From: Mark Zhang <markz@xxxxxxxxxxxx>
When read the global counter and there's any dynamic counter allocated,
the value of a hwcounter is the sum of the default counter and all
dynamic counters. So the number of hwcounters of a dynamically allocated
counter must be same as of the default counter, otherwise there will
be read violations.
This fixes the KASAN slab-out-of-bounds bug:
[ 1736.292832] ==================================================================
[ 1736.294938] BUG: KASAN: slab-out-of-bounds in rdma_counter_get_hwstat_value+0x36d/0x390 [ib_core]
[ 1736.297071] Read of size 8 at addr ffff8884192a5778 by task rdma/10138
[ 1736.298462]
[ 1736.299257] CPU: 7 PID: 10138 Comm: rdma Not tainted 5.5.0-for-upstream-dbg-2020-02-06_18-30-19-27 #1
[ 1736.301476] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.12.1-0-ga5cab58e9a3f-prebuilt.qemu.org 04/01/2014
[ 1736.303923] Call Trace:
[ 1736.304771] dump_stack+0xb7/0x10b
[ 1736.305750] print_address_description.constprop.4+0x1e2/0x400
[ 1736.307072] ? rdma_counter_get_hwstat_value+0x36d/0x390 [ib_core]
[ 1736.308429] __kasan_report+0x15c/0x1e0
[ 1736.309467] ? mlx5_ib_query_q_counters+0x13f/0x270 [mlx5_ib]
[ 1736.310780] ? rdma_counter_get_hwstat_value+0x36d/0x390 [ib_core]
[ 1736.312139] kasan_report+0xe/0x20
[ 1736.313124] rdma_counter_get_hwstat_value+0x36d/0x390 [ib_core]
[ 1736.314477] ? rdma_counter_query_stats+0xd0/0xd0 [ib_core]
[ 1736.315742] ? memcpy+0x34/0x50
[ 1736.316683] ? nla_put+0xe2/0x170
[ 1736.317665] nldev_stat_get_doit+0x9c7/0x14f0 [ib_core]
...
[ 1736.350888] do_syscall_64+0x95/0x490
[ 1736.351901] entry_SYSCALL_64_after_hwframe+0x49/0xbe
[ 1736.353095] RIP: 0033:0x7fcc457fe65a
[ 1736.354089] Code: bb 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 8b 05 fa f1 2b 00 45 89 c9 4c 63 d1 48 63 ff 85 c0 75 15 b8 2c 00 00 00 0f 05 <48> 3d 00 f0 ff ff 77 76 f3 c3 0f 1f 40 00 41 55 41 54 4d 89 c5 55
[ 1736.358062] RSP: 002b:00007ffc0586f868 EFLAGS: 00000246 ORIG_RAX: 000000000000002c
[ 1736.360021] RAX: ffffffffffffffda RBX: 0000000000000000 RCX: 00007fcc457fe65a
[ 1736.361490] RDX: 0000000000000020 RSI: 00000000013db920 RDI: 0000000000000003
[ 1736.362967] RBP: 00007ffc0586fa90 R08: 00007fcc45ac10e0 R09: 000000000000000c
[ 1736.364438] R10: 0000000000000000 R11: 0000000000000246 R12: 00000000004089c0
[ 1736.365909] R13: 0000000000000000 R14: 00007ffc0586fab0 R15: 00000000013dc9a0
[ 1736.367423]
[ 1736.368139] Allocated by task 9700:
[ 1736.369129] save_stack+0x19/0x80
[ 1736.370091] __kasan_kmalloc.constprop.7+0xa0/0xd0
[ 1736.371261] mlx5_ib_counter_alloc_stats+0xd1/0x1d0 [mlx5_ib]
[ 1736.372557] rdma_counter_alloc+0x16d/0x3f0 [ib_core]
[ 1736.373758] rdma_counter_bind_qpn_alloc+0x216/0x4e0 [ib_core]
[ 1736.375067] nldev_stat_set_doit+0x8c2/0xb10 [ib_core]
[ 1736.376283] rdma_nl_rcv_msg+0x3d2/0x730 [ib_core]
[ 1736.377451] rdma_nl_rcv+0x2a8/0x400 [ib_core]
[ 1736.378565] netlink_unicast+0x448/0x620
[ 1736.379605] netlink_sendmsg+0x731/0xd10
[ 1736.380649] sock_sendmsg+0xb1/0xf0
[ 1736.381635] __sys_sendto+0x25d/0x2c0
[ 1736.382647] __x64_sys_sendto+0xdd/0x1b0
[ 1736.383694] do_syscall_64+0x95/0x490
[ 1736.384700] entry_SYSCALL_64_after_hwframe+0x49/0xbe
[ 1736.385897]
...
[ 1736.409522]
[ 1736.410242] The buggy address belongs to the object at ffff8884192a5600
[ 1736.410242] which belongs to the cache kmalloc-512 of size 512
[ 1736.412928] The buggy address is located 376 bytes inside of
[ 1736.412928] 512-byte region [ffff8884192a5600, ffff8884192a5800)
[ 1736.415511] The buggy address belongs to the page:
[ 1736.416693] page:ffffea001064a800 refcount:1 head refcount:1 mapcount:0 mapping:ffff88841c011300 index:0x0 compound_mapcount:0 compound_pincount:0
[ 1736.419474] raw: 002fffff80010200 ffffea0010173208 ffffea0010034808 ffff88841c011300
[ 1736.421472] raw: 0000000000000000 0000000000150015 00000001ffffffff 0000000000000000
[ 1736.423492] page dumped because: kasan: bad access detected
[ 1736.424787]
[ 1736.425508] Memory state around the buggy address:
[ 1736.426694] ffff8884192a5600: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 1736.428648] ffff8884192a5680: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
[ 1736.430610] >ffff8884192a5700: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 fc
[ 1736.432556] ^
[ 1736.434071] ffff8884192a5780: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[ 1736.436027] ffff8884192a5800: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
[ 1736.437987] ==================================================================
[ 1736.439939] Disabling lock debugging due to kernel taint
Fixes: 18d422ce8ccf ("IB/mlx5: Add counter_alloc_stats() and counter_update_stats() support")
Signed-off-by: Mark Zhang <markz@xxxxxxxxxxxx>
Signed-off-by: Leon Romanovsky <leonro@xxxxxxxxxxxx>
---
drivers/infiniband/hw/mlx5/main.c | 5 +++--
1 file changed, 3 insertions(+), 2 deletions(-)
diff --git a/drivers/infiniband/hw/mlx5/main.c b/drivers/infiniband/hw/mlx5/main.c
index 4d0780d9114c..1279aeabf651 100644
--- a/drivers/infiniband/hw/mlx5/main.c
+++ b/drivers/infiniband/hw/mlx5/main.c
@@ -5720,9 +5720,10 @@ mlx5_ib_counter_alloc_stats(struct rdma_counter *counter)
const struct mlx5_ib_counters *cnts =
get_counters(dev, counter->port - 1);
- /* Q counters are in the beginning of all counters */
return rdma_alloc_hw_stats_struct(cnts->names,
- cnts->num_q_counters,
+ cnts->num_q_counters +
+ cnts->num_cong_counters +
+ cnts->num_ext_ppcnt_counters,
RDMA_HW_STATS_DEFAULT_LIFESPAN);
}
--
2.24.1
