crypto: AF_ALG: handle 0 lengths in af_alg_make_sg
PATCH 1/1 crypto: AF_ALG: handle 0 lengths in af_alg_make_sg
Hi Herbert,
If you concur with the patch, I think it should go to 4.11 as well as to stable.
Ciao
Stephan
---8<---
The function af_alg_make_sg converts user-provided IOVECs into an SGL. Thus it operates directly on the user-space provided number of IOVECs. When user space provides 0 for the number of IOVECs iov_iter_get_pages returns a bogus number of bytes. This in turn will cause a crash when the SGL is processed.
The fix initializes an SGL with one entry for handling chaining operation but does not contain data.
In addition, the patch changes variable type of len from int to size_t to be consistent with the data type of the invoker and the data type where len is used.
--- crypto/af_alg.c | 10 +++++++++- include/crypto/if_alg.h | 2 +- 2 files changed, 10 insertions(+), 2 deletions(-)
crypto/af_alg.c +9 -1
@@ -399,12 +399,20 @@ static const struct net_proto_family alg_family = { .owner = THIS_MODULE, }; -int af_alg_make_sg (struct af_alg_sgl *sgl,struct iov_iter *iter,int len) +int af_alg_make_sg (struct af_alg_sgl *sgl,struct iov_iter *iter,size_t len) {size_t off;ssize_t n;int npages, i; +if (!len) { +/* init one for linking */ +sg_init_table (sgl->sg,1 ); +sg_mark_end (sgl->sg); + sgl->npages =0 ; +return 0 ; + } + n =iov_iter_get_pages (iter, sgl->pages, len,ALG_MAX_PAGES , &off);if (n <0 )return n;
include/crypto/if_alg.h +1 -1
@@ -76,7 +76,7 @@ int af_alg_release(struct socket *sock);void af_alg_release_parent (struct sock *sk);int af_alg_accept (struct sock *sk,struct socket *newsock); -int af_alg_make_sg (struct af_alg_sgl *sgl,struct iov_iter *iter,int len); +int af_alg_make_sg (struct af_alg_sgl *sgl,struct iov_iter *iter,size_t len);void af_alg_free_sg (struct af_alg_sgl *sgl);void af_alg_link_sg (struct af_alg_sgl *sgl_prev,struct af_alg_sgl *sgl_new);
Signature · 1 line
2.9.3
Am Samstag, 1. April 2017, 17:04:28 CEST schrieb Stephan Müller:
Hi Herbert,
Quoted 4 lines
Hi Herbert,
If you concur with the patch, I think it should go to 4.11 as well as to stable.
After checking this issue again, I see that it is not triggerable in the current code as the different af_alg users make sure that this function is not called with 0.
I only triggered this issue during experimenting with the algif_skcipher and algif_aead revamp as requested by you. During those experiments, I invoked af_alg_make_sg with a len = 0.
Thus, this patch is not applicable for stable and 4.11.
Yet, I would suggest to consider this patch as a safeguard for any potential programming errors.
Ciao
Stephan
Herbert Xu
maintainerQuoted 13 lines · Stephan Müller
Am Samstag, 1. April 2017, 17:04:28 CEST schrieb Stephan Müller:
Hi Herbert,
Quoted 4 lines
Hi Herbert,
If you concur with the patch, I think it should go to 4.11 as well as to stable.
After checking this issue again, I see that it is not triggerable in the current code as the different
af_algusers make sure that this function is not called with 0.
Quoted 8 lines
I only triggered this issue during experimenting with the
algif_skcipherandalgif_aeadrevamp as requested by you. During those experiments, I invokedaf_alg_make_sgwith a len = 0.Thus, this patch is not applicable for stable and 4.11.
Yet, I would suggest to consider this patch as a safeguard for any potential programming errors.
So this is only possible with patches that have not been applied yet? In that case please fold this into whichever patch series that needs it.
Thanks,
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
Am Mittwoch, 5. April 2017, 14:50:12 CEST schrieb Herbert Xu:
Hi Herbert,
Quoted 2 lines
So this is only possible with patches that have not been applied yet?
Correct.
Quoted 2 lines
In that case please fold this into whichever patch series that needs it.
Ok, let us defer it.
Ciao
Stephan