Bug 3765
Summary: | (net b44) Network link down, when writting to sata disk AND network in use | ||
---|---|---|---|
Product: | Drivers | Reporter: | Surakshan Mendis (bugtracker) |
Component: | Network | Assignee: | Jeff Garzik (jgarzik) |
Status: | REJECTED UNREPRODUCIBLE | ||
Severity: | high | CC: | bunk, protasnb |
Priority: | P2 | ||
Hardware: | i386 | ||
OS: | Linux | ||
Kernel Version: | 2.6.9, 2.6.13, 2.6.17-rc4 | Subsystem: | |
Regression: | --- | Bisected commit-id: |
Description
Surakshan Mendis
2004-11-17 21:13:32 UTC
I belive I'm eliminated "power" as a reason, i've disconnected a 7200rpm non vital drive (IDE), disconnected cd and dvd rom, disconnected non vital fans. I've also moved the sata cable to a different sata port on the expansion card. Result hasn't changed, problem persists added realtek LAN card and 8139too driver. I can confirm the bug no longer occuers with this network card. So its got something to do with the Broadcom 4400 driver in the kernel, or something whack with the hardware on the mobo. I do not have a similar machine to test this. Is this problem still present in kernel 2.6.13-rc6? Problem is still there on the stabe 2.6.13 from kernel.org. Today I tried by doing an FTP transfer where the netword card was the Broadcom 4400 onboard using the b44 driver. Test 1: saved large file to IDE disk. No problem Test 2: saved large file to SATA disk (connected to sata controller), network drops... Sep 19 08:51:58 synapse b44: eth0: Link is down. Sep 19 08:52:01 synapse b44: eth0: Link is up at 100 Mbps, full duplex. Sep 19 08:52:01 synapse b44: eth0: Flow control is on for TX and on for RX. Sep 19 08:52:03 synapse b44: eth0: Link is down. Sep 19 08:52:06 synapse b44: eth0: Link is up at 100 Mbps, full duplex. Sep 19 08:52:06 synapse b44: eth0: Flow control is on for TX and on for RX. ... ..... TEST 3: Transfers data from IDE disk to sata disk, network drops same as output from the system logger. I'm using another PCI network card that uses the natsemi driver -- the problem doesn't happen on it. Also as mentioned before I've really stripped down the system by removing all but the SATA controller and the problem continues. I'm seeing the same occur here. P4PE motherboard, Promise SATA300 TX4 card installed. Googling shows others having this problem without mention of the TX4 card. Disabling ACPI has no effect. I have a feeling that the PCI bus being flooded is the problem. I get the problem doing /any/ network traffic, not just traffic resulting in writes to disk. In the case I'm looking at, the system is resyncing a raid10 array. Doing a "make oldconfig" (or any other network traffic, it seems) in an nfs-mounted partition causes the link bouncing behaviour. Fiddling with the raid rebuild rate DOES affect things - reduces the occurence of the problem, but at a raid10 rebuild rate of 20M/second the problem still occurs. Wild-assed guess at it being a PCI bus problem. Removed the PCI sata card. Can still reproduce problems by connecting two SATA drives to the integrated SATA controller on this motherboard (ASUS P4PE). Integrated SATA controller is on the same bus as the b44 and the motherboard's PCI slots. Both SATA controllers (on-board and PCI card) are Promise chipsets, but I'm thinking that b44 is simply unhappy when the PCI bus is busy. I don't have any other convenient way of flooding that bus ATM. (kernel is 2.6.17rc4) Any update on this? Is it still a problem with recent kernel? Thanks. Since the subsystem has changes so much and no new updates - closing the bug (hopefully everything works now). |