Hello, On antradienis 29 Rugsėjis 2009 15:31:54 Alan Jenkins wrote: > Alan Jenkins wrote: > > Alan Jenkins wrote: > >> Modestas Vainius wrote: > >>> Hello, > >>> > >>> On antradienis 29 Rugsėjis 2009 12:12:41 Alan Jenkins wrote: > >>> > >>> > >>> Probably not, but I think the message could be more generic like > >>> "Incomplete data in /sys/module/*/ or failed to read /proc/modules". > > How about this: > > WARNING: /sys/module/ not present or too old, and /proc/modules does not > exist. ACK. > >>> Ok, your intentions are much clearer now. But beware, my comment still > >>> applies. You did not patch code in module_in_sysfs() so it will still > >>> return 0 if /sys/module/<modulename> is present but > >>> /sys/module/<modulename>/initstate is not present (<= 2.6.19). This is > >>> because read_attribute() returns 0 if file is NOT present which is the > >>> case here. Therefore, module_in_kernel() will not fallback to > >>> module_in_procfs(), but return 0 and the bug will not be fixed. > >> > >> Ah. How about if I add this to the patch: > > > > Nevermind, that's not quite right. > > Ok, I've written a separate patch for this. It passes the existing > tests, and it works for me on a mock-up of the old sysfs. Looks fine. Commit the whole patchset to github and I will test it on 2.6.18. -- Modestas Vainius <modestas@xxxxxxxxxx>
Attachment:
signature.asc
Description: This is a digitally signed message part.
