> Also note that my song is in one fat32 partition and > mozilla tar is on another fat32. The destination of > untar is on "/" which is logging. Is the pcfs driver > the culprit here? I have doubt about that because > bzip2 never figured in the above mentioned 5 seconds > in iosnoop. Moreover, the symptoms of skips/freezes > also show up when firefox loads and tries to open > around 12 tabs; this operation is all "/" bound.
No, I don't think it's fat32 / the pcfs filesystem. I've been able to reproduce the problem on a S10 GA x86 system that uses ufs filesystems only (with and without logging). Is this the fsflush_do_pages() loop in fsflush.c that is responsible for this huge amout of I/O request ?? http://cvs.opensolaris.org/source/xref/usr/src/uts/common/fs/fsflush.c#fsflush_do_pages It scans for modified pages and syncs them to HDD. Apparently, the more memory is installed in the machine, the more (modified) pages could be synced to HDD with one function to fsflush_do_pages() That would be the "nscan" variable, computed in fsflush_do_pages(): http://cvs.opensolaris.org/source/xref/usr/src/uts/common/fs/fsflush.c#149 149 nscan = (last_total_pages * (tune.t_fsflushr))/v.v_autoup; nscan can be reduced by removing memory from the system; by lowering tune.t_fsflushr (but apparently that is already reduced to 1 on S10 GA); or by increasing the v.v_autoup tunable. This message posted from opensolaris.org _______________________________________________ opensolaris-discuss mailing list [email protected]
