Yes, this does correct the problem with unloading the module. I'll take a look this evening and see if I can find anything else out about the problem with afsd sucking cpu. -- Nathan > -----Original Message----- > From: Neulinger, Nathan R. [mailto:[EMAIL PROTECTED]] > Sent: Thursday, November 30, 2000 1:39 PM > To: 'Michael Pronath' > Cc: Derek Atkins; [EMAIL PROTECTED] > Subject: [OpenAFS-devel] RE: problem with openafs on i386_linux22 SMP > with 2.2.17, 2.2.18p re17 > > > Ok. not sure if this fixes it or not, hosed my home machine > trying... (can't > get to it at the moment) > > I do know though that it does _not_ help the afsd going > cpu-hog problem I > experienced on that box. > > I'll let ya know as soon as I find out whether it corrects > the memory oops. > > -- Nathan > > > -----Original Message----- > > From: Michael Pronath [mailto:[EMAIL PROTECTED]] > > Sent: Thursday, November 30, 2000 6:50 AM > > To: Nathan Neulinger > > Cc: Derek Atkins; [EMAIL PROTECTED] > > Subject: problem with openafs on i386_linux22 SMP with 2.2.17, > > 2.2.18pre17 > > > > > > Your post to openafs-devel on Fri, 03 Nov 2000: > > > > This is apparently a bug in the memory allocation code of > > the Linux module > > libafs-xxx.o . > > The hash structures for memory allocation are setup the > > first time memory > > is allocated. If we insert the module with insmod, no > allocation has > > happened so far and the structures are not initialized. If > > you remove the > > module with rmmod right after that, the function > > osi_linux_free_afs_memory (src/afs/LINUX/osi_alloc.c:322) > is called. > > It mistakenly assumes that the hash lh_mem_tab is setup > > already and uses > > it. This leads to the segfault. > > > > The following patch should fix this problem. > > Apply it with > > > > cd src/afs/LINUX > > patch < alloc.patch > > > > > > Patch file: > > cut here >---------------------------------------------- > > --- src/afs/LINUX/osi_alloc.c Mon Oct 30 19:07:41 2000 > > +++ src/afs/LINUX/osi_alloc.c_new Thu Nov 30 13:38:10 2000 > > @@ -323,6 +323,7 @@ > > { > > down(&afs_linux_alloc_sem); > > > > + if (allocator_init) { > > /* iterate through all elements in the hash table and > free both > > * the chunk and the atom associated with it. > > */ > > @@ -336,7 +337,7 @@ > > > > /* change the state so that the allocator is now > > uninitialized. */ > > allocator_init = 0; > > - > > + } > > up(&afs_linux_alloc_sem); > > } > > cut here >---------------------------------------------- > > > > > > > > Michael Pronath > > > > > > > > -- > > > > Dipl.-Ing. Michael Pronath > > Lehrstuhl fuer Entwurfsautomatisierung Tel: +49 > > 89 289 23659 > > Technische Universitaet Muenchen Fax: +49 > > 89 289 63666 > > Arcisstr.21, D-80333 Muenchen TU-Innenhof, Gebaeude > > 9, Raum 4915 > > Mailto:[EMAIL PROTECTED] > > > _______________________________________________ > OpenAFS-devel mailing list > [EMAIL PROTECTED] > https://lists.openafs.org/mailman/listinfo.cgi/openafs-devel > _______________________________________________ OpenAFS-devel mailing list [EMAIL PROTECTED] https://lists.openafs.org/mailman/listinfo.cgi/openafs-devel
