Messages in this thread | | | Date | Sat, 18 Dec 1999 16:05:39 -0500 | From | TenThumbs <> | Subject | Re: More on the IDE multiwrite problem |
| |
Mark Lord wrote: > > TenThumbs wrote: > > Mark Lord wrote: > > > Okay, the error reported must be from the second ide_multwrite() call, > > > after the "END IRQ" in Alan's scenario above. This could all make sense. > > > > > > If so, reinstating the disable/enable calls *for this fragment only* > > > should cure the problem 100%. > .. > > I'm still seeing the problem in 2.2.14pre15 although it may have moved. > .. > > Do you have a patch to try? > > Yup. Attached. > > Could you tell me more about your system? > Eg.. what IDE chipset is having the problem, > does it share interrupts with anything, > and is your kernel SMP, or UP?
Worse, unfortunately. Now I'm seeing "lost interrupt" messages, the system becomes sluggish, the drive light sometimes stays on with no disk activity, and eventually X locks up. Here's an example: I was running a ppp connection, X, Netscape, and squid. Squid has its cache on hdd, Netscape (and most of the system) use hdc, and the swap partition is on hdb. I get these messages and then X locked up.
Dec 18 15:07:39 perfect kernel: hdb: lost interrupt Dec 18 15:08:19 perfect last message repeated 4 times Dec 18 15:09:07 perfect last message repeated 3 times Dec 18 15:09:43 perfect kernel: hdb: lost interrupt Dec 18 15:11:17 perfect last message repeated 2 times Dec 18 15:11:58 perfect last message repeated 3 times Dec 18 15:13:26 perfect last message repeated 2 times Dec 18 15:13:57 perfect last message repeated 3 times Dec 18 15:16:01 perfect kernel: hdb: lost interrupt
I'm testing on and old 486 which only understands PIO. Not sure what the IDE chipset is but it isn't new. Obviously it's a UP kernel and it's not sharing interrupts. There are two controllers on the motherboard and 4 drives.
It seems interesting that hdb lost interrupts. Is the fact that I'm using two controllers make a difference?
-- "Now you will see that Evil will always triumph because Good is dumb." -- Dark Helmet
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
| |