I believe I have detected a significant problem with the "atl1e" driver for the Attansic Technology Corp. Atheros AR8121/AR8113/AR8114 PCI-E Ethernet Controller (rev b0) when running Fedora 11 preview with the latest updates.
This controller is integrated on the ASUS M4A78 PRO motherboard.
Although my problem occurred when I was running scp, I believe that the problem could also occur with other forms of data transfer and only show up as corrupted data (files). Thus, I thought this email appropriate to warn other users. My current "solution" is to install another NIC.
This problem has been reported: https://bugzilla.redhat.com/show_bug.cgi?id=503288 http://bugzilla.kernel.org/show_bug.cgi?id=13404
This is no "show stopper" but may be of concern to other users. I am posting a separate copies of this email to the test and user mailing lists.
Gene
On 05/30/2009 04:36 PM, Gene Czarcinski wrote:
I believe I have detected a significant problem with the "atl1e" driver for the Attansic Technology Corp. Atheros AR8121/AR8113/AR8114 PCI-E Ethernet Controller (rev b0) when running Fedora 11 preview with the latest updates.
I added to the bug that I'm seeing this on two machines with the same NIC - one Asus-board desktop, and an eeePC, both on RC2. I thought I was seeing it sooner on the desktop, but that appears to have just been random bad luck, both can go for several minutes (gigabit) without errors.
Although my problem occurred when I was running scp, I believe that the problem could also occur with other forms of data transfer and only show up as corrupted data (files).
I'm glad I use rsync. One thing it does is send a checksum when it's done transferring a file. Inevitably, I try something else and wind up using rsync in the end!
Thus, I thought this email appropriate to warn other users. My current "solution" is to install another NIC.
That's one heck of a recommended workaround for a regression. I can't seem to find the PCIe connector on the netbook. :)
-Bill
This is actually pretty bad. Repeatable, patterned silent data corruption, using multiple common NIC's on multiple architectures.
Perhaps somebody who understands the kernel can shed some insight onto the regularity of the corruption noted on the bug. I can see the pattern, but I have no idea where it could come from or what else to test.
-Bill