On 2016-11-22 18:04, Felix Fietkau wrote: > On 2016-11-22 17:57, David Lang wrote: >> On Tue, 22 Nov 2016, Felix Fietkau wrote: >> >>> On 2016-11-22 17:48, David Lang wrote: >>>> On Tue, 22 Nov 2016, Felix Fietkau wrote: >>>> >>>>> On 2016-11-22 17:43, David Lang wrote: >>>>>> On Tue, 22 Nov 2016, Felix Fietkau wrote: >>>>>> >>>>>>>> On a 16M filesystem, we probably want to use a 1K block size and have >>>>>>>> an inode >>>>>>>> for every couple of blocks. >>>>>>> I'd say on a 16M filesystem we probably want to use squashfs+ext4 >>>>>>> instead of plain ext4 and avoid the inode issue altogether. >>>>>> >>>>>> I'm not sure I understand how this would work? >>>>> Pad a plain squashfs image to the intended target size, include mke2fs >>>>> and mkf2fs (for bigger sizes) in the image. >>>>> fstools will take care of the rest at boot time. >>>> >>>> you still have to set the parameters for mkfs to use. >>> It will be created as an empty overlay filesystem, so there's lots of >>> free inodes available. >> >> but by default, it wouldn't have lots of free inodes, the default is one >> inode >> per 16K of filesystem, which is how we get to 1024 inodes on a 16M >> filesystem. >> >> We need to be able to say that this is a tiny filesystem, and should really >> use >> a smaller blocksize and more inodes. > FWIW, I just pushed a commit to my staging tree that adds squashfs > support for the uml target. > I tested it and it's usable with 16M rootfs. Here's some more information: it creates the overlay filesystem with a block size of 1K, and it has 3493 free inodes after booting up.
- Felix _______________________________________________ Lede-dev mailing list Lede-dev@lists.infradead.org http://lists.infradead.org/mailman/listinfo/lede-dev
