s390/crc32-vx: Fix checksum calculation for small sizes

The current prealign logic will fail for sizes < alignment,
as the new datalen passed to the vector function is smaller
than zero. Being a size_t this gets wrapped to a huge
number causing memory overruns and wrong data.

Let's add an early exit if the size is smaller than the minimal
size with alignment. This will also avoid calling the software
fallback twice for all sizes smaller than the minimum size
(prealign + remaining)

Signed-off-by: Christian Borntraeger <borntraeger@de.ibm.com>
Fixes: f848dbd3bc ("s390/crc32-vx: add crypto API module for optimized CRC-32 algorithms")
Reviewed-by: Hendrik Brueckner <brueckner@linux.vnet.ibm.com>
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
This commit is contained in:
Christian Borntraeger 2016-08-03 17:55:00 +02:00 committed by Martin Schwidefsky
parent bf47dc572a
commit 134a24cd89

View File

@ -51,6 +51,9 @@ u32 crc32c_le_vgfm_16(u32 crc, unsigned char const *buf, size_t size);
struct kernel_fpu vxstate; \
unsigned long prealign, aligned, remaining; \
\
if (datalen < VX_MIN_LEN + VX_ALIGN_MASK) \
return ___crc32_sw(crc, data, datalen); \
\
if ((unsigned long)data & VX_ALIGN_MASK) { \
prealign = VX_ALIGNMENT - \
((unsigned long)data & VX_ALIGN_MASK); \
@ -59,9 +62,6 @@ u32 crc32c_le_vgfm_16(u32 crc, unsigned char const *buf, size_t size);
data = (void *)((unsigned long)data + prealign); \
} \
\
if (datalen < VX_MIN_LEN) \
return ___crc32_sw(crc, data, datalen); \
\
aligned = datalen & ~VX_ALIGN_MASK; \
remaining = datalen & VX_ALIGN_MASK; \
\