USB stack built for BONE crashes in UHCI finish thread
|Reported by:||jonas.kirilla||Owned by:||mmlr|
Feedback as per requested in the last paragraph: http://haiku-os.org/blog/mmlr/2007-04-20/using_the_haiku_usb_stack
" Problems in uhci finish thread : segment violation address: 0x80233070, op: 1 pc: 0x00175563, tmid: 2 thid: 10 kernel debugger: Welcome to Kernel Debugging Land...
eax 00000000 ebp 0122c944 cs 0008 | area 004b3768 (kernel_intel_text) ebx 00000040 esp 0122c8f8 ss 0010 | addr 00100000 size 00086000 ecx 00000040 edi 300f6600 ds 0010 | edx 00000000 esi 80233070 es 0010 | Thread: uhci finish thread eip 00175563 flag 00010246 fs 0000 | Team: kernel_team
trap 0000000e err 00000000 gs 0000 | Stack Trace follows:
00000000 00175563 memcpy_pIII+0047 0122c944 0021a012 CreateFilledTransfer4UHCIP8TransferPP7uhci_tdPP7uhci_qh+00aa 0122c984 002196ef FinishTransfers4UHCI+0387 0122c9e4 0021935f FinishThread4UHCIPv+001f "
This is on R5/Bone, at the very start of copying a 233MB file from /boot to a USB flash drive. I can't try the same thing in Haiku as the umass (usb_scsi) driver hasn't been adapted to Haiku, IIRC.
The CPU is a P4/Celeron, the motherboard is an Asus P4-XP (IIRC) with the intel 845 chipset. The only other USB device connected, that I know of, is a USB mouse by Microsoft.
I can read from the USB flash media without problems, but writing to it KDLs. I get the feeling that creating folders and small files is less likely to crash than is creating large files.
The media is in FAT format.
The full syslog will be uploaded shortly.
A thought: memcpy_pIII, P4-class CPU? Another one: Is too much memory allocated by the USB driver for the transfer?