On Thu, 21 Jan 2010, Karel Zak wrote: > > Done. Now the code calls read() for necessary data only. Results: Beautiful. And what a difference it makes: [root@EeePC util-linux-ng]# time /sbin/blkid -p /dev/sdc /dev/sdc: PTTYPE="dos" real 0m0.269s user 0m0.002s sys 0m0.004s [root@EeePC util-linux-ng]# time /sbin/blkid -p /dev/sdc1 /dev/sdc1: SEC_TYPE="msdos" VERSION="FAT12" TYPE="vfat" USAGE="filesystem" real 0m0.271s user 0m0.002s sys 0m0.003s so now blkid is no longer the problem. There are still other things that are slower than I'd like, and it takes a total of about 8 seconds for the device to show up, but now it's one second of storage settling time, and 1.3s for the actual 'mount' itself to happen etc etc. And part of it is that while I installed the self-compiled util-linux-ng, I didn't install the fixes to DeviceKit-disk, so there's two seconds of the slow devkit-disks-part-id and hald-probe-storage. There's at least a second and a half to be gained there from the trivial fadvise() thing. But at this point, it's now just plain "that's an effing slow device", and no longer "the tools are just broken" So thanks. Please do try to push this out as a release so that distros pick it up. And I'll do the same, and make 'delay_use' default to 0 at the beginning of the next merge window, to see if it really is needed. And we'll have a much snappier desktop. Linus -- To unsubscribe from this list: send the line "unsubscribe util-linux-ng" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
