On Mon, Sep 28, 2009 at 5:31 PM, avital sela <[email protected]> wrote:
>
> Hi Herbert,
>
> This is the error message I'm getting with the unmodifed testmgr:
>
> alg: hash: Test 3 failed for hmac(sha1-hw)
>
> 00000000: d6 bf 3f 79 2a 14 11 24 f0 5c 3b 80 7e 81 08 fe
>
> 00000010: 09 ba 7a fb
>
> Attached below is the source code for the sha driver.? Like I said previously, if I add printk-s or delays in testmgr then the test passes and all subsequent sha1 crypto operations are successful.
>
> Thanks for your help,
> Avital
>
> static void hw_sha1_init(struct crypto_tfm *tfm)
> {
> ??? struct hw_sha1_ctx??? ??? ??? ??? ??? *sctx = crypto_tfm_ctx(tfm);
> ??? unsigned long ??? ??? ??? ??? ??? ??? sha_cmd;
> ??? static const struct hw_sha1_ctx initstate = {
> ??? ? 0,
> ??? ? { SHA1_H0, SHA1_H1, SHA1_H2, SHA1_H3, SHA1_H4 },
> ??? ? { 0, }
> ??? };
>
> ??? *sctx = initstate;
>
> ??? /* construct SHA HW command */
> ??? sha_cmd =??? SHA_MEM_MODE_DMA_CPU2K???? |
> ??? ??? ??? ??? SHA_IPSEC_OP_SEL_ENCRYPT?? |
> ??? ??? ??? ??? SHA_MODE_SHA1????????????? |
> ??? ??? ??? ??? SHA_REV_BIT_OFF??????????? |
> ??? ??? ??? ??? (SET_SHA_CONTROL_REG_END_SEL(0x77))?? ;
> ?? SHA_HW_WRITE(SHA_CONTROL_REG,sha_cmd);
>
> }
>
> static void hw_sha1_update(struct crypto_tfm *tfm, const u8 *data,unsigned int len)
> {
> ?? struct hw_sha1_ctx??? ??? *sctx;
>
> ?? /* copy the new chunk of data to the HW buffer */
> ?? sctx = crypto_tfm_ctx(tfm);
> ?? memcpy((void*)(SHA_DATA_BUFFER + (unsigned long)sctx->count),(void*)data,len);
> ?? sctx->count += len;
> }
>
> static void hw_sha1_final(struct crypto_tfm *tfm, u8 *out)
> {
> ?? unsigned long??? ??? ??? length;
> ?? struct hw_sha1_ctx??? ??? *sctx;
>
> ?? sctx = crypto_tfm_ctx(tfm);
>
> ?? SHA_HW_WRITE(SHA_MESSAGE_SIZE_REG,sctx->count);
> ?? SHA_HW_WRITE(SHA_START_REG,0x0);
> ?? SHA_HW_WAIT_FOR_DONE();
> ?? memcpy((void*)out,(void*)SHA_DATA_BUFFER,20);
>
> ?? /* Wipe context */
> ?? memset(sctx, 0, sizeof *sctx);
> }
>
> static struct crypto_alg hw_sha1_alg = {
> ??? .cra_name??? ??? =??? "sha1",
> ??? .cra_driver_name=??? "sha1-hw",
> ??? .cra_flags??? ??? =??? CRYPTO_ALG_TYPE_DIGEST,
> ??? .cra_priority??? =??? 300,
> ??? .cra_blocksize??? =??? SHA1_BLOCK_SIZE,
> ??? .cra_ctxsize??? =??? sizeof(struct hw_sha1_ctx),
> ??? .cra_module??? ??? =??? THIS_MODULE,
> ??? .cra_list?????? =?? LIST_HEAD_INIT(hw_sha1_alg.cra_list),
> ??? .cra_u??? ??? =
> ??? {
> ??? ??? .digest =
> ??? ??? {
> ??? ??? ??? .dia_digestsize??? =??? SHA1_DIGEST_SIZE,
> ??? ??? ??? .dia_init?? ??? = ??? hw_sha1_init,
> ??? ??? ??? .dia_update ??? =??? hw_sha1_update,
> ??? ??? ??? .dia_final? ??? =??? hw_sha1_final
> ??? ??? }
> ??? }
> };
>
> /*
> ?* ========================================================================================
> ?* This is the module's init function
> ?* ========================================================================================
> ?*/
>
> static int __init percello_SHA_init_module (void) {
> ? int ??? ??? ??? ??? ??? ??? ??? ret = 0;
>
> ? return crypto_register_alg(&hw_sha1_alg);;
> }
>
>
>
>
>
>
>
>
>
> On Tue, Sep 22, 2009 at 4:20 PM, Herbert Xu <[email protected]> wrote:
>>
>> On Tue, Sep 22, 2009 at 04:17:20PM +0300, avital sela wrote:
>> >
>> > Any ideas as to what could be the reason for the failure?
>>
>> I don't see the failure message in your testmgr output?
>>
>> Please also attach your source code as otherwise it'd be impossible
>> to guess.
>>
>> Thanks,
>> --
>> Visit Openswan at http://www.openswan.org/
>> Email: Herbert Xu ~{PmV>HI~} <[email protected]>
>> Home Page: http://gondor.apana.org.au/~herbert/
>> PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
>
* avital sela | 2009-09-30 09:28:01 [+0200]:
>On Mon, Sep 28, 2009 at 5:31 PM, avital sela <[email protected]> wrote:
>> Attached below is the source code for the sha driver.? Like I said previously, if I add printk-s or delays in testmgr then the test passes and all subsequent sha1 crypto operations are successful.
This isn't the complete source code. I have no idea what hides behind
SHA_DATA_BUFFER.
Anyway if it works with delays and you are making a HW driver than I
would say that you forgot to flush the data from cache to physical
memory after / before your memcpy().
>>
>> Thanks for your help,
>> Avital
Sebastian