othermark wrote: >Gregory Nou wrote: > > >>I'm currently experiencing a weird problem : some of my folders are >>completely empty (ls -a doesn't even mention . or ..) >>But, I cannot remove them, and a ls is very very slow (but just in those >>folders, and it is quite rare, but now I cannot update firefox, nor >>openoffice, because make clean will fail). >> >> > >Yeah, I've seen this. For me, this is the direct result of the ServerWorks >chipset on the box not properly doing DMA, and in fact causes corruption on >the bus when DMA is enabled for the IDE controller. Usually you see this >when doing large extracts or deletes (like the openoffice of >mozilla/firefox ports). You need to fsck, hopefully before you even read >this (you'll see lots of softupdate problems being fixed). > >You'll also notice that your daily automatic reports sent to root will >mention problems with traversing files or directories. > >For reference this bad boy: > >atapci0: <ServerWorks ROSB4 UDMA33 controller> port >0x1f0-0x1f7,0x3f6,0x170-0x177,0x376,0xfc90-0xfc9f at device 15.1 on pci0 > >atapci0_at_pci0:15:1: class=0x01018a card=0x00000000 chip=0x02111166 >rev=0x00 >hdr=0x00 > vendor = 'ServerWorks (Was: Reliance Computer Corp)' > device = 'OSB4 PCI EIDE Controller' > class = mass storage > subclass = ATA > >The only solution is to run with dma off (see atacontrol, or shut it off in >BIOS). > > > Thanks for the answer. However, I did not have the chance to test it, because I had the very bad idea to reboot. I tried to fsck and had lots of : ad2: FAILURE - READ_DMA status=51 <READY, DSC, ERROR> error=40 <UNCORRECTABLE> LBA=xxxxxxx So the system was marked as dirty, and I could not boot. I just could backup datas and reinstall all. making a newfs on the damaged slice made the errors disppear. I suppose it's normal. But I cannot understand why fsck was unable to correct it, because it asked me if I wanted to correct, but told me then that it could not write on this sector. That's weird . -- GregoryReceived on Sat Sep 10 2005 - 07:39:03 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:43 UTC