Description of problem: pci_free_consistent would NOT "free" the "pci_alloc_consistent"-ed memory, with more than 4GB memory on Intel platform, work with Areca SATA host adapter. Version-Release number of selected component (if applicable): kernel-2.6.9-55.EL SMP x86_64, with more than 4GB memory Intel platform 1.20.0X.13-61107 CLI V1.72.250_70306 How reproducible: Steps to Reproduce: 1. Make and install arcmsr.ko for Areca SATA host adapter 2. Looping cli of Areca (while true;do ./cli64 sys info; done) 3. 1M/s redusing. Actual results: Expected results: Additional info:
I think we are facing the same issue in a different context. On an Intel based HP machine (this one: http://h71028.www7.hp.com/enterprise/cache/445600-0-0-0-121.html) with 16 GB memory, same kernel, x86_64: the cciss driver causes memory leaks. In the driver specific ioctl called with CCISS_PASSTHRU it uses pci_alloc_consistent / pci_free_consistent and doing such calls more and more memory is reported "in use". How to reproduce: Install the hpasm Software from HP (can be obtained here: http://h20293.www2.hp.com/portal/swdepot/displayProductInfo.do?productNumber=T2812AA) and start it. To see memory being wasted faster, edit /opt/compaq/storage/etc/cmaidad and set PFLAGS="-p 1 -s OK" A process is visible: cmaidad -p 1 -s OK Attaching strace to this process, one can see the ioctl calls on the cciss driver. Calling "free" repeatedly, the memory "in use" increases rapidly.
Yes. The same issue.
Nick, please could you check this issue ?
Hi, FYI As far as i can say, the fix is in kernel-2.6.9-55.9 , at least with this kernel we don't see the problem anymore. Thanks !
It seems to me that the problem is the same as the one in bz #233399 where it was cleared. Unless someone will report that also newer kernels are having this problem, I'll close this bz in a couple of days.
yes, it seems to happen in the specific kernel version. In newer versions, we don't see the problem any more.
Thanks nickcheng and Albert Fluegel. Tomas, this bz may be colsed soon, as you wish.
Thanks to all who participated on this. I'm closing this bz now, as the problem only appeared in kernel-2.6.9-55.EL.