Re: Fwd: [PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-24 Thread Herbert Xu
On Thu, Sep 21, 2006 at 04:59:37PM -0700, Jesse Brandeburg wrote: > Herbert can you help describe why we need this patch on this thread on > netdev? We reproduced the race without it and it appears to go away > with it. Of course you can resolve it by adding a lock to the txclean path. However, t

Re: [PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-20 Thread Auke Kok
Jeff Garzik wrote: Actually, I rescind the ACK. The code should be inside a spinlock, and therefore not need this additional check. If this check were truly needed, then SMP code all over the kernel would be broken. I will drop the patch for now. Once Jesse is back next week he gets to exp

Re: [PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-19 Thread Stephen Hemminger
On Tue, 19 Sep 2006 16:26:12 -0700 Auke Kok <[EMAIL PROTECTED]> wrote: > Stephen Hemminger wrote: > > On Tue, 19 Sep 2006 16:45:06 -0400 > > Jeff Garzik <[EMAIL PROTECTED]> wrote: > > > >> looks OK except for the tasklet, which may starve if the lock is being > >> held upon entry > >> > >> > >

Re: [PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-19 Thread Auke Kok
Stephen Hemminger wrote: On Tue, 19 Sep 2006 16:45:06 -0400 Jeff Garzik <[EMAIL PROTECTED]> wrote: looks OK except for the tasklet, which may starve if the lock is being held upon entry Why would the tasklet starve anymore than NAPI? Worst case, the transmitters fill the ring completely an

Re: [PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-19 Thread Stephen Hemminger
On Tue, 19 Sep 2006 16:45:06 -0400 Jeff Garzik <[EMAIL PROTECTED]> wrote: > looks OK except for the tasklet, which may starve if the lock is being > held upon entry > > Why would the tasklet starve anymore than NAPI? Worst case, the transmitters fill the ring completely and have to wait for th

Re: [PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-19 Thread Jeff Garzik
looks OK except for the tasklet, which may starve if the lock is being held upon entry - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: [PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-19 Thread Stephen Hemminger
My preference would be for drivers not to use LLTX (except loopback) and just use the dev->xmit_lock via netif_tx_lock if possible. Something like this for e1000. Subject: [PATCH] e1000 no lltx Get rid of lockless transmit for e1000. Use netif_tx_lock instead of having to do locking in device.

Re: [PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-19 Thread Jeff Garzik
ACK - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: [PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-19 Thread Jeff Garzik
Actually, I rescind the ACK. The code should be inside a spinlock, and therefore not need this additional check. If this check were truly needed, then SMP code all over the kernel would be broken. - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message

[PATCH 12/23] e1000: Maybe stop TX if not enough free descriptors

2006-09-19 Thread Kok, Auke
Cc: Herbert Xu <[EMAIL PROTECTED]> Signed-off-by: Jesse Brandeburg <[EMAIL PROTECTED]> Signed-off-by: Auke Kok <[EMAIL PROTECTED]> --- drivers/net/e1000/e1000_main.c | 52 1 files changed, 42 insertions(+), 10 deletions(-) diff --git a/drivers/net/e100