On Tue, Oct 03, 2017 at 08:44:31AM -0400, Rob Clark wrote:
On Tue, Oct 3, 2017 at 7:04 AM, Fabio Estevam festevam@gmail.com wrote:
On Tue, Oct 3, 2017 at 7:57 AM, Rob Clark robdclark@gmail.com wrote:
btw, if I had to take a guess, I'd say that perhaps blksz is smaller than 'legacy_mbr', so maybe rather than allocating blksize, it should be DIV_ROUND_UP(sizeof(legacy_mbr), dev_desc->blksz).. or I guess that could be simplified to not use division if blksz is a power of two
Yes, it does seem to be size related as we are size constraint in SPL.
Just tried your suggestion:
--- a/disk/part_dos.c +++ b/disk/part_dos.c @@ -89,7 +89,9 @@ static int test_block_type(unsigned char *buffer)
static int part_test_dos(struct blk_desc *dev_desc) {
ALLOC_CACHE_ALIGN_BUFFER(legacy_mbr, mbr, dev_desc->blksz);
ALLOC_CACHE_ALIGN_BUFFER(legacy_mbr, mbr,DIV_ROUND_UP(sizeof(legacy_mbr),dev_desc->blksz)); if (blk_dread(dev_desc, 0, 1, (ulong *)mbr) != 1) return -1;and it does work for me :-)
Ok, I guess if blksz can actually be less than the mbr, we probably also need a similar fix in is_gpt_valid() (and also to pass the correct # of blks to blk_dread()).. I'll make a patch in a few..
If you want to re-work / include my changes in https://patchwork.ozlabs.org/patch/820884/ in v2 of https://patchwork.ozlabs.org/patch/820920/ please feel free (and make sure we don't have / introduce more similar ones), thanks!