Red Hat Bugzilla – Bug 106947
Unable to transfer files with 3c575bt NIC
Last modified: 2015-01-04 17:03:30 EST
From Bugzilla Helper: User-Agent: Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.4.1) Gecko/20031009 Description of problem: Using a 3Com 575 BT PCMCIA NIC (cardbus), I am able to maintain normal external connections to the internet. However, any attempt to transfer large amounts of data of the LAN stalls, with the following error message: Oct 13 17:13:04 main kernel: nfs: server mail2.tqmcube.com not responding, still trying Oct 13 17:13:04 main last message repeated 4 times Oct 13 17:13:09 main kernel: NETDEV WATCHDOG: eth0: transmit timed out Oct 13 17:13:09 main kernel: eth0: transmit timed out, tx_status 00 status e601. Oct 13 17:13:09 main kernel: diagnostics: net 0cc2 media a800 dma 0000003a. Oct 13 17:13:09 main kernel: eth0: Interrupt posted but not delivered -- IRQ blocked by another device? Oct 13 17:13:09 main kernel: Flags; bus-master 1, dirty 39658(10) current 39658(10) Oct 13 17:13:09 main kernel: Transmit list 00000000 vs. c036d480. Version-Release number of selected component (if applicable): 3.1.31-13 How reproducible: Always Steps to Reproduce: 1. Attempt to move larger files or directories from the server to a client using NFS, FTP or SFTP. 2. 3. Actual Results: Transfer stalls and system becomes unstable for a couple of minutes Expected Results: Normal data transfer. Additional info: Booting the client to WinXP provides perfect results (same server, NIC and router). Moreover, in Win the card advertises as full duplex. In linux it advertises as half duplex and cannot be forced to full duplex which is probably the heart of the problem.
Problem continues with no improvement in Fedora Core 1, kernel 2.4.22-1.2115.nptl (in fact it seems a tad worse). Custom compiling with the driver compiled into the kernel provides no improvement either. Setting the partner NIC to HD offers some improvement but the issue is still present.
Same problem, here, on ThinkPad T21. Ejecting card and re-inserting "fixes" the issue. Seems to be a problem with resource allocation: Interrupt posted but not delivered -- IRQ blocked by another device? Did not have this problem on same machine running RedHat 9 or Debian Woody.