Skip to contentSkip to conversation

linux-arm-2.5.59 problem connecting from win 98

DISCUSSION Marcus Hall 2 participants 2 messages Aug 26, 2003 – Aug 27, 2003

lore
Share & tools
lore
#0 lore raw

I have a PXA-255 (ARM xscale) based system running the 2.5.59-rmk1-pxa1 (+ some trizeps patches) and am having a problem with connecting to the system from a windows box.

Things work just fine with http/ftp/telnet from a linux box, but if I try to connect from a win 98 system, linux panics (don't really blame it..) with the error message 'kernel BUG at net/core/skbuff.c:323', which appears to say that an skbuff is being freed while still on a list.

Looking at the packets being sent, it appears that the windows box is releating the SYN and ACK packets. This should be legal, but may be what is confusing the linux system. The packets I see are:

	windows		linux
		SYN ->
		SYN ->
		<- SYN ACK
		ACK ->
		ACK ->
		data ->
		...

I don't know if the problem occurs on the 2nd SYN, or the 2nd ACK (or, I guess somewhere else...)

I don't believe that there are any changes in the core networking in the arm/xscale patches applied to the base 2.5.59 kernel, just some tweaking of the smc9194 and cs89x0 modules. It seems unlikely that this problem would exist in the "official" kernel for long, but it also seems unlikely to be particular to the arm either... Is it a known bug (hopefully with a patch somewhere)?

Thank You!

Marcus Hall
CorAccess Systems

#1 lore raw
Quoted 5 lines · Marcus Hall

Things work just fine with http/ftp/telnet from a linux box, but if I try to connect from a win 98 system, linux panics (don't really blame it..) with the error message 'kernel BUG at net/core/skbuff.c:323', which appears to say that an skbuff is being freed while still on a list.

Please supply the full backtrace - it may be a driver doing something it shouldn't.

Quoted 6 lines

I don't believe that there are any changes in the core networking in the arm/xscale patches applied to the base 2.5.59 kernel, just some tweaking of the smc9194 and cs89x0 modules. It seems unlikely that this problem would exist in the "official" kernel for long, but it also seems unlikely to be particular to the arm either... Is it a known bug (hopefully with a patch somewhere)?

The best I can advise is to upgrade to something more recent (eg, a 2.6.0-test kernel.)

Signature · 2 lines
Russell King ([email protected])                The developer of ARM Linux
             http://www.arm.linux.org.uk/personal/aboutme.html