Arch Linux User Repository

I've disowned the package so justin can take it over. Sorry for being absent.

heh, I see you noticed I updated the package. That -8 release uses a udev hook instead of the standard one. From my tests it has been much more reliable. Even if you put the regular hook inbetween block and filesystem, it would try to run the script before all block devices had activated. The udev way appears to work much better. If bobpaul doesn't respond in time I'll take over this package and update the archwiki with the udev stuff.

Broken because of changes in the Makefile. :( Here is the link with the changes. http://evilpiepirate.org/git/bcache-tools.git/tree/Makefile Btw Justin8 link is https://www.dray.be/arch/pkgbuilds/bcache-tools-git-0.r83.89f11b1-8.src.tar.gz :)

Since there has been no updates in 6 months. Please see the pkgbuild here in the mean time for updated patches against the latest revision and using recent pkgbuild standards and versioning: https://www.dray.be/arch/pkgbuilds/bcache-tools-git-0.r83.89f11b1-1.src.tar.gz

That's strange about raid, it wastes cache space. Maybe for raid0? And I know about the resize thing, I have a patch for that that Kent won't accept[1]. The udev stuff is meant to reduce the upstream delta. [1] https://github.com/g2p/linux/commits/bcache

G2P: From the IRC chats, it doesn't sound like they recommend putting bcache on top of other block managers. Specifically, when asked about how to build raid with bcache, they've said it's better to do: /dev/sda1->bcache0->md-raid /dev/sdb1->bcache1->md-raid /dev/sdc1->bcache2->md-raid and then have bcache0, bcache1, and bcache2 all share a single cache device. I would assume the same holds for LLVM build your LLVM on top of many bcache back ends. That's not to say you can't put bcache on LLVM, but since bcache ISN'T runtime dynamic (you need to reboot to resize, etc) it's probably easier to manage things if you treat bcache backends as your low level block device(s). That said, I'll look into this when I get time in the next week or so. The bcache dev is using fedora/redhat and I'd like to keep the changes between upstream and arch as small as possible, though.

The initcpio hook won't work if the backing device is lvm. There's probably a way to make initcpio use the existing udev hooks instead.

Thanks bobpaul, build works as expected. Now I can finally play with my new SSD :D.

Updated PKGBUILD and patch to work with current upstream HEAD. added _gitrev="" variable to check out a specific revision from upstream. The particular revision I built the patches against is commented out, and may be uncommented if this happens again in the future.

Sorry laoras... haven't been on the internet for a few days. It seems upstream changes have broken my patch. My patch is needed to provide the bcache mkinitcpio hook. You only need this hook if your root device is on bcache (it does the echo /dev/DEVICE > /sys/fs/bcache/register part... if your root is not on bcache, then the udev rule will do this later on in the boot process for your other devices) So you can either edit the PKGBUILD and comment out the patch line if you don't need to register at bootup, or you can checkout an older version of upstream that the patch still applies to by editing the PKGBUILD like this: cd "$srcdir/$_gitname-build" git checkout 42c182c6ab1530b5e3b18893f284da3eaec46e74 patch -Np1 -i "${srcdir}/../initcpio-arch.patch" I'll have a fixed patch and PKGBUILD out as soon as I have time, hopefully by tomorrow.