Bug 6936
Summary: | BUG: warning at net/core/dev.c:1171/skb_checksum_help() | ||
---|---|---|---|
Product: | Networking | Reporter: | Pawel Staszewski (pstaszewski) |
Component: | IPV4 | Assignee: | Herbert Xu (herbert) |
Status: | RESOLVED CODE_FIX | ||
Severity: | normal | ||
Priority: | P2 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.18-rc3 | Subsystem: | |
Regression: | --- | Bisected commit-id: |
Description
Pawel Staszewski
2006-07-31 19:04:54 UTC
On Tue, Aug 15, 2006 at 11:29:59AM -0700, bugme-daemon@bugzilla.kernel.org wrote: > http://bugzilla.kernel.org/show_bug.cgi?id=6936 It's actually a bug in the bridging code :) [BRIDGE]: Disable SG/GSO if TX checksum is off When the bridge recomputes features, it does not maintain the constraint that SG/GSO must be off if TX checksum is off. This patch adds that constraint. On a completely unrelated note, I've also added TSO6 and TSO_ECN feature bits if GSO is enabled on the underlying device through the new NETIF_F_GSO_SOFTWARE macro. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Cheers, -- Visit Openswan at http://www.openswan.org/ Email: Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt -- diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h index 75f02d8..5d95da5 100644 --- a/include/linux/netdevice.h +++ b/include/linux/netdevice.h @@ -320,6 +320,9 @@ #define NETIF_F_GSO_ROBUST (SKB_GSO_DODG #define NETIF_F_TSO_ECN (SKB_GSO_TCP_ECN << NETIF_F_GSO_SHIFT) #define NETIF_F_TSO6 (SKB_GSO_TCPV6 << NETIF_F_GSO_SHIFT) + /* List of features with software fallbacks. */ +#define NETIF_F_GSO_SOFTWARE (NETIF_F_TSO | NETIF_F_TSO_ECN | NETIF_F_TSO6) + #define NETIF_F_GEN_CSUM (NETIF_F_NO_CSUM | NETIF_F_HW_CSUM) #define NETIF_F_ALL_CSUM (NETIF_F_IP_CSUM | NETIF_F_GEN_CSUM) diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c index f55ef68..b1211d5 100644 --- a/net/bridge/br_if.c +++ b/net/bridge/br_if.c @@ -386,12 +386,17 @@ void br_features_recompute(struct net_br checksum = 0; if (feature & NETIF_F_GSO) - feature |= NETIF_F_TSO; + feature |= NETIF_F_GSO_SOFTWARE; feature |= NETIF_F_GSO; features &= feature; } + if (!(checksum & NETIF_F_ALL_CSUM)) + features &= ~NETIF_F_SG; + if (!(features & NETIF_F_SG)) + features &= ~NETIF_F_GSO_MASK; + br->dev->features = features | checksum | NETIF_F_LLTX | NETIF_F_GSO_ROBUST; } On Wed, 16 Aug 2006 11:29:00 +1000 Herbert Xu <herbert@gondor.apana.org.au> wrote: > On Tue, Aug 15, 2006 at 11:29:59AM -0700, bugme-daemon@bugzilla.kernel.org wrote: > > http://bugzilla.kernel.org/show_bug.cgi?id=6936 > > It's actually a bug in the bridging code :) > > [BRIDGE]: Disable SG/GSO if TX checksum is off > > When the bridge recomputes features, it does not maintain the > constraint that SG/GSO must be off if TX checksum is off. > This patch adds that constraint. > > On a completely unrelated note, I've also added TSO6 and TSO_ECN > feature bits if GSO is enabled on the underlying device through > the new NETIF_F_GSO_SOFTWARE macro. > > Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> > > Cheers, agree. I assume this came in with the new GSO for 2.6.18 or do we need to fix 2.6.17 as well. On Wed, Aug 16, 2006 at 11:06:31AM -0700, Stephen Hemminger wrote:
>
> agree. I assume this came in with the new GSO for 2.6.18 or do we need
> to fix 2.6.17 as well.
I think 2.6.17 should be fine because it would only happen there if
the underlying device itself had an inconsistent configuration (TX
checksum off and SG/TSO on).
Cheers,
Reply-To: davem@davemloft.net From: Herbert Xu <herbert@gondor.apana.org.au> Date: Wed, 16 Aug 2006 11:29:00 +1000 > On Tue, Aug 15, 2006 at 11:29:59AM -0700, bugme-daemon@bugzilla.kernel.org wrote: > > http://bugzilla.kernel.org/show_bug.cgi?id=6936 > > It's actually a bug in the bridging code :) > > [BRIDGE]: Disable SG/GSO if TX checksum is off Applied, thanks a lot Herbert. |