I sent mail to this list a couple of days ago with a fix for a bug in
COW file systems where the kernel walks off the end of the bitmap. I've been
running this fix for several days now in several different instances all with
COW file systems that have nice even sizes (like 512M, 1G, 2G, 2.5G, 3G). I
didn't see any discussion/responses to the patch I posted and wanted to ensure
that some solution (doesn't have to be mine, but...) is in future versions of
the uml kernel.
I was also reviewing the list of bugs to be fixed before 1.0 and see at leaset one
that I can probably knock off quickly. Does anyone have any more details on:
# Fix the ubd rounding bug spotted by Steve Schmidtke and Roman Zippel.
If the number of sectors isn't a multiple of 8, then the bitmap will be a a byte
shorter than it should be.
I ran into some interesting corruption on hostfs reads on a Mandrake 9.0 UML, but
it also seemed to coincide with my host having /tmp filled up so I initially dismissed
it. Now I wonder if it could be related to the hostfs related UML building problem:
# I built UML inside itself twice on hostfs, with the output going to a log. The first
failed, and when the second ran, the log filled with garbage.
In my case, the source files were getting random non-ascii bytes inserted and
generated lots of errors. The weird thing was that if I ran make again, it would
compile a few more files each time (only took 20 or so invocations of make
to complete the build :-). I'm still trying to determine if the filling of /tmp
was related to the suspected memory leak. I'm running a host kernel 2.4.18-24
(with host-skas3.patch applied), and my UML kernels are all 2.4.19-51um.
Lynn Kerby <mailto:lfk@...>