don't attach socket to struct file
   until it's set up.

Take reserving descriptor into the caller, move fd_install() to the
caller, sanitize failure exits and calling conventions.

Acked-by: Tom Herbert <tom@herbertland.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fcntl: don't cap l_start and l_end values for F_GETLK64 in compat syscall

commit 4d2dc2cc766c3b51929658cacbc6e34fc8e242fb upstream.

Currently, we're capping the values too low in the F_GETLK64 case. The
fields in that structure are 64-bit values, so we shouldn't need to do
any sort of fixup there.

Make sure we check that assumption at build time in the future however
by ensuring that the sizes we're copying will fit.

With this, we no longer need COMPAT_LOFF_T_MAX either, so remove it.

Fixes: 94073ad77fff2 (fs/locks: don't mess with the address limit in compat_fcntl64)
Reported-by: Vitaly Lipatov <lav@etersoft.ru>
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Reviewed-by: David Howells <dhowells@redhat.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: gadget: ffs: Forbid usb_ep_alloc_request from sleeping

commit 30bf90ccdec1da9c8198b161ecbff39ce4e5a9ba upstream.

Found using DEBUG_ATOMIC_SLEEP while submitting an AIO read operation:

[  100.853642] BUG: sleeping function called from invalid context at mm/slab.h:421
[  100.861148] in_atomic(): 1, irqs_disabled(): 1, pid: 1880, name: python
[  100.867954] 2 locks held by python/1880:
[  100.867961]  #0:  (&epfile->mutex){....}, at: [<f8188627>] ffs_mutex_lock+0x27/0x30 [usb_f_fs]
[  100.868020]  #1:  (&(&ffs->eps_lock)->rlock){....}, at: [<f818ad4b>] ffs_epfile_io.isra.17+0x24b/0x590 [usb_f_fs]
[  100.868076] CPU: 1 PID: 1880 Comm: python Not tainted 4.14.0-edison+ #118
[  100.868085] Hardware name: Intel Corporation Merrifield/BODEGA BAY, BIOS 542 2015.01.21:18.19.48
[  100.868093] Call Trace:
[  100.868122]  dump_stack+0x47/0x62
[  100.868156]  ___might_sleep+0xfd/0x110
[  100.868182]  __might_sleep+0x68/0x70
[  100.868217]  kmem_cache_alloc_trace+0x4b/0x200
[  100.868248]  ? dwc3_gadget_ep_alloc_request+0x24/0xe0 [dwc3]
[  100.868302]  dwc3_gadget_ep_alloc_request+0x24/0xe0 [dwc3]
[  100.868343]  usb_ep_alloc_request+0x16/0xc0 [udc_core]
[  100.868386]  ffs_epfile_io.isra.17+0x444/0x590 [usb_f_fs]
[  100.868424]  ? _raw_spin_unlock_irqrestore+0x27/0x40
[  100.868457]  ? kiocb_set_cancel_fn+0x57/0x60
[  100.868477]  ? ffs_ep0_poll+0xc0/0xc0 [usb_f_fs]
[  100.868512]  ffs_epfile_read_iter+0xfe/0x157 [usb_f_fs]
[  100.868551]  ? security_file_permission+0x9c/0xd0
[  100.868587]  ? rw_verify_area+0xac/0x120
[  100.868633]  aio_read+0x9d/0x100
[  100.868692]  ? __fget+0xa2/0xd0
[  100.868727]  ? __might_sleep+0x68/0x70
[  100.868763]  SyS_io_submit+0x471/0x680
[  100.868878]  do_int80_syscall_32+0x4e/0xd0
[  100.868921]  entry_INT80_32+0x2a/0x2a
[  100.868932] EIP: 0xb7fbb676
[  100.868941] EFLAGS: 00000292 CPU: 1
[  100.868951] EAX: ffffffda EBX: b7aa2000 ECX: 00000002 EDX: b7af8368
[  100.868961] ESI: b7fbb660 EDI: b7aab000 EBP: bfb6c658 ESP: bfb6c638
[  100.868973]  DS: 007b ES: 007b FS: 0000 GS: 0033 SS: 007b

Signed-off-by: Vincent Pelletier <plr.vincent@gmail.com>
Signed-off-by: Felipe Balbi <felipe.balbi@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipmi: Stop timers before cleaning up the module

commit 4f7f5551a760eb0124267be65763008169db7087 upstream.

System may crash after unloading ipmi_si.ko module
because a timer may remain and fire after the module cleaned up resources.

cleanup_one_si() contains the following processing.

        /*
         * Make sure that interrupts, the timer and the thread are
         * stopped and will not run again.
         */
        if (to_clean->irq_cleanup)
                to_clean->irq_cleanup(to_clean);
        wait_for_timer_and_thread(to_clean);

        /*
         * Timeouts are stopped, now make sure the interrupts are off
         * in the BMC.  Note that timers and CPU interrupts are off,
         * so no need for locks.
         */
        while (to_clean->curr_msg || (to_clean->si_state != SI_NORMAL)) {
                poll(to_clean);
                schedule_timeout_uninterruptible(1);
        }

si_state changes as following in the while loop calling poll(to_clean).

  SI_GETTING_MESSAGES
    => SI_CHECKING_ENABLES
     => SI_SETTING_ENABLES
      => SI_GETTING_EVENTS
       => SI_NORMAL

As written in the code comments above,
timers are expected to stop before the polling loop and not to run again.
But the timer is set again in the following process
when si_state becomes SI_SETTING_ENABLES.

  => poll
     => smi_event_handler
       => handle_transaction_done
          // smi_info->si_state == SI_SETTING_ENABLES
         => start_getting_events
           => start_new_msg
            => smi_mod_timer
              => mod_timer

As a result, before the timer set in start_new_msg() expires,
the polling loop may see si_state becoming SI_NORMAL
and the module clean-up finishes.

For example, hard LOCKUP and panic occurred as following.
smi_timeout was called after smi_event_handler,
kcs_event and hangs at port_inb()
trying to access I/O port after release.

    [exception RIP: port_inb+19]
    RIP: ffffffffc0473053  RSP: ffff88069fdc3d80  RFLAGS: 00000006
    RAX: ffff8806800f8e00  RBX: ffff880682bd9400  RCX: 0000000000000000
    RDX: 0000000000000ca3  RSI: 0000000000000ca3  RDI: ffff8806800f8e40
    RBP: ffff88069fdc3d80   R8: ffffffff81d86dfc   R9: ffffffff81e36426
    R10: 00000000000509f0  R11: 0000000000100000  R12: 0000000000]:000000
    R13: 0000000000000000  R14: 0000000000000246  R15: ffff8806800f8e00
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0000
 --- <NMI exception stack> ---

To fix the problem I defined a flag, timer_can_start,
as member of struct smi_info.
The flag is enabled immediately after initializing the timer
and disabled immediately before waiting for timer deletion.

Fixes: 0cfec916e86d ("ipmi: Start the timer and thread on internal msgs")
Signed-off-by: Yamazaki Masamitsu <m-yamazaki@ah.jp.nec.com>
[Some fairly major changes went into the IPMI driver in 4.15, so this
 required a backport as the code had changed and moved to a different
 file.]
Signed-off-by: Corey Minyard <cminyard@mvista.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

sctp: use right member as the param of list_for_each_entry

[ Upstream commit a8dd397903a6e57157f6265911f7d35681364427 ]

Commit d04adf1b3551 ("sctp: reset owner sk for data chunks on out queues
when migrating a sock") made a mistake that using 'list' as the param of
list_for_each_entry to traverse the retransmit, sacked and abandoned
queues, while chunks are using 'transmitted_list' to link into these
queues.

It could cause NULL dereference panic if there are chunks in any of these
queues when peeling off one asoc.

So use the chunk member 'transmitted_list' instead in this patch.

Fixes: d04adf1b3551 ("sctp: reset owner sk for data chunks on out queues when migrating a sock")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

cls_bpf: don't decrement net's refcount when offload fails

[ Upstream commit 25415cec502a1232b19fffc85465882b19a90415 ]

When cls_bpf offload was added it seemed like a good idea to
call cls_bpf_delete_prog() instead of extending the error
handling path, since the software state is fully initialized
at that point.  This handling of errors without jumping to
the end of the function is error prone, as proven by later
commit missing that extra call to __cls_bpf_delete_prog().

__cls_bpf_delete_prog() is now expected to be invoked with
a reference on exts->net or the field zeroed out.  The call
on the offload's error patch does not fullfil this requirement,
leading to each error stealing a reference on net namespace.

Create a function undoing what cls_bpf_set_parms() did and
use it from __cls_bpf_delete_prog() and the error path.

Fixes: aae2c35ec892 ("cls_bpf: use tcf_exts_get_net() before call_rcu()")
Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Reviewed-by: Simon Horman <simon.horman@netronome.com>
Acked-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Cong Wang <xiyou.wangcong@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: openvswitch: datapath: fix data type in queue_gso_packets

[ Upstream commit 2734166e89639c973c6e125ac8bcfc2d9db72b70 ]

gso_type is being used in binary AND operations together with SKB_GSO_UDP.
The issue is that variable gso_type is of type unsigned short and
SKB_GSO_UDP expands to more than 16 bits:

SKB_GSO_UDP = 1 << 16

this makes any binary AND operation between gso_type and SKB_GSO_UDP to
be always zero, hence making some code unreachable and likely causing
undesired behavior.

Fix this by changing the data type of variable gso_type to unsigned int.

Addresses-Coverity-ID: 1462223
Fixes: 0c19f846d582 ("net: accept UFO datagrams from tuntap and packet")
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
Acked-by: Willem de Bruijn <willemb@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: accept UFO datagrams from tuntap and packet

[ Upstream commit 0c19f846d582af919db66a5914a0189f9f92c936 ]

Tuntap and similar devices can inject GSO packets. Accept type
VIRTIO_NET_HDR_GSO_UDP, even though not generating UFO natively.

Processes are expected to use feature negotiation such as TUNSETOFFLOAD
to detect supported offload types and refrain from injecting other
packets. This process breaks down with live migration: guest kernels
do not renegotiate flags, so destination hosts need to expose all
features that the source host does.

Partially revert the UFO removal from 182e0b6b5846~1..d9d30adf5677.
This patch introduces nearly(*) no new code to simplify verification.
It brings back verbatim tuntap UFO negotiation, VIRTIO_NET_HDR_GSO_UDP
insertion and software UFO segmentation.

It does not reinstate protocol stack support, hardware offload
(NETIF_F_UFO), SKB_GSO_UDP tunneling in SKB_GSO_SOFTWARE or reception
of VIRTIO_NET_HDR_GSO_UDP packets in tuntap.

To support SKB_GSO_UDP reappearing in the stack, also reinstate
logic in act_csum and openvswitch. Achieve equivalence with v4.13 HEAD
by squashing in commit 939912216fa8 ("net: skb_needs_check() removes
CHECKSUM_UNNECESSARY check for tx.") and reverting commit 8d63bee643f1
("net: avoid skb_warn_bad_offload false positives on UFO").

(*) To avoid having to bring back skb_shinfo(skb)->ip6_frag_id,
ipv6_proxy_select_ident is changed to return a __be32 and this is
assigned directly to the frag_hdr. Also, SKB_GSO_UDP is inserted
at the end of the enum to minimize code churn.

Tested
  Booted a v4.13 guest kernel with QEMU. On a host kernel before this
  patch `ethtool -k eth0` shows UFO disabled. After the patch, it is
  enabled, same as on a v4.13 host kernel.

  A UFO packet sent from the guest appears on the tap device:
    host:
      nc -l -p -u 8000 &
      tcpdump -n -i tap0

    guest:
      dd if=/dev/zero of=payload.txt bs=1 count=2000
      nc -u 192.16.1.1 8000 < payload.txt

  Direct tap to tap transmission of VIRTIO_NET_HDR_GSO_UDP succeeds,
  packets arriving fragmented:

    ./with_tap_pair.sh ./tap_send_ufo tap0 tap1
    (from https://github.com/wdebruij/kerneltools/tree/master/tests)

Changes
  v1 -> v2
    - simplified set_offload change (review comment)
    - documented test procedure

Link: http://lkml.kernel.org/r/<CAF=yD-LuUeDuL9YWPJD9ykOZ0QCjNeznPDr6whqZ9NGMNF12Mw@mail.gmail.com>
Fixes: fb652fdfe837 ("macvlan/macvtap: Remove NETIF_F_UFO advertisement.")
Reported-by: Michal Kubecek <mkubecek@suse.cz>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tun: fix rcu_read_lock imbalance in tun_build_skb

[ Upstream commit 654d573845f35017dc397840fa03610fef3d08b0 ]

rcu_read_lock in tun_build_skb is used to rcu_dereference tun->xdp_prog
safely, rcu_read_unlock should be done in every return path.

Now I could see one place missing it, where it returns NULL in switch-case
XDP_REDIRECT,  another palce using rcu_read_lock wrongly, where it returns
NULL in if (xdp_xmit) chunk.

So fix both in this patch.

Fixes: 761876c857cb ("tap: XDP support")
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: ipv6: Fixup device for anycast routes during copy

[ Upstream commit 98d11291d189cb5adf49694d0ad1b971c0212697 ]

Florian reported a breakage with anycast routes due to commit
4832c30d5458 ("net: ipv6: put host and anycast routes on device with
address"). Prior to this commit anycast routes were added against the
loopback device causing repetitive route entries with no insight into
why they existed. e.g.:
  $ ip -6 ro ls  table local type anycast
  anycast 2001:db8:1:: dev lo proto kernel metric 0 pref medium
  anycast 2001:db8:2:: dev lo proto kernel metric 0 pref medium
  anycast fe80:: dev lo proto kernel metric 0 pref medium
  anycast fe80:: dev lo proto kernel metric 0 pref medium

The point of commit 4832c30d5458 is to add the routes using the device
with the address which is causing the route to be added. e.g.,:
  $ ip -6 ro ls  table local type anycast
  anycast 2001:db8:1:: dev eth1 proto kernel metric 0 pref medium
  anycast 2001:db8:2:: dev eth2 proto kernel metric 0 pref medium
  anycast fe80:: dev eth2 proto kernel metric 0 pref medium
  anycast fe80:: dev eth1 proto kernel metric 0 pref medium

For traffic to work as it did before, the dst device needs to be switched
to the loopback when the copy is created similar to local routes.

Fixes: 4832c30d5458 ("net: ipv6: put host and anycast routes on device with address")
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tun: free skb in early errors

[ Upstream commit c33ee15b3820a03cf8229ba9415084197b827f8c ]

tun_recvmsg() supports accepting skb by msg_control after
commit ac77cfd4258f ("tun: support receiving skb through msg_control"),
the skb if presented should be freed no matter how far it can go
along, otherwise it would be leaked.

This patch fixes several missed cases.

Signed-off-by: Wei Xu <wexu@redhat.com>
Reported-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: when scheduling TLP, time of RTO should account for current ACK

[ Upstream commit ed66dfaf236c04d414de1d218441296e57fb2bd2 ]

Fix the TLP scheduling logic so that when scheduling a TLP probe, we
ensure that the estimated time at which an RTO would fire accounts for
the fact that ACKs indicating forward progress should push back RTO
times.

After the following fix:

df92c8394e6e ("tcp: fix xmit timer to only be reset if data ACKed/SACKed")

we had an unintentional behavior change in the following kind of
scenario: suppose the RTT variance has been very low recently. Then
suppose we send out a flight of N packets and our RTT is 100ms:

t=0: send a flight of N packets
t=100ms: receive an ACK for N-1 packets

The response before df92c8394e6e that was:
  -> schedule a TLP for now + RTO_interval

The response after df92c8394e6e is:
  -> schedule a TLP for t=0 + RTO_interval

Since RTO_interval = srtt + RTT_variance, this means that we have
scheduled a TLP timer at a point in the future that only accounts for
RTT_variance. If the RTT_variance term is small, this means that the
timer fires soon.

Before df92c8394e6e this would not happen, because in that code, when
we receive an ACK for a prefix of flight, we did:

    1) Near the top of tcp_ack(), switch from TLP timer to RTO
       at write_queue_head->paket_tx_time + RTO_interval:
            if (icsk->icsk_pending == ICSK_TIME_LOSS_PROBE)
                   tcp_rearm_rto(sk);

    2) In tcp_clean_rtx_queue(), update the RTO to now + RTO_interval:
            if (flag & FLAG_ACKED) {
                   tcp_rearm_rto(sk);

    3) In tcp_ack() after tcp_fastretrans_alert() switch from RTO
       to TLP at now + RTO_interval:
            if (icsk->icsk_pending == ICSK_TIME_RETRANS)
                   tcp_schedule_loss_probe(sk);

In df92c8394e6e we removed that 3-phase dance, and instead directly
set the TLP timer once: we set the TLP timer in cases like this to
write_queue_head->packet_tx_time + RTO_interval. So if the RTT
variance is small, then this means that this is setting the TLP timer
to fire quite soon. This means if the ACK for the tail of the flight
takes longer than an RTT to arrive (often due to delayed ACKs), then
the TLP timer fires too quickly.

Fixes: df92c8394e6e ("tcp: fix xmit timer to only be reset if data ACKed/SACKed")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: Yuchung Cheng <ycheng@google.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Soheil Hassas Yeganeh <soheil@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tap: free skb if flags error

[ Upstream commit 61d78537843e676e7f56ac6db333db0c0529b892 ]

tap_recvmsg() supports accepting skb by msg_control after
commit 3b4ba04acca8 ("tap: support receiving skb from msg_control"),
the skb if presented should be freed within the function, otherwise
it would be leaked.

Signed-off-by: Wei Xu <wexu@redhat.com>
Reported-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: sched: cbq: create block for q->link.block

[ Upstream commit d51aae68b142f48232257e96ce317db25445418d ]

q->link.block is not initialized, that leads to EINVAL when one tries to
add filter there. So initialize it properly.

This can be reproduced by:
$ tc qdisc add dev eth0 root handle 1: cbq avpkt 1000 rate 1000Mbit bandwidth 1000Mbit
$ tc filter add dev eth0 parent 1: protocol ip prio 100 u32 match ip protocol 0 0x00 flowid 1:1

Reported-by: Jaroslav Aster <jaster@redhat.com>
Reported-by: Ivan Vecera <ivecera@redhat.com>
Fixes: 6529eaba33f0 ("net: sched: introduce tcf block infractructure")
Signed-off-by: Jiri Pirko <jiri@mellanox.com>
Acked-by: Eelco Chaudron <echaudro@redhat.com>
Reviewed-by: Ivan Vecera <ivecera@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: use current time in tcp_rcv_space_adjust()

[ Upstream commit 8632385022f2b05a6ca0b9e0f95575865de0e2ce ]

When I switched rcv_rtt_est to high resolution timestamps, I forgot
that tp->tcp_mstamp needed to be refreshed in tcp_rcv_space_adjust()

Using an old timestamp leads to autotuning lags.

Fixes: 645f4c6f2ebd ("tcp: switch rcv_rtt_est and rcvq_space to high resolution timestamps")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Wei Wang <weiwan@google.com>
Cc: Neal Cardwell <ncardwell@google.com>
Cc: Yuchung Cheng <ycheng@google.com>
Acked-by: Neal Cardwell <ncardwell@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tipc: call tipc_rcv() only if bearer is up in tipc_udp_recv()

[ Upstream commit c7799c067c2ae33e348508c8afec354f3257ff25 ]

Remove the second tipc_rcv() call in tipc_udp_recv(). We have just
checked that the bearer is not up, and calling tipc_rcv() with a bearer
that is not up leads to a TIPC div-by-zero crash in
tipc_node_calculate_timer(). The crash is rare in practice, but can
happen like this:

  We're enabling a bearer, but it's not yet up and fully initialized.
  At the same time we receive a discovery packet, and in tipc_udp_recv()
  we end up calling tipc_rcv() with the not-yet-initialized bearer,
  causing later the div-by-zero crash in tipc_node_calculate_timer().

Jon Maloy explains the impact of removing the second tipc_rcv() call:
  "link setup in the worst case will be delayed until the next arriving
   discovery messages, 1 sec later, and this is an acceptable delay."

As the tipc_rcv() call is removed, just leave the function via the
rcu_out label, so that we will kfree_skb().

[   12.590450] Own node address <1.1.1>, network identity 1
[   12.668088] divide error: 0000 [#1] SMP
[   12.676952] CPU: 2 PID: 0 Comm: swapper/2 Not tainted 4.14.2-dirty #1
[   12.679225] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-2.fc27 04/01/2014
[   12.682095] task: ffff8c2a761edb80 task.stack: ffffa41cc0cac000
[   12.684087] RIP: 0010:tipc_node_calculate_timer.isra.12+0x45/0x60 [tipc]
[   12.686486] RSP: 0018:ffff8c2a7fc838a0 EFLAGS: 00010246
[   12.688451] RAX: 0000000000000000 RBX: ffff8c2a5b382600 RCX: 0000000000000000
[   12.691197] RDX: 0000000000000000 RSI: ffff8c2a5b382600 RDI: ffff8c2a5b382600
[   12.693945] RBP: ffff8c2a7fc838b0 R08: 0000000000000001 R09: 0000000000000001
[   12.696632] R10: 0000000000000000 R11: 0000000000000000 R12: ffff8c2a5d8949d8
[   12.699491] R13: ffffffff95ede400 R14: 0000000000000000 R15: ffff8c2a5d894800
[   12.702338] FS:  0000000000000000(0000) GS:ffff8c2a7fc80000(0000) knlGS:0000000000000000
[   12.705099] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[   12.706776] CR2: 0000000001bb9440 CR3: 00000000bd009001 CR4: 00000000003606e0
[   12.708847] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[   12.711016] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[   12.712627] Call Trace:
[   12.713390]  <IRQ>
[   12.714011]  tipc_node_check_dest+0x2e8/0x350 [tipc]
[   12.715286]  tipc_disc_rcv+0x14d/0x1d0 [tipc]
[   12.716370]  tipc_rcv+0x8b0/0xd40 [tipc]
[   12.717396]  ? minmax_running_min+0x2f/0x60
[   12.718248]  ? dst_alloc+0x4c/0xa0
[   12.718964]  ? tcp_ack+0xaf1/0x10b0
[   12.719658]  ? tipc_udp_is_known_peer+0xa0/0xa0 [tipc]
[   12.720634]  tipc_udp_recv+0x71/0x1d0 [tipc]
[   12.721459]  ? dst_alloc+0x4c/0xa0
[   12.722130]  udp_queue_rcv_skb+0x264/0x490
[   12.722924]  __udp4_lib_rcv+0x21e/0x990
[   12.723670]  ? ip_route_input_rcu+0x2dd/0xbf0
[   12.724442]  ? tcp_v4_rcv+0x958/0xa40
[   12.725039]  udp_rcv+0x1a/0x20
[   12.725587]  ip_local_deliver_finish+0x97/0x1d0
[   12.726323]  ip_local_deliver+0xaf/0xc0
[   12.726959]  ? ip_route_input_noref+0x19/0x20
[   12.727689]  ip_rcv_finish+0xdd/0x3b0
[   12.728307]  ip_rcv+0x2ac/0x360
[   12.728839]  __netif_receive_skb_core+0x6fb/0xa90
[   12.729580]  ? udp4_gro_receive+0x1a7/0x2c0
[   12.730274]  __netif_receive_skb+0x1d/0x60
[   12.730953]  ? __netif_receive_skb+0x1d/0x60
[   12.731637]  netif_receive_skb_internal+0x37/0xd0
[   12.732371]  napi_gro_receive+0xc7/0xf0
[   12.732920]  receive_buf+0x3c3/0xd40
[   12.733441]  virtnet_poll+0xb1/0x250
[   12.733944]  net_rx_action+0x23e/0x370
[   12.734476]  __do_softirq+0xc5/0x2f8
[   12.734922]  irq_exit+0xfa/0x100
[   12.735315]  do_IRQ+0x4f/0xd0
[   12.735680]  common_interrupt+0xa2/0xa2
[   12.736126]  </IRQ>
[   12.736416] RIP: 0010:native_safe_halt+0x6/0x10
[   12.736925] RSP: 0018:ffffa41cc0cafe90 EFLAGS: 00000246 ORIG_RAX: ffffffffffffff4d
[   12.737756] RAX: 0000000000000000 RBX: ffff8c2a761edb80 RCX: 0000000000000000
[   12.738504] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000000
[   12.739258] RBP: ffffa41cc0cafe90 R08: 0000014b5b9795e5 R09: ffffa41cc12c7e88
[   12.740118] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000002
[   12.740964] R13: ffff8c2a761edb80 R14: 0000000000000000 R15: 0000000000000000
[   12.741831]  default_idle+0x2a/0x100
[   12.742323]  arch_cpu_idle+0xf/0x20
[   12.742796]  default_idle_call+0x28/0x40
[   12.743312]  do_idle+0x179/0x1f0
[   12.743761]  cpu_startup_entry+0x1d/0x20
[   12.744291]  start_secondary+0x112/0x120
[   12.744816]  secondary_startup_64+0xa5/0xa5
[   12.745367] Code: b9 f4 01 00 00 48 89 c2 48 c1 ea 02 48 3d d3 07 00
00 48 0f 47 d1 49 8b 0c 24 48 39 d1 76 07 49 89 14 24 48 89 d1 31 d2 48
89 df <48> f7 f1 89 c6 e8 81 6e ff ff 5b 41 5c 5d c3 66 90 66 2e 0f 1f
[   12.747527] RIP: tipc_node_calculate_timer.isra.12+0x45/0x60 [tipc] RSP: ffff8c2a7fc838a0
[   12.748555] ---[ end trace 1399ab83390650fd ]---
[   12.749296] Kernel panic - not syncing: Fatal exception in interrupt
[   12.750123] Kernel Offset: 0x13200000 from 0xffffffff82000000
(relocation range: 0xffffffff80000000-0xffffffffbfffffff)
[   12.751215] Rebooting in 60 seconds..

Fixes: c9b64d492b1f ("tipc: add replicast peer discovery")
Signed-off-by: Tommi Rantala <tommi.t.rantala@nokia.com>
Cc: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: use IPCB instead of TCP_SKB_CB in inet_exact_dif_match()

[ Usptream commit b4d1605a8ea608fd7dc45b926a05d75d340bde4b ]

After this fix : ("tcp: add tcp_v4_fill_cb()/tcp_v4_restore_cb()"),
socket lookups happen while skb->cb[] has not been mangled yet by TCP.

Fixes: a04a480d4392 ("net: Require exact match for TCP socket lookups if dif is l3mdev")
Signed-off-by: David Ahern <dsahern@gmail.com>
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/qeth: fix GSO throughput regression

[ Upstream commit 6d69b1f1eb7a2edf8a3547f361c61f2538e054bb ]

Using GSO with small MTUs currently results in a substantial throughput
regression - which is caused by how qeth needs to map non-linear skbs
into its IO buffer elements:
compared to a linear skb, each GSO-segmented skb effectively consumes
twice as many buffer elements (ie two instead of one) due to the
additional header-only part. This causes the Output Queue to be
congested with low-utilized IO buffers.

Fix this as follows:
If the MSS is low enough so that a non-SG GSO segmentation produces
order-0 skbs (currently ~3500 byte), opt out from NETIF_F_SG. This is
where we anticipate the biggest savings, since an SG-enabled
GSO segmentation produces skbs that always consume at least two
buffer elements.

Larger MSS values continue to get a SG-enabled GSO segmentation, since
1) the relative overhead of the additional header-only buffer element
becomes less noticeable, and
2) the linearization overhead increases.

With the throughput regression fixed, re-enable NETIF_F_SG by default to
reap the significant CPU savings of GSO.

Fixes: 5722963a8e83 ("qeth: do not turn on SG per default")
Reported-by: Nils Hoppmann <niho@de.ibm.com>
Signed-off-by: Julian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/qeth: fix thinko in IPv4 multicast address tracking

[ Upsteam commit bc3ab70584696cb798b9e1e0ac8e6ced5fd4c3b8 ]

Commit 5f78e29ceebf ("qeth: optimize IP handling in rx_mode callback")
reworked how secondary addresses are managed for qeth devices.
Instead of dropping & subsequently re-adding all addresses on every
ndo_set_rx_mode() call, qeth now keeps track of the addresses that are
currently registered with the HW.
On a ndo_set_rx_mode(), we thus only need to do (de-)registration
requests for the addresses that have actually changed.

On L3 devices, the lookup for IPv4 Multicast addresses checks the wrong
hashtable - and thus never finds a match. As a result, we first delete
*all* such addresses, and then re-add them again. So each set_rx_mode()
causes a short period where the IPv4 Multicast addresses are not
registered, and the card stops forwarding inbound traffic for them.

Fix this by setting the ->is_multicast flag on the lookup object, thus
enabling qeth_l3_ip_from_hash() to search the correct hashtable and
find a match there.

Fixes: 5f78e29ceebf ("qeth: optimize IP handling in rx_mode callback")
Signed-off-by: Julian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/qeth: build max size GSO skbs on L2 devices

[ Upstream commit 0cbff6d4546613330a1c5f139f5c368e4ce33ca1 ]

The current GSO skb size limit was copy&pasted over from the L3 path,
where it is needed due to a TSO limitation.
As L2 devices don't offer TSO support (and thus all GSO skbs are
segmented before they reach the driver), there's no reason to restrict
the stack in how large it may build the GSO skbs.

Fixes: d52aec97e5bc ("qeth: enable scatter/gather in layer 2 mode")
Signed-off-by: Julian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp/dccp: block bh before arming time_wait timer

[ Upstream commit cfac7f836a715b91f08c851df915d401a4d52783 ]

Maciej Żenczykowski reported some panics in tcp_twsk_destructor()
that might be caused by the following bug.

timewait timer is pinned to the cpu, because we want to transition
timwewait refcount from 0 to 4 in one go, once everything has been
initialized.

At the time commit ed2e92394589 ("tcp/dccp: fix timewait races in timer
handling") was merged, TCP was always running from BH habdler.

After commit 5413d1babe8f ("net: do not block BH while processing
socket backlog") we definitely can run tcp_time_wait() from process
context.

We need to block BH in the critical section so that the pinned timer
has still its purpose.

This bug is more likely to happen under stress and when very small RTO
are used in datacenter flows.

Fixes: 5413d1babe8f ("net: do not block BH while processing socket backlog")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Maciej Żenczykowski <maze@google.com>
Acked-by: Maciej Żenczykowski <maze@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

stmmac: reset last TSO segment size after device open

[ Upstream commit 45ab4b13e46325d00f4acdb365d406e941a15f81 ]

The mss variable tracks the last max segment size sent to the TSO
engine. We do not update the hardware as long as we receive skb:s with
the same value in gso_size.

During a network device down/up cycle (mapped to stmmac_release() and
stmmac_open() callbacks) we issue a reset to the hardware and it
forgets the setting for mss. However we did not zero out our mss
variable so the next transmission of a gso packet happens with an
undefined hardware setting.

This triggers a hang in the TSO engine and eventuelly the netdev
watchdog will bark.

Fixes: f748be531d70 ("stmmac: support new GMAC4")
Signed-off-by: Lars Persson <larper@axis.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: remove hlist_nulls_add_tail_rcu()

[ Upstream commit d7efc6c11b277d9d80b99b1334a78bfe7d7edf10 ]

Alexander Potapenko reported use of uninitialized memory [1]

This happens when inserting a request socket into TCP ehash,
in __sk_nulls_add_node_rcu(), since sk_reuseport is not initialized.

Bug was added by commit d894ba18d4e4 ("soreuseport: fix ordering for
mixed v4/v6 sockets")

Note that d296ba60d8e2 ("soreuseport: Resolve merge conflict for v4/v6
ordering fix") missed the opportunity to get rid of
hlist_nulls_add_tail_rcu() :

Both UDP sockets and TCP/DCCP listeners no longer use
__sk_nulls_add_node_rcu() for their hash insertion.

Since all other sockets have unique 4-tuple, the reuseport status
has no special meaning, so we can always use hlist_nulls_add_head_rcu()
for them and save few cycles/instructions.

[1]

==================================================================
BUG: KMSAN: use of uninitialized memory in inet_ehash_insert+0xd40/0x1050
CPU: 0 PID: 0 Comm: swapper/0 Not tainted 4.13.0+ #3288
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs 01/01/2011
Call Trace:
 <IRQ>
 __dump_stack lib/dump_stack.c:16
 dump_stack+0x185/0x1d0 lib/dump_stack.c:52
 kmsan_report+0x13f/0x1c0 mm/kmsan/kmsan.c:1016
 __msan_warning_32+0x69/0xb0 mm/kmsan/kmsan_instr.c:766
 __sk_nulls_add_node_rcu ./include/net/sock.h:684
 inet_ehash_insert+0xd40/0x1050 net/ipv4/inet_hashtables.c:413
 reqsk_queue_hash_req net/ipv4/inet_connection_sock.c:754
 inet_csk_reqsk_queue_hash_add+0x1cc/0x300 net/ipv4/inet_connection_sock.c:765
 tcp_conn_request+0x31e7/0x36f0 net/ipv4/tcp_input.c:6414
 tcp_v4_conn_request+0x16d/0x220 net/ipv4/tcp_ipv4.c:1314
 tcp_rcv_state_process+0x42a/0x7210 net/ipv4/tcp_input.c:5917
 tcp_v4_do_rcv+0xa6a/0xcd0 net/ipv4/tcp_ipv4.c:1483
 tcp_v4_rcv+0x3de0/0x4ab0 net/ipv4/tcp_ipv4.c:1763
 ip_local_deliver_finish+0x6bb/0xcb0 net/ipv4/ip_input.c:216
 NF_HOOK ./include/linux/netfilter.h:248
 ip_local_deliver+0x3fa/0x480 net/ipv4/ip_input.c:257
 dst_input ./include/net/dst.h:477
 ip_rcv_finish+0x6fb/0x1540 net/ipv4/ip_input.c:397
 NF_HOOK ./include/linux/netfilter.h:248
 ip_rcv+0x10f6/0x15c0 net/ipv4/ip_input.c:488
 __netif_receive_skb_core+0x36f6/0x3f60 net/core/dev.c:4298
 __netif_receive_skb net/core/dev.c:4336
 netif_receive_skb_internal+0x63c/0x19c0 net/core/dev.c:4497
 napi_skb_finish net/core/dev.c:4858
 napi_gro_receive+0x629/0xa50 net/core/dev.c:4889
 e1000_receive_skb drivers/net/ethernet/intel/e1000/e1000_main.c:4018
 e1000_clean_rx_irq+0x1492/0x1d30
drivers/net/ethernet/intel/e1000/e1000_main.c:4474
 e1000_clean+0x43aa/0x5970 drivers/net/ethernet/intel/e1000/e1000_main.c:3819
 napi_poll net/core/dev.c:5500
 net_rx_action+0x73c/0x1820 net/core/dev.c:5566
 __do_softirq+0x4b4/0x8dd kernel/softirq.c:284
 invoke_softirq kernel/softirq.c:364
 irq_exit+0x203/0x240 kernel/softirq.c:405
 exiting_irq+0xe/0x10 ./arch/x86/include/asm/apic.h:638
 do_IRQ+0x15e/0x1a0 arch/x86/kernel/irq.c:263
 common_interrupt+0x86/0x86

Fixes: d894ba18d4e4 ("soreuseport: fix ordering for mixed v4/v6 sockets")
Fixes: d296ba60d8e2 ("soreuseport: Resolve merge conflict for v4/v6 ordering fix")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: Alexander Potapenko <glider@google.com>
Acked-by: Craig Gallek <kraig@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usbnet: fix alignment for frames with no ethernet header

[ Upstream commit a4abd7a80addb4a9547f7dfc7812566b60ec505c ]

The qmi_wwan minidriver support a 'raw-ip' mode where frames are
received without any ethernet header. This causes alignment issues
because the skbs allocated by usbnet are "IP aligned".

Fix by allowing minidrivers to disable the additional alignment
offset. This is implemented using a per-device flag, since the same
minidriver also supports 'ethernet' mode.

Fixes: 32f7adf633b9 ("net: qmi_wwan: support "raw IP" mode")
Reported-and-tested-by: Jay Foster <jay@systech.com>
Signed-off-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: remove buggy call to tcp_v6_restore_cb()

[ Upstream commit 3016dad75b48279e579117ee3ed566ba90a3b023 ]

tcp_v6_send_reset() expects to receive an skb with skb->cb[] layout as
used in TCP stack.
MD5 lookup uses tcp_v6_iif() and tcp_v6_sdif() and thus
TCP_SKB_CB(skb)->header.h6

This patch probably fixes RST packets sent on behalf of a timewait md5
ipv6 socket.

Before Florian patch, tcp_v6_restore_cb() was needed before jumping to
no_tcp_socket label.

Fixes: 271c3b9b7bda ("tcp: honour SO_BINDTODEVICE for TW_RST case too")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Cc: Florian Westphal <fw@strlen.de>
Acked-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net/packet: fix a race in packet_bind() and packet_notifier()

[ Upstream commit 15fe076edea787807a7cdc168df832544b58eba6 ]

syzbot reported crashes [1] and provided a C repro easing bug hunting.

When/if packet_do_bind() calls __unregister_prot_hook() and releases
po->bind_lock, another thread can run packet_notifier() and process an
NETDEV_UP event.

This calls register_prot_hook() and hooks again the socket right before
first thread is able to grab again po->bind_lock.

Fixes this issue by temporarily setting po->num to 0, as suggested by
David Miller.

[1]
dev_remove_pack: ffff8801bf16fa80 not found
------------[ cut here ]------------
kernel BUG at net/core/dev.c:7945!  ( BUG_ON(!list_empty(&dev->ptype_all)); )
invalid opcode: 0000 [#1] SMP KASAN
Dumping ftrace buffer:
   (ftrace buffer empty)
Modules linked in:
device syz0 entered promiscuous mode
CPU: 0 PID: 3161 Comm: syzkaller404108 Not tainted 4.14.0+ #190
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
task: ffff8801cc57a500 task.stack: ffff8801cc588000
RIP: 0010:netdev_run_todo+0x772/0xae0 net/core/dev.c:7945
RSP: 0018:ffff8801cc58f598 EFLAGS: 00010293
RAX: ffff8801cc57a500 RBX: dffffc0000000000 RCX: ffffffff841f75b2
RDX: 0000000000000000 RSI: 1ffff100398b1ede RDI: ffff8801bf1f8810
device syz0 entered promiscuous mode
RBP: ffff8801cc58f898 R08: 0000000000000001 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000000 R12: ffff8801bf1f8cd8
R13: ffff8801cc58f870 R14: ffff8801bf1f8780 R15: ffff8801cc58f7f0
FS:  0000000001716880(0000) GS:ffff8801db400000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 0000000020b13000 CR3: 0000000005e25000 CR4: 00000000001406f0
DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
Call Trace:
 rtnl_unlock+0xe/0x10 net/core/rtnetlink.c:106
 tun_detach drivers/net/tun.c:670 [inline]
 tun_chr_close+0x49/0x60 drivers/net/tun.c:2845
 __fput+0x333/0x7f0 fs/file_table.c:210
 ____fput+0x15/0x20 fs/file_table.c:244
 task_work_run+0x199/0x270 kernel/task_work.c:113
 exit_task_work include/linux/task_work.h:22 [inline]
 do_exit+0x9bb/0x1ae0 kernel/exit.c:865
 do_group_exit+0x149/0x400 kernel/exit.c:968
 SYSC_exit_group kernel/exit.c:979 [inline]
 SyS_exit_group+0x1d/0x20 kernel/exit.c:977
 entry_SYSCALL_64_fastpath+0x1f/0x96
RIP: 0033:0x44ad19

Fixes: 30f7ea1c2b5f ("packet: race condition in packet_bind")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Cc: Francesco Ruggeri <fruggeri@aristanetworks.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

packet: fix crash in fanout_demux_rollover()

syzkaller found a race condition fanout_demux_rollover() while removing
a packet socket from a fanout group.

po->rollover is read and operated on during packet_rcv_fanout(), via
fanout_demux_rollover(), but the pointer is currently cleared before the
synchronization in packet_release().   It is safer to delay the cleanup
until after synchronize_net() has been called, ensuring all calls to
packet_rcv_fanout() for this socket have finished.

To further simplify synchronization around the rollover structure, set
po->rollover in fanout_add() only if there are no errors.  This removes
the need for rcu in the struct and in the call to
packet_getsockopt(..., PACKET_ROLLOVER_STATS, ...).

Crashing stack trace:
 fanout_demux_rollover+0xb6/0x4d0 net/packet/af_packet.c:1392
 packet_rcv_fanout+0x649/0x7c8 net/packet/af_packet.c:1487
 dev_queue_xmit_nit+0x835/0xc10 net/core/dev.c:1953
 xmit_one net/core/dev.c:2975 [inline]
 dev_hard_start_xmit+0x16b/0xac0 net/core/dev.c:2995
 __dev_queue_xmit+0x17a4/0x2050 net/core/dev.c:3476
 dev_queue_xmit+0x17/0x20 net/core/dev.c:3509
 neigh_connected_output+0x489/0x720 net/core/neighbour.c:1379
 neigh_output include/net/neighbour.h:482 [inline]
 ip6_finish_output2+0xad1/0x22a0 net/ipv6/ip6_output.c:120
 ip6_finish_output+0x2f9/0x920 net/ipv6/ip6_output.c:146
 NF_HOOK_COND include/linux/netfilter.h:239 [inline]
 ip6_output+0x1f4/0x850 net/ipv6/ip6_output.c:163
 dst_output include/net/dst.h:459 [inline]
 NF_HOOK.constprop.35+0xff/0x630 include/linux/netfilter.h:250
 mld_sendpack+0x6a8/0xcc0 net/ipv6/mcast.c:1660
 mld_send_initial_cr.part.24+0x103/0x150 net/ipv6/mcast.c:2072
 mld_send_initial_cr net/ipv6/mcast.c:2056 [inline]
 ipv6_mc_dad_complete+0x99/0x130 net/ipv6/mcast.c:2079
 addrconf_dad_completed+0x595/0x970 net/ipv6/addrconf.c:4039
 addrconf_dad_work+0xac9/0x1160 net/ipv6/addrconf.c:3971
 process_one_work+0xbf0/0x1bc0 kernel/workqueue.c:2113
 worker_thread+0x223/0x1990 kernel/workqueue.c:2247
 kthread+0x35e/0x430 kernel/kthread.c:231
 ret_from_fork+0x2a/0x40 arch/x86/entry/entry_64.S:432

Fixes: 0648ab70afe6 ("packet: rollover prepare: per-socket state")
Fixes: 509c7a1ecc860 ("packet: avoid panic in packet_getsockopt()")
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Mike Maloney <maloney@google.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tcp: add tcp_v4_fill_cb()/tcp_v4_restore_cb()

[ Upstream commit eeea10b83a139451130df1594f26710c8fa390c8 ]

James Morris reported kernel stack corruption bug [1] while
running the SELinux testsuite, and bisected to a recent
commit bffa72cf7f9d ("net: sk_buff rbnode reorg")

We believe this commit is fine, but exposes an older bug.

SELinux code runs from tcp_filter() and might send an ICMP,
expecting IP options to be found in skb->cb[] using regular IPCB placement.

We need to defer TCP mangling of skb->cb[] after tcp_filter() calls.

This patch adds tcp_v4_fill_cb()/tcp_v4_restore_cb() in a very
similar way we added them for IPv6.

[1]
[  339.806024] SELinux: failure in selinux_parse_skb(), unable to parse packet
[  339.822505] Kernel panic - not syncing: stack-protector: Kernel stack is corrupted in: ffffffff81745af5
[  339.822505]
[  339.852250] CPU: 4 PID: 3642 Comm: client Not tainted 4.15.0-rc1-test #15
[  339.868498] Hardware name: LENOVO 10FGS0VA1L/30BC, BIOS FWKT68A   01/19/2017
[  339.885060] Call Trace:
[  339.896875]  <IRQ>
[  339.908103]  dump_stack+0x63/0x87
[  339.920645]  panic+0xe8/0x248
[  339.932668]  ? ip_push_pending_frames+0x33/0x40
[  339.946328]  ? icmp_send+0x525/0x530
[  339.958861]  ? kfree_skbmem+0x60/0x70
[  339.971431]  __stack_chk_fail+0x1b/0x20
[  339.984049]  icmp_send+0x525/0x530
[  339.996205]  ? netlbl_skbuff_err+0x36/0x40
[  340.008997]  ? selinux_netlbl_err+0x11/0x20
[  340.021816]  ? selinux_socket_sock_rcv_skb+0x211/0x230
[  340.035529]  ? security_sock_rcv_skb+0x3b/0x50
[  340.048471]  ? sk_filter_trim_cap+0x44/0x1c0
[  340.061246]  ? tcp_v4_inbound_md5_hash+0x69/0x1b0
[  340.074562]  ? tcp_filter+0x2c/0x40
[  340.086400]  ? tcp_v4_rcv+0x820/0xa20
[  340.098329]  ? ip_local_deliver_finish+0x71/0x1a0
[  340.111279]  ? ip_local_deliver+0x6f/0xe0
[  340.123535]  ? ip_rcv_finish+0x3a0/0x3a0
[  340.135523]  ? ip_rcv_finish+0xdb/0x3a0
[  340.147442]  ? ip_rcv+0x27c/0x3c0
[  340.158668]  ? inet_del_offload+0x40/0x40
[  340.170580]  ? __netif_receive_skb_core+0x4ac/0x900
[  340.183285]  ? rcu_accelerate_cbs+0x5b/0x80
[  340.195282]  ? __netif_receive_skb+0x18/0x60
[  340.207288]  ? process_backlog+0x95/0x140
[  340.218948]  ? net_rx_action+0x26c/0x3b0
[  340.230416]  ? __do_softirq+0xc9/0x26a
[  340.241625]  ? do_softirq_own_stack+0x2a/0x40
[  340.253368]  </IRQ>
[  340.262673]  ? do_softirq+0x50/0x60
[  340.273450]  ? __local_bh_enable_ip+0x57/0x60
[  340.285045]  ? ip_finish_output2+0x175/0x350
[  340.296403]  ? ip_finish_output+0x127/0x1d0
[  340.307665]  ? nf_hook_slow+0x3c/0xb0
[  340.318230]  ? ip_output+0x72/0xe0
[  340.328524]  ? ip_fragment.constprop.54+0x80/0x80
[  340.340070]  ? ip_local_out+0x35/0x40
[  340.350497]  ? ip_queue_xmit+0x15c/0x3f0
[  340.361060]  ? __kmalloc_reserve.isra.40+0x31/0x90
[  340.372484]  ? __skb_clone+0x2e/0x130
[  340.382633]  ? tcp_transmit_skb+0x558/0xa10
[  340.393262]  ? tcp_connect+0x938/0xad0
[  340.403370]  ? ktime_get_with_offset+0x4c/0xb0
[  340.414206]  ? tcp_v4_connect+0x457/0x4e0
[  340.424471]  ? __inet_stream_connect+0xb3/0x300
[  340.435195]  ? inet_stream_connect+0x3b/0x60
[  340.445607]  ? SYSC_connect+0xd9/0x110
[  340.455455]  ? __audit_syscall_entry+0xaf/0x100
[  340.466112]  ? syscall_trace_enter+0x1d0/0x2b0
[  340.476636]  ? __audit_syscall_exit+0x209/0x290
[  340.487151]  ? SyS_connect+0xe/0x10
[  340.496453]  ? do_syscall_64+0x67/0x1b0
[  340.506078]  ? entry_SYSCALL64_slow_path+0x25/0x25

Fixes: 971f10eca186 ("tcp: better TCP_SKB_CB layout to reduce cache line misses")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: James Morris <james.l.morris@oracle.com>
Tested-by: James Morris <james.l.morris@oracle.com>
Tested-by: Casey Schaufler <casey@schaufler-ca.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

sit: update frag_off info

[ Upstream commit f859b4af1c52493ec21173ccc73d0b60029b5b88 ]

After parsing the sit netlink change info, we forget to update frag_off in
ipip6_tunnel_update(). Fix it by assigning frag_off with new value.

Reported-by: Jianlin Shi <jishi@redhat.com>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Acked-by: Nicolas Dichtel <nicolas.dichtel@6wind.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

rds: Fix NULL pointer dereference in __rds_rdma_map

[ Upstream commit f3069c6d33f6ae63a1668737bc78aaaa51bff7ca ]

This is a fix for syzkaller719569, where memory registration was
attempted without any underlying transport being loaded.

Analysis of the case reveals that it is the setsockopt() RDS_GET_MR
(2) and RDS_GET_MR_FOR_DEST (7) that are vulnerable.

Here is an example stack trace when the bug is hit:

BUG: unable to handle kernel NULL pointer dereference at 00000000000000c0
IP: __rds_rdma_map+0x36/0x440 [rds]
PGD 2f93d03067 P4D 2f93d03067 PUD 2f93d02067 PMD 0
Oops: 0000 [#1] SMP
Modules linked in: bridge stp llc tun rpcsec_gss_krb5 nfsv4
dns_resolver nfs fscache rds binfmt_misc sb_edac intel_powerclamp
coretemp kvm_intel kvm irqbypass crct10dif_pclmul c rc32_pclmul
ghash_clmulni_intel pcbc aesni_intel crypto_simd glue_helper cryptd
iTCO_wdt mei_me sg iTCO_vendor_support ipmi_si mei ipmi_devintf nfsd
shpchp pcspkr i2c_i801 ioatd ma ipmi_msghandler wmi lpc_ich mfd_core
auth_rpcgss nfs_acl lockd grace sunrpc ip_tables ext4 mbcache jbd2
mgag200 i2c_algo_bit drm_kms_helper ixgbe syscopyarea ahci sysfillrect
sysimgblt libahci mdio fb_sys_fops ttm ptp libata sd_mod mlx4_core drm
crc32c_intel pps_core megaraid_sas i2c_core dca dm_mirror
dm_region_hash dm_log dm_mod
CPU: 48 PID: 45787 Comm: repro_set2 Not tainted 4.14.2-3.el7uek.x86_64 #2
Hardware name: Oracle Corporation ORACLE SERVER X5-2L/ASM,MOBO TRAY,2U, BIOS 31110000 03/03/2017
task: ffff882f9190db00 task.stack: ffffc9002b994000
RIP: 0010:__rds_rdma_map+0x36/0x440 [rds]
RSP: 0018:ffffc9002b997df0 EFLAGS: 00010202
RAX: 0000000000000000 RBX: ffff882fa2182580 RCX: 0000000000000000
RDX: 0000000000000000 RSI: ffffc9002b997e40 RDI: ffff882fa2182580
RBP: ffffc9002b997e30 R08: 0000000000000000 R09: 0000000000000002
R10: ffff885fb29e3838 R11: 0000000000000000 R12: ffff882fa2182580
R13: ffff882fa2182580 R14: 0000000000000002 R15: 0000000020000ffc
FS:  00007fbffa20b700(0000) GS:ffff882fbfb80000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00000000000000c0 CR3: 0000002f98a66006 CR4: 00000000001606e0
Call Trace:
 rds_get_mr+0x56/0x80 [rds]
 rds_setsockopt+0x172/0x340 [rds]
 ? __fget_light+0x25/0x60
 ? __fdget+0x13/0x20
 SyS_setsockopt+0x80/0xe0
 do_syscall_64+0x67/0x1b0
 entry_SYSCALL64_slow_path+0x25/0x25
RIP: 0033:0x7fbff9b117f9
RSP: 002b:00007fbffa20aed8 EFLAGS: 00000293 ORIG_RAX: 0000000000000036
RAX: ffffffffffffffda RBX: 00000000000c84a4 RCX: 00007fbff9b117f9
RDX: 0000000000000002 RSI: 0000400000000114 RDI: 000000000000109b
RBP: 00007fbffa20af10 R08: 0000000000000020 R09: 00007fbff9dd7860
R10: 0000000020000ffc R11: 0000000000000293 R12: 0000000000000000
R13: 00007fbffa20b9c0 R14: 00007fbffa20b700 R15: 0000000000000021

Code: 41 56 41 55 49 89 fd 41 54 53 48 83 ec 18 8b 87 f0 02 00 00 48
89 55 d0 48 89 4d c8 85 c0 0f 84 2d 03 00 00 48 8b 87 00 03 00 00 <48>
83 b8 c0 00 00 00 00 0f 84 25 03 00 0 0 48 8b 06 48 8b 56 08

The fix is to check the existence of an underlying transport in
__rds_rdma_map().

Signed-off-by: Håkon Bugge <haakon.bugge@oracle.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Acked-by: Santosh Shilimkar <santosh.shilimkar@oracle.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

vhost: fix skb leak in handle_rx()

[ Upstream commit 6e474083f3daf3a3546737f5d7d502ad12eb257c ]

Matthew found a roughly 40% tcp throughput regression with commit
c67df11f(vhost_net: try batch dequing from skb array) as discussed
in the following thread:
https://www.mail-archive.com/netdev@vger.kernel.org/msg187936.html

Eventually we figured out that it was a skb leak in handle_rx()
when sending packets to the VM. This usually happens when a guest
can not drain out vq as fast as vhost fills in, afterwards it sets
off the traffic jam and leaks skb(s) which occurs as no headcount
to send on the vq from vhost side.

This can be avoided by making sure we have got enough headcount
before actually consuming a skb from the batched rx array while
transmitting, which is simply done by moving checking the zero
headcount a bit ahead.

Signed-off-by: Wei Xu <wexu@redhat.com>
Reported-by: Matthew Rosato <mjrosato@linux.vnet.ibm.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tipc: fix memory leak in tipc_accept_from_sock()

[ Upstream commit a7d5f107b4978e08eeab599ee7449af34d034053 ]

When the function tipc_accept_from_sock() fails to create an instance of
struct tipc_subscriber it omits to free the already created instance of
struct tipc_conn instance before it returns.

We fix that with this commit.

Reported-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Jon Maloy <jon.maloy@ericsson.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

s390/qeth: fix early exit from error path

[ Upstream commit 83cf79a2fec3cf499eb6cb9eb608656fc2a82776 ]

When the allocation of the addr buffer fails, we need to free
our refcount on the inetdevice before returning.

Signed-off-by: Julian Wiedmann <jwi@linux.vnet.ibm.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: realtek: r8169: implement set_link_ksettings()

[ Upstream commit 9e77d7a5549dc4d4999a60676373ab3fd1dae4db ]

Commit 6fa1ba61520576cf1346c4ff09a056f2950cb3bf partially
implemented the new ethtool API, by replacing get_settings()
with get_link_ksettings(). This breaks ethtool, since the
userspace tool (according to the new API specs) never tries
the legacy set() call, when the new get() call succeeds.

All attempts to chance some setting from userspace result in:
> Cannot set new settings: Operation not supported

Implement the missing set() call.

Signed-off-by: Tobias Jakobi <tjakobi@math.uni-bielefeld.de>
Tested-by: Holger Hoffstätte <holger@applied-asynchrony.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: thunderx: Fix TCP/UDP checksum offload for IPv4 pkts

[ Upstream commit 134059fd2775be79e26c2dff87d25cc2f6ea5626 ]

Offload IP header checksum to NIC.

This fixes a previous patch which disabled checksum offloading
for both IPv4 and IPv6 packets.  So L3 checksum offload was
getting disabled for IPv4 pkts.  And HW is dropping these pkts
for some reason.

Without this patch, IPv4 TSO appears to be broken:

WIthout this patch I get ~16kbyte/s, with patch close to 2mbyte/s
when copying files via scp from test box to my home workstation.

Looking at tcpdump on sender it looks like hardware drops IPv4 TSO skbs.
This patch restores performance for me, ipv6 looks good too.

Fixes: fa6d7cb5d76c ("net: thunderx: Fix TCP/UDP checksum offload for IPv6 pkts")
Cc: Sunil Goutham <sgoutham@cavium.com>
Cc: Aleksey Makarov <aleksey.makarov@auriga.com>
Cc: Eric Dumazet <edumazet@google.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: thunderx: Fix TCP/UDP checksum offload for IPv6 pkts

[ Upstream commit fa6d7cb5d76cf0467c61420fc9238045aedfd379 ]

Don't offload IP header checksum to NIC.

This fixes a previous patch which enabled checksum offloading
for both IPv4 and IPv6 packets.  So L3 checksum offload was
getting enabled for IPv6 pkts.  And HW is dropping these pkts
as it assumes the pkt is IPv4 when IP csum offload is set
in the SQ descriptor.

Fixes:  3a9024f52c2e ("net: thunderx: Enable TSO and checksum offloads for ipv6")
Signed-off-by: Sunil Goutham <sgoutham@cavium.com>
Signed-off-by: Aleksey Makarov <aleksey.makarov@auriga.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

net: qmi_wwan: add Quectel BG96 2c7c:0296

[ Upstream commit f9409e7f086fa6c4623769b4b2f4f17a024d8143 ]

Quectel BG96 is an Qualcomm MDM9206 based IoT modem, supporting both
CAT-M and NB-IoT. Tested hardware is BG96 mounted on Quectel development
board (EVB). The USB id is added to qmi_wwan.c to allow QMI
communication with the BG96.

Signed-off-by: Sebastian Sjoholm <ssjoholm@mac.com>
Acked-by: Bjørn Mork <bjorn@mork.no>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Linux 4.14.6

afs: Connect up the CB.ProbeUuid

[ Upstream commit f4b3526d83c40dd8bf5948b9d7a1b2c340f0dcc8 ]

The handler for the CB.ProbeUuid operation in the cache manager is
implemented, but isn't listed in the switch-statement of operation
selection, so won't be used.  Fix this by adding it.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

afs: Fix total-length calculation for multiple-page send

[ Upstream commit 1199db603511d7463d9d3840f96f61967affc766 ]

Fix the total-length calculation in afs_make_call() when the operation
being dispatched has data from a series of pages attached.

Despite the patched code looking like that it should reduce mathematically
to the current code, it doesn't because the 32-bit unsigned arithmetic
being used to calculate the page-offset-difference doesn't correctly extend
to a 64-bit value when the result is effectively negative.

Without this, some FS.StoreData operations that span multiple pages fail,
reporting too little or too much data.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

IB/mlx5: Assign send CQ and recv CQ of UMR QP

[ Upstream commit 31fde034a8bd964a5c7c1a5663fc87a913158db2 ]

The UMR's QP is created by calling mlx5_ib_create_qp directly, and
therefore the send CQ and the recv CQ on the ibqp weren't assigned.

Assign them right after calling the mlx5_ib_create_qp to assure
that any access to those pointers will work as expected and won't
crash the system as might happen as part of reset flow.

Fixes: e126ba97dba9 ("mlx5: Add driver for Mellanox Connect-IB adapters")
Signed-off-by: Majd Dibbiny <majd@mellanox.com>
Reviewed-by: Yishai Hadas <yishaih@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

IB/mlx4: Increase maximal message size under UD QP

[ Upstream commit 5f22a1d87c5315a98981ecf93cd8de226cffe6ca ]

Maximal message should be used as a limit to the max message payload allowed,
without the headers. The ConnectX-3 check is done against this value includes
the headers. When the payload is 4K this will cause the NIC to drop packets.

Increase maximal message to 8K as workaround, this shouldn't change current
behaviour because we continue to set the MTU to 4k.

To reproduce;
set MTU to 4296 on the corresponding interface, for example:
ifconfig eth0 mtu 4296 (both server and client)

On server:
ib_send_bw -c UD -d mlx4_0 -s 4096 -n 1000000 -i1 -m 4096

On client:
ib_send_bw -d mlx4_0 -c UD <server_ip> -s 4096 -n 1000000 -i 1 -m 4096

Fixes: 6e0d733d9215 ("IB/mlx4: Allow 4K messages for UD QPs")
Signed-off-by: Mark Bloch <markb@mellanox.com>
Reviewed-by: Majd Dibbiny <majd@mellanox.com>
Signed-off-by: Leon Romanovsky <leon@kernel.org>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bnxt_re: changing the ip address shouldn't affect new connections

[ Upstream commit 063fb5bd1a01937094f40169a20e4aa5ca030db1 ]

While adding a new gid, the driver currently does not return the context
back to the stack. A subsequent del_gid() (e.g, when ip address is changed)
doesn't find the right context in the driver and it ends up dropping that
request. This results in the HW caching a stale gid entry and traffic fails
because of that. Fix by returning the proper context in bnxt_re_add_gid().

Signed-off-by: Sriharsha Basavapatna <sriharsha.basavapatna@broadcom.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

f2fs: fix to clear FI_NO_PREALLOC

[ Upstream commit 28cfafb73853f0494b06649716687a3ea07681d5 ]

We need to clear FI_NO_PREALLOC flag in error path of f2fs_file_write_iter,
otherwise we will lose the chance to preallocate blocks in latter write()
at one time.

Fixes: dc91de78e5e1 ("f2fs: do not preallocate blocks which has wrong buffer")
Signed-off-by: Chao Yu <yuchao0@huawei.com>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xfrm: Copy policy family in clone_policy

[ Upstream commit 0e74aa1d79a5bbc663e03a2804399cae418a0321 ]

The syzbot found an ancient bug in the IPsec code.  When we cloned
a socket policy (for example, for a child TCP socket derived from a
listening socket), we did not copy the family field.  This results
in a live policy with a zero family field.  This triggers a BUG_ON
check in the af_key code when the cloned policy is retrieved.

This patch fixes it by copying the family field over.

Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

tls: Use kzalloc for aead_request allocation

[ Upstream commit 61ef6da622aa7b66bf92991bd272490eea6c712e ]

Use kzalloc for aead_request allocation as
we don't set all the bits in the request.

Fixes: 3c4d7559159b ('tls: kernel TLS support')
Signed-off-by: Ilya Lesokhin <ilyal@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

jump_label: Invoke jump_label_test() via early_initcall()

[ Upstream commit 92ee46efeb505ead3ab06d3c5ce695637ed5f152 ]

Fengguang Wu reported that running the rcuperf test during boot can cause
the jump_label_test() to hit a WARN_ON(). The issue is that the core jump
label code relies on kernel_text_address() to detect when it can no longer
update branches that may be contained in __init sections. The
kernel_text_address() in turn assumes that if the system_state variable is
greter than or equal to SYSTEM_RUNNING then __init sections are no longer
valid (since the assumption is that they have been freed). However, when
rcuperf is setup to run in early boot it can call kernel_power_off() which
sets the system_state to SYSTEM_POWER_OFF.

Since rcuperf initialization is invoked via a module_init(), we can make
the dependency of jump_label_test() needing to complete before rcuperf
explicit by calling it via early_initcall().

Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Jason Baron <jbaron@akamai.com>
Acked-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/1510609727-2238-1-git-send-email-jbaron@akamai.com
Signed-off-by: Ingo Molnar <mingo@kernel.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

atm: horizon: Fix irq release error

[ Upstream commit bde533f2ea607cbbbe76ef8738b36243939a7bc2 ]

atm_dev_register() can fail here and passed parameters to free irq
which is not initialised. Initialization of 'dev->irq' happened after
the 'goto out_free_irq'. So using 'irq' insted of 'dev->irq' in
free_irq().

Signed-off-by: Arvind Yadav <arvind.yadav.cs@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

kbuild: rpm-pkg: fix jobserver unavailable warning

[ Upstream commit 606625be47bc87b6fab0af10cd57aaa675cb9e42 ]

If "make rpm-pkg" or "make binrpm-pkg" is run with -j[jobs] option,
the following warning message is displayed.

  warning: jobserver unavailable: using -j1.  Add '+' to parent make rule.

Follow the suggestion.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mailbox: mailbox-test: don't rely on rx_buffer content to signal data ready

[ Upstream commit e339c80af95e14de3712d69ddea09a3868fa14cd ]

Currently we rely on the first byte of the Rx buffer to check if there's
any data available to be read. If the first byte of the received buffer
is zero (i.e. null character), then we fail to signal that data is
available even when it's available.

Instead introduce a boolean variable to track the data availability and
update it in the channel receive callback as ready and clear it when the
data is read.

Signed-off-by: Sudeep Holla <sudeep.holla@arm.com>
Signed-off-by: Jassi Brar <jaswinder.singh@linaro.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: hi3660: fix incorrect uart3 clock freqency

[ Upstream commit d33fb1b9f0fcb67f2b9f8b1891465a088a9480f8 ]

UART3 clock rate is doubled in previous commit.

This error is not detected until recently a mezzanine board which makes
real use of uart3 port (through LS connector of 96boards) was setup
and tested on hi3660-hikey960 board.

This patch changes clock source rate of clk_factor_uart3 to 100000000.

Signed-off-by: Zhong Kaihua <zhongkaihua@huawei.com>
Signed-off-by: Guodong Xu <guodong.xu@linaro.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: uniphier: fix DAPLL2 clock rate of Pro5

[ Upstream commit 67affb78a4e4feb837953e3434c8402a5c3b272f ]

The parent of DAPLL2 should be DAPLL1.  Fix the clock connection.

Signed-off-by: Masahiro Yamada <yamada.masahiro@socionext.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: qcom: common: fix legacy board-clock registration

[ Upstream commit 43a51019cc8ff1b1cd2ba72e86563beb40d356fc ]

Make sure to search only the child nodes of "/clocks", rather than the
whole device-tree depth-first starting at "/clocks" when determining
whether to register a fixed clock in the legacy board-clock registration
helper.

Fixes: ee15faffef11 ("clk: qcom: common: Add API to register board clocks backwards compatibly")
Signed-off-by: Johan Hovold <johan@kernel.org>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: sunxi-ng: a83t: Fix i2c buses bits

[ Upstream commit cc54c0955d6f8618a38a999eecdc3d95306b90de ]

i2c1 and i2c2 bits for CCU are not bit 0 but bit 1 and bit 2.
Because of that, the i2c0 (bit 0) was not correctly configured.
Fixed the correct bits for i2c1 and i2c2.

Fixes: 05359be1176b ("clk: sunxi-ng: Add driver for A83T CCU")

Signed-off-by: Mylène Josserand <mylene.josserand@free-electrons.com>
Acked-by: Maxime Ripard <maxime.ripard@free-electrons.com>
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

clk: stm32h7: fix test of clock config

[ Upstream commit c1ea839c41d049604a3f64ef72712d1c7c6639d0 ]

fix test of composite clock config (bad copy / past)

Signed-off-by: Gabriel Fernandez <gabriel.fernandez@st.com>
Fixes: 3e4d618b0722 ("clk: stm32h7: Add stm32h743 clock driver")
Signed-off-by: Stephen Boyd <sboyd@codeaurora.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bpf: fix lockdep splat

[ Upstream commit 89ad2fa3f043a1e8daae193bcb5fe34d5f8caf28 ]

pcpu_freelist_pop() needs the same lockdep awareness than
pcpu_freelist_populate() to avoid a false positive.

 [ INFO: SOFTIRQ-safe -> SOFTIRQ-unsafe lock order detected ]

 switchto-defaul/12508 [HC0[0]:SC0[6]:HE0:SE0] is trying to acquire:
  (&htab->buckets[i].lock){......}, at: [<ffffffff9dc099cb>] __htab_percpu_map_update_elem+0x1cb/0x300

 and this task is already holding:
  (dev_queue->dev->qdisc_class ?: &qdisc_tx_lock#2){+.-...}, at: [<ffffffff9e135848>] __dev_queue_xmit+0
x868/0x1240
 which would create a new lock dependency:
  (dev_queue->dev->qdisc_class ?: &qdisc_tx_lock#2){+.-...} -> (&htab->buckets[i].lock){......}

 but this new dependency connects a SOFTIRQ-irq-safe lock:
  (dev_queue->dev->qdisc_class ?: &qdisc_tx_lock#2){+.-...}
 ... which became SOFTIRQ-irq-safe at:
   [<ffffffff9db5931b>] __lock_acquire+0x42b/0x1f10
   [<ffffffff9db5b32c>] lock_acquire+0xbc/0x1b0
   [<ffffffff9da05e38>] _raw_spin_lock+0x38/0x50
   [<ffffffff9e135848>] __dev_queue_xmit+0x868/0x1240
   [<ffffffff9e136240>] dev_queue_xmit+0x10/0x20
   [<ffffffff9e1965d9>] ip_finish_output2+0x439/0x590
   [<ffffffff9e197410>] ip_finish_output+0x150/0x2f0
   [<ffffffff9e19886d>] ip_output+0x7d/0x260
   [<ffffffff9e19789e>] ip_local_out+0x5e/0xe0
   [<ffffffff9e197b25>] ip_queue_xmit+0x205/0x620
   [<ffffffff9e1b8398>] tcp_transmit_skb+0x5a8/0xcb0
   [<ffffffff9e1ba152>] tcp_write_xmit+0x242/0x1070
   [<ffffffff9e1baffc>] __tcp_push_pending_frames+0x3c/0xf0
   [<ffffffff9e1b3472>] tcp_rcv_established+0x312/0x700
   [<ffffffff9e1c1acc>] tcp_v4_do_rcv+0x11c/0x200
   [<ffffffff9e1c3dc2>] tcp_v4_rcv+0xaa2/0xc30
   [<ffffffff9e191107>] ip_local_deliver_finish+0xa7/0x240
   [<ffffffff9e191a36>] ip_local_deliver+0x66/0x200
   [<ffffffff9e19137d>] ip_rcv_finish+0xdd/0x560
   [<ffffffff9e191e65>] ip_rcv+0x295/0x510
   [<ffffffff9e12ff88>] __netif_receive_skb_core+0x988/0x1020
   [<ffffffff9e130641>] __netif_receive_skb+0x21/0x70
   [<ffffffff9e1306ff>] process_backlog+0x6f/0x230
   [<ffffffff9e132129>] net_rx_action+0x229/0x420
   [<ffffffff9da07ee8>] __do_softirq+0xd8/0x43d
   [<ffffffff9e282bcc>] do_softirq_own_stack+0x1c/0x30
   [<ffffffff9dafc2f5>] do_softirq+0x55/0x60
   [<ffffffff9dafc3a8>] __local_bh_enable_ip+0xa8/0xb0
   [<ffffffff9db4c727>] cpu_startup_entry+0x1c7/0x500
   [<ffffffff9daab333>] start_secondary+0x113/0x140

 to a SOFTIRQ-irq-unsafe lock:
  (&head->lock){+.+...}
 ... which became SOFTIRQ-irq-unsafe at:
 ...  [<ffffffff9db5971f>] __lock_acquire+0x82f/0x1f10
   [<ffffffff9db5b32c>] lock_acquire+0xbc/0x1b0
   [<ffffffff9da05e38>] _raw_spin_lock+0x38/0x50
   [<ffffffff9dc0b7fa>] pcpu_freelist_pop+0x7a/0xb0
   [<ffffffff9dc08b2c>] htab_map_alloc+0x50c/0x5f0
   [<ffffffff9dc00dc5>] SyS_bpf+0x265/0x1200
   [<ffffffff9e28195f>] entry_SYSCALL_64_fastpath+0x12/0x17

 other info that might help us debug this:

 Chain exists of:
   dev_queue->dev->qdisc_class ?: &qdisc_tx_lock#2 --> &htab->buckets[i].lock --> &head->lock

  Possible interrupt unsafe locking scenario:

        CPU0                    CPU1
        ----                    ----
   lock(&head->lock);
                                local_irq_disable();
                                lock(dev_queue->dev->qdisc_class ?: &qdisc_tx_lock#2);
                                lock(&htab->buckets[i].lock);
   <Interrupt>
     lock(dev_queue->dev->qdisc_class ?: &qdisc_tx_lock#2);

  *** DEADLOCK ***

Fixes: e19494edab82 ("bpf: introduce percpu_freelist")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

geneve: fix fill_info when link down

[ Upstream commit fd7eafd02121d6ef501ef1a4a891e6061366c952 ]

geneve->sock4/6 were added with geneve_open and released with geneve_stop.
So when geneve link down, we will not able to show remote address and
checksum info after commit 11387fe4a98 ("geneve: fix fill_info when using
collect_metadata").

Fix this by avoid passing *_REMOTE{,6} for COLLECT_METADATA since they are
mutually exclusive, and always show UDP_ZERO_CSUM6_RX info.

Fixes: 11387fe4a98 ("geneve: fix fill_info when using collect_metadata")
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fcntl: don't leak fd reference when fixup_compat_flock fails

[ Upstream commit 9280a601e6080c9ff658468c1c775ff6514099a6 ]

Currently we just return err here, but we need to put the fd reference
first.

Fixes: 94073ad77fff (fs/locks: don't mess with the address limit in compat_fcntl64)
Signed-off-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

sctp: use the right sk after waking up from wait_buf sleep

[ Upstream commit cea0cc80a6777beb6eb643d4ad53690e1ad1d4ff ]

Commit dfcb9f4f99f1 ("sctp: deny peeloff operation on asocs with threads
sleeping on it") fixed the race between peeloff and wait sndbuf by
checking waitqueue_active(&asoc->wait) in sctp_do_peeloff().

But it actually doesn't work, as even if waitqueue_active returns false
the waiting sndbuf thread may still not yet hold sk lock. After asoc is
peeled off, sk is not asoc->base.sk any more, then to hold the old sk
lock couldn't make assoc safe to access.

This patch is to fix this by changing to hold the new sk lock if sk is
not asoc->base.sk, meanwhile, also set the sk in sctp_sendmsg with the
new sk.

With this fix, there is no more race between peeloff and waitbuf, the
check 'waitqueue_active' in sctp_do_peeloff can be removed.

Thanks Marcelo and Neil for making this clear.

v1->v2:
  fix it by changing to lock the new sock instead of adding a flag in asoc.

Suggested-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

sctp: do not free asoc when it is already dead in sctp_sendmsg

[ Upstream commit ca3af4dd28cff4e7216e213ba3b671fbf9f84758 ]

Now in sctp_sendmsg sctp_wait_for_sndbuf could schedule out without
holding sock sk. It means the current asoc can be freed elsewhere,
like when receiving an abort packet.

If the asoc is just created in sctp_sendmsg and sctp_wait_for_sndbuf
returns err, the asoc will be freed again due to new_asoc is not nil.
An use-after-free issue would be triggered by this.

This patch is to fix it by setting new_asoc with nil if the asoc is
already dead when cpu schedules back, so that it will not be freed
again in sctp_sendmsg.

v1->v2:
  set new_asoc as nil in sctp_sendmsg instead of sctp_wait_for_sndbuf.

Suggested-by: Neil Horman <nhorman@tuxdriver.com>
Reported-by: Dmitry Vyukov <dvyukov@google.com>
Signed-off-by: Xin Long <lucien.xin@gmail.com>
Acked-by: Neil Horman <nhorman@tuxdriver.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

slub: fix sysfs duplicate filename creation when slub_debug=O

[ Upstream commit 11066386efa692f77171484c32ea30f6e5a0d729 ]

When slub_debug=O is set.  It is possible to clear debug flags for an
"unmergeable" slab cache in kmem_cache_open().  It makes the "unmergeable"
cache became "mergeable" in sysfs_slab_add().

These caches will generate their "unique IDs" by create_unique_id(), but
it is possible to create identical unique IDs.  In my experiment,
sgpool-128, names_cache, biovec-256 generate the same ID ":Ft-0004096" and
the kernel reports "sysfs: cannot create duplicate filename
'/kernel/slab/:Ft-0004096'".

To repeat my experiment, set disable_higher_order_debug=1,
CONFIG_SLUB_DEBUG_ON=y in kernel-4.14.

Fix this issue by setting unmergeable=1 if slub_debug=O and the the
default slub_debug contains any no-merge flags.

call path:
kmem_cache_create()
  __kmem_cache_alias() -> we set SLAB_NEVER_MERGE flags here
  create_cache()
    __kmem_cache_create()
      kmem_cache_open() -> clear DEBUG_METADATA_FLAGS
      sysfs_slab_add() -> the slab cache is mergeable now

  sysfs: cannot create duplicate filename '/kernel/slab/:Ft-0004096'
  ------------[ cut here ]------------
  WARNING: CPU: 0 PID: 1 at fs/sysfs/dir.c:31 sysfs_warn_dup+0x60/0x7c
  Modules linked in:
  CPU: 0 PID: 1 Comm: swapper/0 Tainted: G        W       4.14.0-rc7ajb-00131-gd4c2e9f-dirty #123
  Hardware name: linux,dummy-virt (DT)
  task: ffffffc07d4e0080 task.stack: ffffff8008008000
  PC is at sysfs_warn_dup+0x60/0x7c
  LR is at sysfs_warn_dup+0x60/0x7c
  pc :  lr :  pstate: 60000145
  Call trace:
   sysfs_warn_dup+0x60/0x7c
   sysfs_create_dir_ns+0x98/0xa0
   kobject_add_internal+0xa0/0x294
   kobject_init_and_add+0x90/0xb4
   sysfs_slab_add+0x90/0x200
   __kmem_cache_create+0x26c/0x438
   kmem_cache_create+0x164/0x1f4
   sg_pool_init+0x60/0x100
   do_one_initcall+0x38/0x12c
   kernel_init_freeable+0x138/0x1d4
   kernel_init+0x10/0xfc
   ret_from_fork+0x10/0x18

Link: http://lkml.kernel.org/r/1510365805-5155-1-git-send-email-miles.chen@mediatek.com
Signed-off-by: Miles Chen <miles.chen@mediatek.com>
Acked-by: Christoph Lameter <cl@linux.com>
Cc: Pekka Enberg <penberg@kernel.org>
Cc: David Rientjes <rientjes@google.com>
Cc: Joonsoo Kim <iamjoonsoo.kim@lge.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <alexander.levin@verizon.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>