2007-05-08 12:05:58

by Rusty Russell

[permalink] [raw]
Subject: [PATCH] lguest: two net bugfixes

1) Bridging via host is broken: we need to set "promisc" bit in MAC
address published by the host so the guest sends us everything.
Thanks James Morris for the report (I don't use bridging).

2) Lguest network device uses 0 to mean "noone at this slot". It used to
use 0xFF, and one spot was missed. Minor: it just means we'll have a
useless extra transmit for broadcast packets.

Signed-off-by: Rusty Russell <[email protected]>
---
Documentation/lguest/lguest.c | 2 ++
drivers/net/lguest_net.c | 2 +-
2 files changed, 3 insertions(+), 1 deletion(-)

diff -r 999a9058a151 Documentation/lguest/lguest.c
--- a/Documentation/lguest/lguest.c Tue May 08 19:49:33 2007 +1000
+++ b/Documentation/lguest/lguest.c Tue May 08 21:04:51 2007 +1000
@@ -860,6 +860,10 @@ static void setup_tun_net(const char *ar

/* We are peer 0, ie. first slot. */
configure_device(ipfd, ifr.ifr_name, ip, dev->mem);
+
+ /* Set "promisc" bit: we want every single packet. */
+ *((u8 *)dev->mem) |= 0x1;
+
close(ipfd);

verbose("device %p: tun net %u.%u.%u.%u\n",
diff -r 999a9058a151 drivers/net/lguest_net.c
--- a/drivers/net/lguest_net.c Tue May 08 19:49:33 2007 +1000
+++ b/drivers/net/lguest_net.c Tue May 08 21:03:47 2007 +1000
@@ -249,7 +249,7 @@ static int lguestnet_close(struct net_de
struct lguestnet_info *info = dev->priv;

/* Clear all trace: others might deliver packets, we'll ignore it. */
- memset(&info->peer[info->me], 0xFF, sizeof(info->peer[info->me]));
+ memset(&info->peer[info->me], 0, sizeof(info->peer[info->me]));

/* Deregister sg lists. */
hcall(LHCALL_BIND_DMA, peer_key(info, info->me), __pa(info->dma), 0);





2007-05-08 13:17:20

by James Morris

[permalink] [raw]
Subject: Re: [PATCH] lguest: two net bugfixes

On Tue, 8 May 2007, Rusty Russell wrote:

> 1) Bridging via host is broken: we need to set "promisc" bit in MAC
> address published by the host so the guest sends us everything.
> Thanks James Morris for the report (I don't use bridging).
>
> 2) Lguest network device uses 0 to mean "noone at this slot". It used to
> use 0xFF, and one spot was missed. Minor: it just means we'll have a
> useless extra transmit for broadcast packets.
>
> Signed-off-by: Rusty Russell <[email protected]>

Acked-by: James Morris <[email protected]>



> ---
> Documentation/lguest/lguest.c | 2 ++
> drivers/net/lguest_net.c | 2 +-
> 2 files changed, 3 insertions(+), 1 deletion(-)
>
> diff -r 999a9058a151 Documentation/lguest/lguest.c
> --- a/Documentation/lguest/lguest.c Tue May 08 19:49:33 2007 +1000
> +++ b/Documentation/lguest/lguest.c Tue May 08 21:04:51 2007 +1000
> @@ -860,6 +860,10 @@ static void setup_tun_net(const char *ar
>
> /* We are peer 0, ie. first slot. */
> configure_device(ipfd, ifr.ifr_name, ip, dev->mem);
> +
> + /* Set "promisc" bit: we want every single packet. */
> + *((u8 *)dev->mem) |= 0x1;
> +
> close(ipfd);
>
> verbose("device %p: tun net %u.%u.%u.%u\n",
> diff -r 999a9058a151 drivers/net/lguest_net.c
> --- a/drivers/net/lguest_net.c Tue May 08 19:49:33 2007 +1000
> +++ b/drivers/net/lguest_net.c Tue May 08 21:03:47 2007 +1000
> @@ -249,7 +249,7 @@ static int lguestnet_close(struct net_de
> struct lguestnet_info *info = dev->priv;
>
> /* Clear all trace: others might deliver packets, we'll ignore it. */
> - memset(&info->peer[info->me], 0xFF, sizeof(info->peer[info->me]));
> + memset(&info->peer[info->me], 0, sizeof(info->peer[info->me]));
>
> /* Deregister sg lists. */
> hcall(LHCALL_BIND_DMA, peer_key(info, info->me), __pa(info->dma), 0);
>
>
>

--
James Morris
<[email protected]>