net/mlx4_core: Fix qp mtt size calculation
authorJack Morgenstein <jackm@dev.mellanox.co.il>
Tue, 12 Mar 2019 15:05:49 +0000 (17:05 +0200)
committerDavid S. Miller <davem@davemloft.net>
Tue, 12 Mar 2019 22:00:15 +0000 (15:00 -0700)
Calculation of qp mtt size (in function mlx4_RST2INIT_wrapper)
ultimately depends on function roundup_pow_of_two.

If the amount of memory required by the QP is less than one page,
roundup_pow_of_two is called with argument zero.  In this case, the
roundup_pow_of_two result is undefined.

Calling roundup_pow_of_two with a zero argument resulted in the
following stack trace:

UBSAN: Undefined behaviour in ./include/linux/log2.h:61:13
shift exponent 64 is too large for 64-bit type 'long unsigned int'
CPU: 4 PID: 26939 Comm: rping Tainted: G OE 4.19.0-rc1
Hardware name: Supermicro X9DR3-F/X9DR3-F, BIOS 3.2a 07/09/2015
Call Trace:
dump_stack+0x9a/0xeb
ubsan_epilogue+0x9/0x7c
__ubsan_handle_shift_out_of_bounds+0x254/0x29d
? __ubsan_handle_load_invalid_value+0x180/0x180
? debug_show_all_locks+0x310/0x310
? sched_clock+0x5/0x10
? sched_clock+0x5/0x10
? sched_clock_cpu+0x18/0x260
? find_held_lock+0x35/0x1e0
? mlx4_RST2INIT_QP_wrapper+0xfb1/0x1440 [mlx4_core]
mlx4_RST2INIT_QP_wrapper+0xfb1/0x1440 [mlx4_core]

Fix this by explicitly testing for zero, and returning one if the
argument is zero (assuming that the next higher power of 2 in this case
should be one).

Fixes: c82e9aa0a8bc ("mlx4_core: resource tracking for HCA resources used by guests")
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il>
Signed-off-by: Tariq Toukan <tariqt@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/mellanox/mlx4/resource_tracker.c

index eb13d361816248a7eaea579da55e44df1d4124f2..4356f3a580027602b84f7d550b3b56244e757307 100644 (file)
@@ -2719,13 +2719,13 @@ static int qp_get_mtt_size(struct mlx4_qp_context *qpc)
        int total_pages;
        int total_mem;
        int page_offset = (be32_to_cpu(qpc->params2) >> 6) & 0x3f;
+       int tot;
 
        sq_size = 1 << (log_sq_size + log_sq_sride + 4);
        rq_size = (srq|rss|xrc) ? 0 : (1 << (log_rq_size + log_rq_stride + 4));
        total_mem = sq_size + rq_size;
-       total_pages =
-               roundup_pow_of_two((total_mem + (page_offset << 6)) >>
-                                  page_shift);
+       tot = (total_mem + (page_offset << 6)) >> page_shift;
+       total_pages = !tot ? 1 : roundup_pow_of_two(tot);
 
        return total_pages;
 }