| Message ID | 1509436586-17500-1-git-send-email-gilad@benyossef.com |
|---|---|
| State | New |
| Headers | show |
| Series | crypto: testmgr: don't allocate IV on stack | expand |
On Tue, Oct 31, 2017 at 07:56:26AM +0000, Gilad Ben-Yossef wrote: > The IV was allocated on the stack in testmgr skcipher tests. > Since HW based tfm providers need to DMA the IV to the HW, > this leads to problems and is detected by the DMA-API debug > code. > > Fix it by allocating the IV using kmalloc instead. > > Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com> The driver that is mapping the IV directly should be fixed instead. Only input that is given in the form of SG lists can be mapped. Everything else should be copied if they need to go over DMA. Cheers, -- Email: Herbert Xu <herbert@gondor.apana.org.au> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
On Tue, Oct 31, 2017 at 9:59 AM, Herbert Xu <herbert@gondor.apana.org.au> wrote: > On Tue, Oct 31, 2017 at 07:56:26AM +0000, Gilad Ben-Yossef wrote: >> The IV was allocated on the stack in testmgr skcipher tests. >> Since HW based tfm providers need to DMA the IV to the HW, >> this leads to problems and is detected by the DMA-API debug >> code. >> >> Fix it by allocating the IV using kmalloc instead. >> >> Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com> > > The driver that is mapping the IV directly should be fixed instead. > Only input that is given in the form of SG lists can be mapped. > Everything else should be copied if they need to go over DMA. Got it. Will do. Thanks, Gilad > > Cheers, > -- > Email: Herbert Xu <herbert@gondor.apana.org.au> > Home Page: http://gondor.apana.org.au/~herbert/ > PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt -- Gilad Ben-Yossef Chief Coffee Drinker "If you take a class in large-scale robotics, can you end up in a situation where the homework eats your dog?" -- Jean-Baptiste Queru
On 10/31/2017 10:00 AM, Herbert Xu wrote: > On Tue, Oct 31, 2017 at 07:56:26AM +0000, Gilad Ben-Yossef wrote: >> The IV was allocated on the stack in testmgr skcipher tests. >> Since HW based tfm providers need to DMA the IV to the HW, >> this leads to problems and is detected by the DMA-API debug >> code. >> >> Fix it by allocating the IV using kmalloc instead. >> >> Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com> Gilad, you're not the only one who bumped into this issue: https://www.mail-archive.com/linux-crypto@vger.kernel.org/msg23074.html not to mention previous patches that have been accepted: 96692a7305c4 crypto: tcrypt - do not allocate iv on stack for aead speed tests 9bac019dad80 crypto: testmgr - Fix DMA-API warning and so on. > > The driver that is mapping the IV directly should be fixed instead. > Only input that is given in the form of SG lists can be mapped. > Everything else should be copied if they need to go over DMA. > Herbert, wouldn't it make more sense to follow your previous suggestion: "Perhaps we should change the API so that it gets passed in as an SG list." https://www.mail-archive.com/linux-crypto@vger.kernel.org/msg23082.html Thanks, Horia
diff --git a/crypto/testmgr.c b/crypto/testmgr.c index 7125ba3..88d0c57 100644 --- a/crypto/testmgr.c +++ b/crypto/testmgr.c @@ -1085,12 +1085,16 @@ static int __test_skcipher(struct crypto_skcipher *tfm, int enc, const char *e, *d; struct tcrypt_result result; void *data; - char iv[MAX_IVLEN]; + char *iv; char *xbuf[XBUFSIZE]; char *xoutbuf[XBUFSIZE]; int ret = -ENOMEM; unsigned int ivsize = crypto_skcipher_ivsize(tfm); + iv = kmalloc(MAX_IVLEN, GFP_KERNEL); + if (!iv) + goto out_nobuf; + if (testmgr_alloc_buf(xbuf)) goto out_nobuf; @@ -1331,6 +1335,7 @@ static int __test_skcipher(struct crypto_skcipher *tfm, int enc, testmgr_free_buf(xoutbuf); out_nooutbuf: testmgr_free_buf(xbuf); + kfree(iv); out_nobuf: return ret; }
The IV was allocated on the stack in testmgr skcipher tests. Since HW based tfm providers need to DMA the IV to the HW, this leads to problems and is detected by the DMA-API debug code. Fix it by allocating the IV using kmalloc instead. Signed-off-by: Gilad Ben-Yossef <gilad@benyossef.com> --- crypto/testmgr.c | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) -- 2.7.4