IMFS / memfile bug?
strauman at SLAC.Stanford.EDU
Tue Sep 17 01:50:49 UTC 2002
Joel Sherrill wrote:
>Till Straumann wrote:
>>Sorry, if this is already known - it seems I can't access the bug
>www.oarcorp.com had a security breach. It is undergoing an
>audit and upgrade. :(
Sorry to hear that - what a nuisance :-(
>>There seems to be bug in the IMFS/memfile implementation:
>>Memory for files who are closed after removal (e.g. created
>>by tmpfile()) seems to be never released...
>>Is anybody working on IMFS?
>Jennifer and I wrote the IMFS. We fix bugs as they are reported.
>Please send a test case and we will look into it.
IMO it's easier to look at the source than producing a test case -
you'll see it immediately. I came across it when tmpfile() ate
up all my memory.
tmpfile() opens a file, unlink()s it and returns the open file to the
caller. The file is supposed to be physically removed upon close().
Now, memfile_rmnod() [correctly] refuses to release the memory
occupied by an open file. However, there should be a check in
memfile_close() for this case [i.e. last close of a file that has
already been unlinked from a directory]. Fixing this shouldn't be
hard but I don't have the time right now to study the details
which you and Jennifer are more familiar with anyways...
More information about the users