2018-01-25 10:03:18

by Jia-Ju Bai

[permalink] [raw]
Subject: [PATCH] crypto: crypto_user: Replace GFP_ATOMIC with GFP_KERNEL in crypto_report

After checking all possible call chains to crypto_report here,
my tool finds that crypto_report is never called in atomic context.
And crypto_report calls crypto_alg_match which calls down_read,
thus it proves again that crypto_report can call functions which may sleep.
Thus GFP_ATOMIC is not necessary, and it can be replaced with GFP_KERNEL.

This is found by a static analysis tool named DCNS written by myself.

Signed-off-by: Jia-Ju Bai <[email protected]>
---
crypto/crypto_user.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/crypto/crypto_user.c b/crypto/crypto_user.c
index 0dbe2be7..23baef1 100644
--- a/crypto/crypto_user.c
+++ b/crypto/crypto_user.c
@@ -271,7 +271,7 @@ static int crypto_report(struct sk_buff *in_skb, struct nlmsghdr *in_nlh,
return -ENOENT;

err = -ENOMEM;
- skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
+ skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
if (!skb)
goto drop_alg;

--
1.7.9.5



2018-02-15 16:25:40

by Herbert Xu

[permalink] [raw]
Subject: Re: [PATCH] crypto: crypto_user: Replace GFP_ATOMIC with GFP_KERNEL in crypto_report

On Thu, Jan 25, 2018 at 06:06:02PM +0800, Jia-Ju Bai wrote:
> After checking all possible call chains to crypto_report here,
> my tool finds that crypto_report is never called in atomic context.
> And crypto_report calls crypto_alg_match which calls down_read,
> thus it proves again that crypto_report can call functions which may sleep.
> Thus GFP_ATOMIC is not necessary, and it can be replaced with GFP_KERNEL.
>
> This is found by a static analysis tool named DCNS written by myself.
>
> Signed-off-by: Jia-Ju Bai <[email protected]>

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