Thomas Müller wrote: > I have been digging the mailinglist archives a bit > further. I know this is way old stuff, but the symtoms > described there are just those I experience since > upgrading to 2.4.21 kernel: [snip] > Any help very much appreciated. Put your data on device backed (partition backed) loop. File backed loops may deadlock. This may happen when kernel is running out of free RAM, and when a process is running file system code and is holding some exclusive resource, and calls VM to allocate more RAM saying "allocate me some RAM, but don't re-enter any file system code while doing that". At that point VM will shovel out some dirty RAM pages to satisfy allocation request. Dirty pages that are sent to normal devices, or device backed loops, don't have problems with that. Guess what happens when such shoveled out RAM pages are on file backed loop: before RAM can be freed, they need to be written to file system, and they may need to acquire that same exclusive resource that other thread happens to be holding. IOW, one thread waits for RAM to be freed, other thread waits for exlusive resource other thread is holding. Deadlock. File backed loops work OK when they are used in read-only mode. Regards, Jari Ruusu <jari.ruusu@xxxxxxxxxx> - Linux-crypto: cryptography in and on the Linux system Archive: http://mail.nl.linux.org/linux-crypto/
