linux.git/drivers/infiniband, branch v4.4.181

RDMA/cxgb4: Fix null pointer dereference on alloc_skb failure

2019-06-11T10:23:59+00:00

[ Upstream commit a6d2a5a92e67d151c98886babdc86d530d27111c ]

Currently if alloc_skb fails to allocate the skb a null skb is passed to
t4_set_arp_err_handler and this ends up dereferencing the null skb.  Avoid
the NULL pointer dereference by checking for a NULL skb and returning
early.

Addresses-Coverity: ("Dereference null return")
Fixes: b38a0ad8ec11 ("RDMA/cxgb4: Set arp error handler for PASS_ACCEPT_RPL messages")
Signed-off-by: Colin Ian King 
Acked-by: Potnuri Bharat Teja 
Signed-off-by: Jason Gunthorpe 
Signed-off-by: Sasha Levin

IB/mlx4: Fix race condition between catas error reset and aliasguid flows

2019-04-27T07:33:56+00:00

[ Upstream commit 587443e7773e150ae29e643ee8f41a1eed226565 ]

Code review revealed a race condition which could allow the catas error
flow to interrupt the alias guid query post mechanism at random points.
Thiis is fixed by doing cancel_delayed_work_sync() instead of
cancel_delayed_work() during the alias guid mechanism destroy flow.

Fixes: a0c64a17aba8 ("mlx4: Add alias_guid mechanism")
Signed-off-by: Jack Morgenstein 
Signed-off-by: Leon Romanovsky 
Signed-off-by: Jason Gunthorpe 
Signed-off-by: Sasha Levin

IB/mlx4: Increase the timeout for CM cache

2019-04-27T07:33:50+00:00

[ Upstream commit 2612d723aadcf8281f9bf8305657129bd9f3cd57 ]

Using CX-3 virtual functions, either from a bare-metal machine or
pass-through from a VM, MAD packets are proxied through the PF driver.

Since the VF drivers have separate name spaces for MAD Transaction Ids
(TIDs), the PF driver has to re-map the TIDs and keep the book keeping
in a cache.

Following the RDMA Connection Manager (CM) protocol, it is clear when
an entry has to evicted form the cache. But life is not perfect,
remote peers may die or be rebooted. Hence, it's a timeout to wipe out
a cache entry, when the PF driver assumes the remote peer has gone.

During workloads where a high number of QPs are destroyed concurrently,
excessive amount of CM DREQ retries has been observed

The problem can be demonstrated in a bare-metal environment, where two
nodes have instantiated 8 VFs each. This using dual ported HCAs, so we
have 16 vPorts per physical server.

64 processes are associated with each vPort and creates and destroys
one QP for each of the remote 64 processes. That is, 1024 QPs per
vPort, all in all 16K QPs. The QPs are created/destroyed using the
CM.

When tearing down these 16K QPs, excessive CM DREQ retries (and
duplicates) are observed. With some cat/paste/awk wizardry on the
infiniband_cm sysfs, we observe as sum of the 16 vPorts on one of the
nodes:

cm_rx_duplicates:
      dreq  2102
cm_rx_msgs:
      drep  1989
      dreq  6195
       rep  3968
       req  4224
       rtu  4224
cm_tx_msgs:
      drep  4093
      dreq 27568
       rep  4224
       req  3968
       rtu  3968
cm_tx_retries:
      dreq 23469

Note that the active/passive side is equally distributed between the
two nodes.

Enabling pr_debug in cm.c gives tons of:

[171778.814239]  mlx4_ib_multiplex_cm_handler: id{slave:
1,sl_cm_id: 0xd393089f} is NULL!

By increasing the CM_CLEANUP_CACHE_TIMEOUT from 5 to 30 seconds, the
tear-down phase of the application is reduced from approximately 90 to
50 seconds. Retries/duplicates are also significantly reduced:

cm_rx_duplicates:
      dreq  2460
[]
cm_tx_retries:
      dreq  3010
       req    47

Increasing the timeout further didn't help, as these duplicates and
retries stems from a too short CMA timeout, which was 20 (~4 seconds)
on the systems. By increasing the CMA timeout to 22 (~17 seconds), the
numbers fell down to about 10 for both of them.

Adjustment of the CMA timeout is not part of this commit.

Signed-off-by: Håkon Bugge 
Acked-by: Jack Morgenstein 
Signed-off-by: Jason Gunthorpe 
Signed-off-by: Sasha Levin

IB/{hfi1, qib}: Fix WC.byte_len calculation for UD_SEND_WITH_IMM

2019-03-23T07:44:24+00:00

[ Upstream commit 904bba211acc2112fdf866e5a2bc6cd9ecd0de1b ]

The work completion length for a receiving a UD send with immediate is
short by 4 bytes causing application using this opcode to fail.

The UD receive logic incorrectly subtracts 4 bytes for immediate
value. These bytes are already included in header length and are used to
calculate header/payload split, so the result is these 4 bytes are
subtracted twice, once when the header length subtracted from the overall
length and once again in the UD opcode specific path.

Remove the extra subtraction when handling the opcode.

Fixes: 7724105686e7 ("IB/hfi1: add driver files")
Reviewed-by: Michael J. Ruhl 
Signed-off-by: Brian Welty 
Signed-off-by: Mike Marciniszyn 
Signed-off-by: Dennis Dalessandro 
Signed-off-by: Jason Gunthorpe 
Signed-off-by: Sasha Levin

RDMA/srp: Rework SCSI device reset handling

2019-03-23T07:44:17+00:00

commit 48396e80fb6526ea5ed267bd84f028bae56d2f9e upstream.

Since .scsi_done() must only be called after scsi_queue_rq() has
finished, make sure that the SRP initiator driver does not call
.scsi_done() while scsi_queue_rq() is in progress. Although
invoking sg_reset -d while I/O is in progress works fine with kernel
v4.20 and before, that is not the case with kernel v5.0-rc1. This
patch avoids that the following crash is triggered with kernel
v5.0-rc1:

BUG: unable to handle kernel NULL pointer dereference at 0000000000000138
CPU: 0 PID: 360 Comm: kworker/0:1H Tainted: G    B             5.0.0-rc1-dbg+ #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
Workqueue: kblockd blk_mq_run_work_fn
RIP: 0010:blk_mq_dispatch_rq_list+0x116/0xb10
Call Trace:
 blk_mq_sched_dispatch_requests+0x2f7/0x300
 __blk_mq_run_hw_queue+0xd6/0x180
 blk_mq_run_work_fn+0x27/0x30
 process_one_work+0x4f1/0xa20
 worker_thread+0x67/0x5b0
 kthread+0x1cf/0x1f0
 ret_from_fork+0x24/0x30

Cc: 
Fixes: 94a9174c630c ("IB/srp: reduce lock coverage of command completion")
Signed-off-by: Bart Van Assche 
Signed-off-by: Jason Gunthorpe 
Signed-off-by: Greg Kroah-Hartman

mm: replace get_user_pages() write/force parameters with gup_flags

2018-12-17T20:55:16+00:00

commit 768ae309a96103ed02eb1e111e838c87854d8b51 upstream.

This removes the 'write' and 'force' from get_user_pages() and replaces
them with 'gup_flags' to make the use of FOLL_FORCE explicit in callers
as use of this flag can result in surprising behaviour (and hence bugs)
within the mm subsystem.

Signed-off-by: Lorenzo Stoakes 
Acked-by: Christian König 
Acked-by: Jesper Nilsson 
Acked-by: Michal Hocko 
Reviewed-by: Jan Kara 
Signed-off-by: Linus Torvalds 
[bwh: Backported to 4.4:
 - Drop changes in rapidio, vchiq, goldfish
 - Keep the "write" variable in amdgpu_ttm_tt_pin_userptr() as it's still
   needed
 - Also update calls from various other places that now use
   get_user_pages_remote() upstream, which were updated there by commit
   9beae1ea8930 "mm: replace get_user_pages_remote() write/force ..."
 - Also update calls from hfi1 and ipath
 - Adjust context]
Signed-off-by: Ben Hutchings 
Signed-off-by: Greg Kroah-Hartman

iser: set sector for ambiguous mr status errors

2018-12-13T08:21:29+00:00

commit 24c3456c8d5ee6fc1933ca40f7b4406130682668 upstream.

If for some reason we failed to query the mr status, we need to make sure
to provide sufficient information for an ambiguous error (guard error on
sector 0).

Fixes: 0a7a08ad6f5f ("IB/iser: Implement check_protection")
Cc: 
Reported-by: Dan Carpenter 
Signed-off-by: Sagi Grimberg 
Signed-off-by: Jason Gunthorpe 
Signed-off-by: Greg Kroah-Hartman

IB/ucm: Fix Spectre v1 vulnerability

2018-11-10T15:41:42+00:00

commit 0295e39595e1146522f2722715dba7f7fba42217 upstream.

hdr.cmd can be indirectly controlled by user-space, hence leading to
a potential exploitation of the Spectre variant 1 vulnerability.

This issue was detected with the help of Smatch:

drivers/infiniband/core/ucm.c:1127 ib_ucm_write() warn: potential
spectre issue 'ucm_cmd_table' [r] (local cap)

Fix this by sanitizing hdr.cmd before using it to index
ucm_cmd_table.

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

[1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2

Cc: stable@vger.kernel.org
Signed-off-by: Gustavo A. R. Silva 
Signed-off-by: Doug Ledford 
Signed-off-by: Greg Kroah-Hartman

RDMA/ucma: Fix Spectre v1 vulnerability

2018-11-10T15:41:42+00:00

commit a3671a4f973ee9d9621d60166cc3b037c397d604 upstream.

hdr.cmd can be indirectly controlled by user-space, hence leading to
a potential exploitation of the Spectre variant 1 vulnerability.

This issue was detected with the help of Smatch:

drivers/infiniband/core/ucma.c:1686 ucma_write() warn: potential
spectre issue 'ucma_cmd_table' [r] (local cap)

Fix this by sanitizing hdr.cmd before using it to index
ucm_cmd_table.

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

[1] https://marc.info/?l=linux-kernel&m=152449131114778&w=2

Cc: stable@vger.kernel.org
Signed-off-by: Gustavo A. R. Silva 
Signed-off-by: Doug Ledford 
Signed-off-by: Greg Kroah-Hartman

ucma: fix a use-after-free in ucma_resolve_ip()

2018-10-13T07:11:34+00:00

commit 5fe23f262e0548ca7f19fb79f89059a60d087d22 upstream.

There is a race condition between ucma_close() and ucma_resolve_ip():

CPU0				CPU1
ucma_resolve_ip():		ucma_close():

ctx = ucma_get_ctx(file, cmd.id);

        list_for_each_entry_safe(ctx, tmp, &file->ctx_list, list) {
                mutex_lock(&mut);
                idr_remove(&ctx_idr, ctx->id);
                mutex_unlock(&mut);
		...
                mutex_lock(&mut);
                if (!ctx->closing) {
                        mutex_unlock(&mut);
                        rdma_destroy_id(ctx->cm_id);
		...
                ucma_free_ctx(ctx);

ret = rdma_resolve_addr();
ucma_put_ctx(ctx);

Before idr_remove(), ucma_get_ctx() could still find the ctx
and after rdma_destroy_id(), rdma_resolve_addr() may still
access id_priv pointer. Also, ucma_put_ctx() may use ctx after
ucma_free_ctx() too.

ucma_close() should call ucma_put_ctx() too which tests the
refcnt and waits for the last one releasing it. The similar
pattern is already used by ucma_destroy_id().

Reported-and-tested-by: syzbot+da2591e115d57a9cbb8b@syzkaller.appspotmail.com
Reported-by: syzbot+cfe3c1e8ef634ba8964b@syzkaller.appspotmail.com
Cc: Jason Gunthorpe 
Cc: Doug Ledford 
Cc: Leon Romanovsky 
Signed-off-by: Cong Wang 
Reviewed-by: Leon Romanovsky 
Signed-off-by: Doug Ledford 
Signed-off-by: Greg Kroah-Hartman