Date: Sat, 20 Sep 2003 10:54:21 -0700 (PDT) From: Maksim Yevmenkin <m_evmenkin@yahoo.com> To: Gary Palmer <gjp@in-addr.com> Cc: current@freebsd.org Subject: Re: ubt0, Bluetooth and kernel memory leaking Message-ID: <20030920175421.37161.qmail@web40303.mail.yahoo.com> In-Reply-To: <40053.1064078735@in-addr.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Hello Gary, > It seems when using rfcomm_pppd (at least) as a Bluetooth <-> LAN > gateway there is a fairly large memory leak which eventually crashes > the box running the ppp server. To track down where, I recorded the > output of kern.malloc every minute until the box crashed. At the > start of the debugging run, I saw: > > devbuf 441 2650K 2861K 2504 > 16,32,64,128,256,512,1024,2048,4096,8192,16384,65536 > > A mere 32 minutes later (right before it crashed), it was up a > considerable ammount: > > devbuf 89943184451K 184451K 92399 > 16,32,64,128,256,512,1024,2048,4096,8192,16384,65536 > > I'm pretty sure that kernel virtual memory got exhausted and the > machine paniced and rebooted. hmmm... please correct me if i wrong, but devbuf means memory allocated with M_DEVBUF. if that is true then ng_ubt(4) is likely not a problem as it never allocated memory from M_DEVBUF. > I'm not even sure where to start looking for this. I'm using a MSI USB > dongle: [...] this all looks normal to me. i seem to recall very similar discussion. someone else had exactly the same problem. i think last time it was tracked back to USB code memory leak. > This is on -current from approx. Aug 14 2003. i just quickly checked the CVS. according to your -current date you should have this fix. can you double check? perhaps there is another bug somewhere. i will take close look at ng_ubt(4) today once again. ================================================= CVS log for src/sys/dev/usb/usb_mem.c Revision 1.4 / (download) - annotate - [select for diffs], Tue Jul 29 05:07:37 2003 UTC (7 weeks, 4 days ago) by jmg Branch: MAIN CVS Tags: HEAD Changes since 1.3: +2 -0 lines Diff to previous 1.3 (colored) fix another bus_dma leak due to not having a size param for our bus_dma allocation function. With this patch, it prevents continous growth of the devbuf memory pool. Tested with ssh <host> dd of=/dev/null < /dev/zero and vmstat -m | grep devbuf ====================================================== thanks, max __________________________________ Do you Yahoo!? Yahoo! SiteBuilder - Free, easy-to-use web site design software http://sitebuilder.yahoo.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20030920175421.37161.qmail>