Skip to contentSkip to conversation

crypto: sun8i-ce: fix multiple memory leaks in sun8i_ce_hash_run

PATCH In review Dongliang Mu 2 participants 8 messages Aug 3, 2021 – Sep 23, 2021

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 crypto: sun8i-ce: fix multiple memory leaks in sun8i_ce_hash_run

In sun8i_ce_hash_run, all the dma_mmap_sg/single will cause memory leak due to no corresponding unmap operation if errors happen.

Fix this by freeing the objects allocated bydma_mmap_sg/single when errors occur.

Fixes 56f6d5aee88d ("crypto: sun8i-ce - support hash algorithms")

---
v1->v2: fix some wording and keep error handling code consistent

 .../crypto/allwinner/sun8i-ce/sun8i-ce-hash.c | 27 ++++++++++---------
 1 file changed, 15 insertions(+), 12 deletions(-)
drivers/crypto/allwinner/sun8i-ce/sun8i-ce-hash.c +15 -12
index 88194718a806..05bb781da701 100644
@@ -288,14 +288,14 @@ int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq)
 	buf = kzalloc(bs * 2, GFP_KERNEL | GFP_DMA);
 	if (!buf) {
 		err = -ENOMEM;
-		goto theend;
+		goto out;
 	}
 	bf = (__le32 *)buf;
 
 	result = kzalloc(digestsize, GFP_KERNEL | GFP_DMA);
 	if (!result) {
 		err = -ENOMEM;
-		goto theend;
+		goto err_buf;
 	}
 
 	flow = rctx->flow;
@@ -321,7 +321,7 @@ int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq)
 	if (nr_sgs <= 0 || nr_sgs > MAX_SG) {
 		dev_err(ce->dev, "Invalid sg number %d\n", nr_sgs);
 		err = -EINVAL;
-		goto theend;
+		goto err_result;
 	}
 
 	len = areq->nbytes;
@@ -334,7 +334,7 @@ int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq)
 	if (len > 0) {
 		dev_err(ce->dev, "remaining len %d\n", len);
 		err = -EINVAL;
-		goto theend;
+		goto err_unmap_sg;
 	}
 	addr_res = dma_map_single(ce->dev, result, digestsize, DMA_FROM_DEVICE);
 	cet->t_dst[0].addr = cpu_to_le32(addr_res);
@@ -342,7 +342,7 @@ int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq)
 	if (dma_mapping_error(ce->dev, addr_res)) {
 		dev_err(ce->dev, "DMA map dest\n");
 		err = -EINVAL;
-		goto theend;
+		goto err_unmap_sg;
 	}
 
 	byte_count = areq->nbytes;
@@ -392,7 +392,7 @@ int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq)
 	if (dma_mapping_error(ce->dev, addr_pad)) {
 		dev_err(ce->dev, "DMA error on padding SG\n");
 		err = -EINVAL;
-		goto theend;
+		goto err_addr_res;
 	}
 
 	if (ce->variant->hash_t_dlen_in_bits)
@@ -405,15 +405,18 @@ int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq)
 	err = sun8i_ce_run_task(ce, flow, crypto_tfm_alg_name(areq->base.tfm));
 
 	dma_unmap_single(ce->dev, addr_pad, j * 4, DMA_TO_DEVICE);
+err_addr_res:
+	dma_unmap_single(ce->dev, addr_res, digestsize, DMA_FROM_DEVICE);
+err_unmap_sg:
 	dma_unmap_sg(ce->dev, areq->src, sg_nents(areq->src),
 		     DMA_TO_DEVICE);
-	dma_unmap_single(ce->dev, addr_res, digestsize, DMA_FROM_DEVICE);
-
-
 	memcpy(areq->result, result, algt->alg.hash.halg.digestsize);
-theend:
-	kfree(buf);
+err_result:
 	kfree(result);
-	crypto_finalize_hash_request(engine, breq, err);
+err_buf:
+	kfree(buf);
+out:
+	if (!err)
+		crypto_finalize_hash_request(engine, breq, err);
 	return 0;
 }
Signature · 1 line
2.25.1
maintainer
#1 lore raw

drivers/crypto/allwinner/sun8i-ce/sun8i-ce-hash.c int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq) · line 421

Quoted 12 lines · Dongliang Mu
drivers/crypto/allwinner/sun8i-ce/sun8i-ce-hash.c +6 -3
-theend:
-	kfree(buf);
+err_result:
 	kfree(result);
-	crypto_finalize_hash_request(engine, breq, err);
+err_buf:
+	kfree(buf);
+out:
+	if (!err)
+		crypto_finalize_hash_request(engine, breq, err);
 	return 0;

This does not look right. You're returning zero in case of an error

Cheers,

Signature · 3 lines
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
#2 lore raw
Quoted 16 lines · Herbert Xu

drivers/crypto/allwinner/sun8i-ce/sun8i-ce-hash.c int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq) · line 421

Quoted 12 lines · Dongliang Mu
drivers/crypto/allwinner/sun8i-ce/sun8i-ce-hash.c +6 -3
-theend:
-     kfree(buf);
+err_result:
      kfree(result);
-     crypto_finalize_hash_request(engine, breq, err);
+err_buf:
+     kfree(buf);
+out:
+     if (!err)
+             crypto_finalize_hash_request(engine, breq, err);
      return 0;

This does not look right. You're returning zero in case of an error

Hi Herbert,

Corentin Labbe said,

For the error code, I am not sure it is needed, error code is already given to user via crypto_finalize_hash_request().
The "return 0" is for crypto/crypto_engine API, returning an error will not change anything since we dont have retry_support.

So I propose you to focus on dma_map_xxx() fix patch.

See details in [1]. P.S., my previous patch returns err variable.

[1] https://lkml.org/lkml/2021/7/26/164

Quoted 6 lines

Cheers,
--
Email: Herbert Xu <[email protected]> Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

maintainer
#3 lore raw
Quoted 18 lines · Dongliang Mu
Quoted 16 lines · Herbert Xu

drivers/crypto/allwinner/sun8i-ce/sun8i-ce-hash.c int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq) · line 421

Quoted 12 lines · Dongliang Mu
drivers/crypto/allwinner/sun8i-ce/sun8i-ce-hash.c +6 -3
-theend:
-     kfree(buf);
+err_result:
      kfree(result);
-     crypto_finalize_hash_request(engine, breq, err);
+err_buf:
+     kfree(buf);
+out:
+     if (!err)
+             crypto_finalize_hash_request(engine, breq, err);
      return 0;

This does not look right. You're returning zero in case of an error

Quoted 6 lines

Hi Herbert,

Corentin Labbe said,

For the error code, I am not sure it is needed, error code is already given to user via crypto_finalize_hash_request().

Yes but your patch changes this. You're now skipping the finalize call and thus throwing away err if it's not zero.

If it's supposed to do this you need to explain it in your patch submission.

Cheers,

Signature · 3 lines
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
#4 lore raw
Quoted 22 lines · Herbert Xu
Quoted 20 lines · Dongliang Mu
Quoted 16 lines · Herbert Xu

drivers/crypto/allwinner/sun8i-ce/sun8i-ce-hash.c int sun8i_ce_hash_run(struct crypto_engine *engine, void *breq) · line 421

Quoted 12 lines · Dongliang Mu
drivers/crypto/allwinner/sun8i-ce/sun8i-ce-hash.c +6 -3
-theend:
-     kfree(buf);
+err_result:
      kfree(result);
-     crypto_finalize_hash_request(engine, breq, err);
+err_buf:
+     kfree(buf);
+out:
+     if (!err)
+             crypto_finalize_hash_request(engine, breq, err);
      return 0;

This does not look right. You're returning zero in case of an error

Hi Herbert,

Quoted 8 lines
Quoted 4 lines

Corentin Labbe said,

For the error code, I am not sure it is needed, error code is already given to user via crypto_finalize_hash_request().

Yes but your patch changes this. You're now skipping the finalize call and thus throwing away err if it's not zero.

I see. Thanks for your comment.

I agree with you and would like to modify "return 0" to "return err" in the next version. Even if the parent function does not care about the return value, it is still necessary to keep the right return value.

Corentin, how do you think?

Quoted 8 lines

If it's supposed to do this you need to explain it in your patch submission.

Cheers,
--
Email: Herbert Xu <[email protected]> Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

maintainer
#5 lore raw
Quoted 5 lines · Dongliang Mu

I agree with you and would like to modify "return 0" to "return err" in the next version. Even if the parent function does not care about the return value, it is still necessary to keep the right return value.

Well if the parent is ignoring the error then returning the error there obviously isn't going to help.

Cheers,

Signature · 3 lines
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
#6 lore raw
Quoted 10 lines · Herbert Xu
Quoted 5 lines · Dongliang Mu

I agree with you and would like to modify "return 0" to "return err" in the next version. Even if the parent function does not care about the return value, it is still necessary to keep the right return value.

Well if the parent is ignoring the error then returning the error there obviously isn't going to help.

If so, do I still need to change my current patch?

Dongliang Mu

Quoted 6 lines

Cheers,
--
Email: Herbert Xu <[email protected]> Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

#7 lore raw
Quoted 10 lines · Dongliang Mu
Quoted 8 lines · Herbert Xu
Quoted 5 lines · Dongliang Mu

I agree with you and would like to modify "return 0" to "return err" in the next version. Even if the parent function does not care about the return value, it is still necessary to keep the right return value.

Quoted 4 lines
Quoted 2 lines

Well if the parent is ignoring the error then returning the error there obviously isn't going to help.

If so, do I still need to change my current patch?

Any update on this thread?

Quoted 9 lines

Dongliang Mu

Quoted 6 lines

Cheers,
--
Email: Herbert Xu <[email protected]> Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt