On Mon, Sep 26, 2011 at 06:12:45PM -0500, Stan Hoeppner wrote: > I'm curious about the footprint of the XFS binary code, specifically > for 32 bit systems, but knowing the 64 bit case would be educational > as well for comparison. I don't use kernel modules or I'd already > have a rough answer. My last build was 2.6.38.6 but I'm sure the > size hasn't changed much in more recent kernels. My XFS build > options were: > > $ grep -i xfs .config > CONFIG_XFS_FS=y > # CONFIG_XFS_QUOTA is not set > # CONFIG_XFS_POSIX_ACL is not set > # CONFIG_XFS_RT is not set > > My resulting kernel image is: > $ la /boot/vmlinuz-2.6.38.6 > -rw-r--r-- 1 root root 1.6M May 17 23:55 vmlinuz-2.6.38.6 > > How much of that 1.6MB is XFS? That's the compressed image that is 1.6MB, so the uncompressed kernel will be quite a bit bigger than that. As it is, the XFs binary size can be found from your source tree via: $ size fs/xfs/xfs.ko text data bss dec hex filename 468288 59632 1400 529320 813a8 fs/xfs/xfs.ko $ So for my current kernel build on x86-64 with these options: CONFIG_XFS_FS=y CONFIG_XFS_QUOTA=y CONFIG_XFS_POSIX_ACL=y CONFIG_XFS_RT=y # CONFIG_XFS_DEBUG is not set the XFS binary footprint is around 520kB. Yours will be smaller given you don't have RT, ACL and quota configured, but it will still probably be the largest single code block in your kernel... > Also, what is the size of the hot > path code that we want to stay in L1 during heavy IO? Simple answer: it will stay in the L2 cache if we are lucky and userspace doesn't trash the L2 cache between syscalls. More complex answer: run perf w/ sampling on L1/L2 cache misses and have a look for yourself. Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx _______________________________________________ xfs mailing list xfs@xxxxxxxxxxx http://oss.sgi.com/mailman/listinfo/xfs
