linux-kernel - RE: [PATCH] RDMA/siw: Fix the sendmsg byte count in siw_tcp

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID:
 <DS0SPRMB006759C349217E60D43F923B995FA@DS0SPRMB0067.namprd15.prod.outlook.com>
Date: Wed, 23 Jul 2025 16:49:30 +0000
From: Bernard Metzler <BMT@...ich.ibm.com>
To: Pedro Falcato <pfalcato@...e.de>
CC: Jason Gunthorpe <jgg@...pe.ca>, Leon Romanovsky <leon@...nel.org>,
        Vlastimil Babka <vbabka@...e.cz>, Jakub Kicinski <kuba@...nel.org>,
        David
 Howells <dhowells@...hat.com>, Tom Talpey <tom@...pey.com>,
        "linux-rdma@...r.kernel.org" <linux-rdma@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "linux-mm@...ck.org" <linux-mm@...ck.org>,
        "stable@...r.kernel.org"
	<stable@...r.kernel.org>,
        kernel test robot <oliver.sang@...el.com>
Subject: RE: [PATCH] RDMA/siw: Fix the sendmsg byte count in siw_tcp_sendpages



> -----Original Message-----
> From: Pedro Falcato <pfalcato@...e.de>
> Sent: Wednesday, 23 July 2025 17:49
> To: Bernard Metzler <BMT@...ich.ibm.com>
> Cc: Jason Gunthorpe <jgg@...pe.ca>; Leon Romanovsky <leon@...nel.org>;
> Vlastimil Babka <vbabka@...e.cz>; Jakub Kicinski <kuba@...nel.org>; David
> Howells <dhowells@...hat.com>; Tom Talpey <tom@...pey.com>; linux-
> rdma@...r.kernel.org; linux-kernel@...r.kernel.org; linux-mm@...ck.org;
> stable@...r.kernel.org; kernel test robot <oliver.sang@...el.com>
> Subject: [EXTERNAL] Re: [PATCH] RDMA/siw: Fix the sendmsg byte count in
> siw_tcp_sendpages
> 
> On Wed, Jul 23, 2025 at 02:52:12PM +0000, Bernard Metzler wrote:
> >
> >
> > > -----Original Message-----
> > > From: Pedro Falcato <pfalcato@...e.de>
> > > Sent: Wednesday, 23 July 2025 12:41
> > > To: Jason Gunthorpe <jgg@...pe.ca>; Bernard Metzler
> <BMT@...ich.ibm.com>;
> > > Leon Romanovsky <leon@...nel.org>; Vlastimil Babka <vbabka@...e.cz>
> > > Cc: Jakub Kicinski <kuba@...nel.org>; David Howells
> <dhowells@...hat.com>;
> > > Tom Talpey <tom@...pey.com>; linux-rdma@...r.kernel.org; linux-
> > > kernel@...r.kernel.org; linux-mm@...ck.org; Pedro Falcato
> > > <pfalcato@...e.de>; stable@...r.kernel.org; kernel test robot
> > > <oliver.sang@...el.com>
> > [snip]
> > > ---
> > >  drivers/infiniband/sw/siw/siw_qp_tx.c | 4 ++--
> > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/drivers/infiniband/sw/siw/siw_qp_tx.c
> > > b/drivers/infiniband/sw/siw/siw_qp_tx.c
> > > index 3a08f57d2211..9576a2b766c4 100644
> > > --- a/drivers/infiniband/sw/siw/siw_qp_tx.c
> > > +++ b/drivers/infiniband/sw/siw/siw_qp_tx.c
> > > @@ -340,11 +340,11 @@ static int siw_tcp_sendpages(struct socket *s,
> struct
> > > page **page, int offset,
> > >  		if (!sendpage_ok(page[i]))
> > >  			msg.msg_flags &= ~MSG_SPLICE_PAGES;
> > >  		bvec_set_page(&bvec, page[i], bytes, offset);
> > > -		iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, size);
> > > +		iov_iter_bvec(&msg.msg_iter, ITER_SOURCE, &bvec, 1, bytes);
> > >
> > >  try_page_again:
> > >  		lock_sock(sk);
> > > -		rv = tcp_sendmsg_locked(sk, &msg, size);
> > > +		rv = tcp_sendmsg_locked(sk, &msg, bytes);
> > >  		release_sock(sk);
> > >
> >
> > Pedro, many thanks for catching this! I completely
> > missed it during my too sloppy review of that patch.
> > It's a serious bug which must be fixed asap.
> > BUT, looking closer, I do not see the offset being taken
> > into account when retrying a current segment. So,
> > resend attempts seem to send old data which are already
> > out. Shouldn't the try_page_again: label be above
> > bvec_set_page()??
> 
> This was raised off-list by Vlastimil - I think it's harmless to bump (but
> not use)
> the offset here, because by reusing the iov_iter we progressively consume
> the data
> (it keeps its own size and offset tracking internally). So the only thing
> we
> need to track is the size we pass to tcp_sendmsg_locked[1].
> 
Ah okay, I didn't know that. Are we sure? I am currently travelling and have
only limited possibilities to try out things. I just looked up other
use cases and found one in net/tls/tls_main.c#L197. Here the loop looks
very similar, but it works as I was suggesting (taking offset into account
and re-initializing new bvec in case of partial send).

> If desired (and if my logic is correct!) I can send a v2 deleting that bit.
> 

So yes if that's all save, please. We shall not have dead code.

Thanks!
Bernard.
> 
> [1] Assuming tcp_sendmsg_locked guarantees it will never consume something
> out
> of the iovec_iter without reporting it as bytes copied, which from a code
> reading
> it seems like it won't...
> 
> 
> --
> Pedro