[<prev] [next>] [day] [month] [year] [list]
Message-ID: <1cd3d67c0901310137m3b15af09rf1db91ad28c8fdb6@mail.gmail.com>
Date: Sat, 31 Jan 2009 10:37:17 +0100
From: Sebastiano Di Paola <sebastiano.dipaola@...il.com>
To: netdev@...r.kernel.org
Cc: linux-kernel@...r.kernel.org, trivial@...nel.org,
"Fred N. van Kempen" <waltje@...lt.NL.Mugnet.ORG>,
Alan Cox <gw4pts@...pts.ampr.org>, paolo.abeni@...il.com,
sebastiano.dipaola@...il.com
Subject: Subject: [PATCH 2.6.28-git] net: packet socket packet_lookup_frame
fix
From: Sebastiano Di Paola and Paolo Abeni
packet_lookup_frames() fails to get user frame if current frame header
status contains extra flags.
This is due to the wrong assumption on the operators precedence during
frame status tests.
Fixed by forcing the right operators precedence order with explicit brackets.
Signed-off-by: Paolo Abeni <paolo.abeni@...il.com>
Signed-off-by: Sebastiano Di Paola <sebastiano.dipaola@...il.com>
---
The issue arises when packet_lookup_frame() is called with status ==
TP_STATUS_USER, that is when the packet socket is memory mapped and a
select() or a poll() syscall is performed on the socket.
If h.h1->tp_status is set to TP_STATUS_USER | TP_STATUS_LOSING (or any
other value with multiple bits set), the tested expression evaluate to
TP_STATUS_USER (!= 0) and a null pointer is returned, even if the
current frame is ready to be read by the user. In such scenario the
poll() and the select() system call stop working.
N.B. This problem is present from version >= 2.6.27
--- linux-2.6.28/net/packet/af_packet.c.orig 2009-01-30 16:29:56.000000000 +0100
+++ linux-2.6.28/net/packet/af_packet.c 2009-01-30 16:27:58.000000000 +0100
@@ -220,13 +220,13 @@ static void *packet_lookup_frame(struct
h.raw = po->pg_vec[pg_vec_pos] + (frame_offset * po->frame_size);
switch (po->tp_version) {
case TPACKET_V1:
- if (status != h.h1->tp_status ? TP_STATUS_USER :
- TP_STATUS_KERNEL)
+ if (status != (h.h1->tp_status ? TP_STATUS_USER :
+ TP_STATUS_KERNEL))
return NULL;
break;
case TPACKET_V2:
- if (status != h.h2->tp_status ? TP_STATUS_USER :
- TP_STATUS_KERNEL)
+ if (status != (h.h2->tp_status ? TP_STATUS_USER :
+ TP_STATUS_KERNEL))
return NULL;
break;
}
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Powered by blists - more mailing lists