Re: linux-next: Tree for March 20 - crypto crash on s390

[Heiko Carstens]

I get the crash below when I compile hardware support for SHA512
(CONFIG_CRYPTO_SHA512_S390=y) into the kernel:

Unable to handle kernel pointer dereference at virtual kernel address 000000003f9b0000
Oops: 0011 [#1] PREEMPT SMP DEBUG_PAGEALLOC 
Modules linked in: 
CPU: 1 Not tainted 2.6.29-rc8-next-20090320-dirty #18 
Process cryptomgr_test (pid: 169, task: 000000003f8326a0, ksp: 000000003fb83d78)
 
Krnl PSW : 0704000180000000 0000000000032b80 (s390_sha_final+0x198/0x1d4) 
           R:0 T:1 IO:1 EX:1 Key:0 M:1 W:0 P:0 AS:0 CC:0 PM:0 EA:3 
Krnl GPRS: 0000000000000003 00000000fffffff3 000000003f9affa4 0000000000fffffc 
           0000000000000080 0000000000000003 0000000000000000 000000003fb83d98 
           000000000060a6f0 000000003f9afc48 0000000000000000 000000003fb83b50 
           000000003f9afc58 0000000000403fb8 00000000001c4152 000000003fb83b50 
Krnl Code: 0000000000032b76: a7f4ff60           brc     15,32a36 
           0000000000032b7a: 1843               lr      %rr4,%rr3 
           0000000000032b7c: a7f4ffbb           brc     15,32af2 
          >0000000000032b80: d7ff20002000       xc      0(256,%rr2),0(%rr2) 
           0000000000032b86: 41202100           la      %rr2,256(%rr2) 
           0000000000032b8a: a737fffb           brctg   %rr3,32b80 
           0000000000032b8e: a7f4ff83           brc     15,32a94 
           0000000000032b92: d2ff30002000       mvc     0(256,%rr3),0(%rr2) 
Call Trace: 
([<000000003fb83c40>] 0x3fb83c40)
 [<00000000001c4152>] crypto_shash_final+0x46/0x70 
 [<00000000001c41f4>] shash_async_final+0x34/0x48 
 [<00000000001c47ce>] shash_async_digest+0x12a/0x158 
 [<00000000001c615e>] alg_test_hash+0x1ae/0x634 
 [<00000000001c802e>] alg_test+0xa2/0x210 
 [<00000000001c51d8>] cryptomgr_test+0x6c/0x70 
 [<0000000000067d4a>] kthread+0x76/0xbc 
 [<000000000001befa>] kernel_thread_starter+0x6/0xc 
 [<000000000001bef4>] kernel_thread_starter+0x0/0xc 
Last Breaking-Event-Address: 
 [<0000000000032b8a>] s390_sha_final+0x1a2/0x1d4 
 <4>---[ end trace 561bb236c800851f ]--- 

The problem here seems to be that
int s390_sha_final(struct shash_desc *desc, u8 *out)

gets called with an shash_desc where

bsize = crypto_shash_blocksize(desc->tfm);

results in bsize with zero.

Later in s390_sha_final we see this:

/* pad with zeros */
memset(ctx->buf + index, 0x00, end - index - 8);

because of bsize being zero the result of "end - index -8" is -12.
Obviously it isn't a very good thing to pass a length of -12 to memset.

I have no idea where the bug is: either the s390 arch specific part is
broken because it cannot handle zero sized block sizes or the common
code is broken because it even calls this function with a size of zero.

Jan, Herbert?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/