On Mon, 1 Mar 2004, Martin Diehl wrote: > After I fixed the "you shall not DMA to the stack while using > usb_control_msg" issue and stopped the tx-path from using transfer_buffers > directly from skb (and thus probably not cache-line aligned) the > disconnect-badness warning didn't appear anymore. Neither does any 5a5a5a > bad entry. So it might be the slab corruption was caused by these driver > issues.
Seems I have to correct myself: the disconnect-badness warning can still be triggered. It seems I was to careful doing ifconfig down before physical disconnect... However, I haven't seen the 5a5a5a slab poisoned thing so far... > BULK-IN and a SETUP TD, the latter one with CC=5, len=0. However, the new > observation is, while it happens with ohci_hcd on NEC-companion with or > without external hub - it seems it never happens when the device is > connected to a SiS USB-1.1 OHCI-HC! So it might even be a silicon issue > with the nec-ohci. Or the pci-interface of this guy is somewhat faster > exposing some race with the hcd. Was too fast drawing this conclusion: meanwhile I've managed to trigger it with the SiS silicon as well. Probably it's just less likely. In fact this was a slightly different one, as the bad entry was taken as an ISO-TD! Note there is no ISO-endpoint in the whole usb-tree. But again, it's an almost-all-zero td and mistakenly considered ISO because it happens to come with TD_ISO/TD_DONE bits set. Anyway, it seems it's not broken HC. ohci_hcd 0000:00:01.2: bad entry 7be5001 ohci_hcd 0000:00:01.2: td_hash[1] ohci_hcd 0000:00:01.2: entry td c7be5040; urb c7cf3bbc index 0; hw next td 07be5080 ohci_hcd 0000:00:01.2: info f0140000 CC=f (CARRY) DI=0 IN R ohci_hcd 0000:00:01.2: cbp 07af9000 be 07af9fff (len 4096) ohci_hcd 0000:00:01.2: bad? td c7be5001 (DONE); urb 00000000 index 0; hw next td 00000000 ohci_hcd 0000:00:01.2: info 00870000 CC=0 FC=0 DI=4 SF=0000 ohci_hcd 0000:00:01.2: bp0 00000000 be 00000000 ohci_hcd 0000:00:01.2: psw [0] = 0, CC=0 SIZE=0 > Hm, just speculation of course, but in td_submit_urb I see quite a number > of writel - could it be there might a readl missing somewhere do deal with > posted-writes on pci? Just tried to add a readl(&ohci->regs->control) at the end of td_submit_urb - seems nothing improved so far... Martin ------------------------------------------------------- SF.Net is sponsored by: Speed Start Your Linux Apps Now. Build and deploy apps & Web services for Linux with a free DVD software kit from IBM. Click Now! http://ads.osdn.com/?ad_id=1356&alloc_id=3438&op=click _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
