[PATCH] rxrpc: sendmsg: Fix variable overwrite in rxrpc_queue_packet

[PATCH] rxrpc: sendmsg: Fix variable overwrite in rxrpc_queue_packet

[Gustavo A. R. Silva]

Value assigned to variable resend_at is overwritten before it can be used.

Fix this by removing the value overwrite as it seems that this is a
leftover code.

Addresses-Coverity-ID: 1462262
Fixes: beb8e5e4f38c ("rxrpc: Express protocol timeouts in terms of RTT")
Signed-off-by: Gustavo A. R. Silva <garsilva@embeddedor.com>
---
 net/rxrpc/sendmsg.c | 1 -
 1 file changed, 1 deletion(-)

diff --git a/net/rxrpc/sendmsg.c b/net/rxrpc/sendmsg.c
index a1c53ac..42060dd 100644
--- a/net/rxrpc/sendmsg.c
+++ b/net/rxrpc/sendmsg.c
@@ -233,7 +233,6 @@ static void rxrpc_queue_packet(struct rxrpc_sock *rx, struct rxrpc_call *call,
 		if (resend_at < 1)
 			resend_at = 1;
 
-		resend_at = now + rxrpc_resend_timeout;
 		WRITE_ONCE(call->resend_at, resend_at);
 		rxrpc_reduce_call_timer(call, resend_at, now,
 					rxrpc_timer_set_for_send);
-- 
2.7.4

Re: [PATCH] rxrpc: sendmsg: Fix variable overwrite in rxrpc_queue_packet

[David Howells]

Gustavo A. R. Silva <garsilva@embeddedor.com> wrote:

> Value assigned to variable resend_at is overwritten before it can be used.
> 
> Fix this by removing the value overwrite as it seems that this is a
> leftover code.

NAK.  Your fix will actually cause the code to break.

The resend_at value used for the timer must be based on the current time
(ie. jiffies in this case), so we can't simply remove that line as the
previously calculated resend_at value is a duration, not a time.

What you need to do is to instead modify the line you wanted to remove to add
'now' to the previously-computed value.

David

Re: [PATCH] rxrpc: sendmsg: Fix variable overwrite in rxrpc_queue_packet

[Gustavo A. R. Silva]

Hi David,

Quoting David Howells <dhowells@redhat.com>:

> Gustavo A. R. Silva <garsilva@embeddedor.com> wrote:
>
>> Value assigned to variable resend_at is overwritten before it can be used.
>>
>> Fix this by removing the value overwrite as it seems that this is a
>> leftover code.
>
> NAK.  Your fix will actually cause the code to break.
>
> The resend_at value used for the timer must be based on the current time
> (ie. jiffies in this case), so we can't simply remove that line as the
> previously calculated resend_at value is a duration, not a time.
>
> What you need to do is to instead modify the line you wanted to remove to add
> 'now' to the previously-computed value.
>

You mean something like the following?

--- a/net/rxrpc/sendmsg.c
+++ b/net/rxrpc/sendmsg.c
@@ -233,7 +233,7 @@ static void rxrpc_queue_packet(struct rxrpc_sock  
*rx, struct rxrpc_call *call,
                 if (resend_at < 1)
                         resend_at = 1;

-               resend_at = now + rxrpc_resend_timeout;
+               resend_at += now;
                 WRITE_ONCE(call->resend_at, resend_at);
                 rxrpc_reduce_call_timer(call, resend_at, now,
                                         rxrpc_timer_set_for_send);


Thanks
--
Gustavo A. R. Silva

Re: [PATCH] rxrpc: sendmsg: Fix variable overwrite in rxrpc_queue_packet

[David Howells]

Gustavo A. R. Silva <garsilva@embeddedor.com> wrote:

> -               resend_at = now + rxrpc_resend_timeout;
> +               resend_at += now;

Yep.  :-)

David

Re: [PATCH] rxrpc: sendmsg: Fix variable overwrite in rxrpc_queue_packet

[Gustavo A. R. Silva]

Quoting David Howells <dhowells@redhat.com>:

> Gustavo A. R. Silva <garsilva@embeddedor.com> wrote:
>
>> -               resend_at = now + rxrpc_resend_timeout;
>> +               resend_at += now;
>
> Yep.  :-)
>

Great!

What about this one: https://lkml.org/lkml/2017/11/27/810

applies the same?

Thanks
--
Gustavo A. R. Silva

Re: [PATCH] rxrpc: sendmsg: Fix variable overwrite in rxrpc_queue_packet

[Gustavo A. R. Silva]

Quoting David Howells <dhowells@redhat.com>:

> Gustavo A. R. Silva <garsilva@embeddedor.com> wrote:
>
>> -               resend_at = now + rxrpc_resend_timeout;
>> +               resend_at += now;
>
> Yep.  :-)
>

Great!

What about this one: https://lkml.org/lkml/2017/11/27/810

applies the same?

Thanks