From mboxrd@z Thu Jan 1 00:00:00 1970 From: "David S. Miller" Subject: Re: Mystery packet killing tg3 Date: Tue, 3 May 2005 15:03:33 -0700 Message-ID: <20050503150333.1ac1c159.davem@davemloft.net> References: <20050502162405.65dfb4a9@localhost.localdomain> <20050502200251.38271b61.davem@davemloft.net> <20050503140528.1bd2ac74@dxpl.pdx.osdl.net> <20050503141314.441c9d75.davem@davemloft.net> <1115152907.15156.26.camel@rh4> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: shemminger@osdl.org, jgarzik@pobox.com, netdev@oss.sgi.com Return-path: To: "Michael Chan" In-Reply-To: <1115152907.15156.26.camel@rh4> Sender: netdev-bounce@oss.sgi.com Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org On Tue, 03 May 2005 13:41:47 -0700 "Michael Chan" wrote: > MSI is buggy on 5703 and tg3 will not enable MSI on that chip. MSI only > works on the latest PCI Express chips. Ok, and he has a pre-MSI copy of the tg3 driver anyways. Michael, there were no master/target abort bits set in the PCI status register from his dump. If one of the DMA units locks up on the tg3, will it still be able to update the PCI_STATUS register appropriately when it encounters a DMA transaction error (ie. master or target abort) or would we also need to look at the PCI host bridge PCI config space registers as well? I strongly suspect one of two things in these AMD system cases of tg3 wedging: 1) some DMA problem 2) register write reordering Lack of master/target abort indication in tg3's PCI status register makes me feel that #2 is more likely to be the problem cause. Just for fun, Stephen, can you make tg3_get_invariants() always set the TG3_FLAG_MBOX_WRITE_REORDER bit in tp->tg3_flags and see if that makes the problem go away? Thanks.