Re: Re: PING: [PATCH] crypto: public_key: fix overflow during implicit conversion

On Thu, 2021-08-19 at 10:03 +0800, zhenwei pi wrote:

On 8/18/21 8:33 PM, Jarkko Sakkinen wrote:

quoted

On Wed, 2021-08-18 at 16:33 +0800, zhenwei pi wrote:

quoted

PING

Please, do not top-post.

You are lacking Herbert Xu:

$ scripts/get_maintainer.pl crypto/asymmetric_keys/public_key.c
David Howells [off-list ref] (maintainer:ASYMMETRIC KEYS)
Herbert Xu [off-list ref] (maintainer:CRYPTO API)
"David S. Miller" [off-list ref] (maintainer:CRYPTO API)
keyrings@vger.kernel.org (open list:ASYMMETRIC KEYS)
linux-crypto@vger.kernel.org (open list:CRYPTO API)
linux-kernel@vger.kernel.org (open list)

quoted

On 8/10/21 2:39 PM, zhenwei pi wrote:

quoted

Hit kernel warning like this, it can be reproduced by verifying
256
bytes datafile by keyctl command.

   WARNING: CPU: 5 PID: 344556 at crypto/rsa-pkcs1pad.c:540
pkcs1pad_verify+0x160/0x190
   ...
   Call Trace:
    public_key_verify_signature+0x282/0x380
    ? software_key_query+0x12d/0x180
    ? keyctl_pkey_params_get+0xd6/0x130
    asymmetric_key_verify_signature+0x66/0x80
    keyctl_pkey_verify+0xa5/0x100
    do_syscall_64+0x35/0xb0
    entry_SYSCALL_64_after_hwframe+0x44/0xae

'.digest_size(u8) = params->in_len(u32)' leads overflow of an
u8

Where is this statement?

In function "static int asymmetric_key_verify_signature(struct 
kernel_pkey_params *params, const void *in, const void *in2)"

quoted

quoted

quoted

value,
so use u32 instead of u8 of digest. And reorder struct
public_key_signature, it could save 8 bytes on a 64 bit
machine.

                                                      ~~~~~
                                                      64-bit
                                                      
What do you mean by "could"? Does it, or does it
not?
                                        				
	

After reordering struct public_key_signature, sizeof(struct 
public_key_signature) gets smaller than the original version.

OK, then just state is as "it saves" instead of "it could save".

Not a requirement but have you been able to trigger this for a
kernel that does not have this fix?

/Jarkko