On 28.03.2023 17:12, Andrew Cooper wrote: > On 28/03/2023 3:19 pm, Jan Beulich wrote: >> On 27.03.2023 21:41, Andrew Cooper wrote: >>> Right now, the boot flow depends on the second pass over >>> bootstrap_map()/find_cpio_data() altering ucode_blob.data to use the >>> directmap >>> alias of the CPIO module, where previously it caches the early boostrap >>> mapping. >>> >>> If the scan is successful, it will be successful the second time too, but >>> there's no point repeating the work. Cache the module index, offset and >>> size >>> to short circuit things the second time around. >> If the scan failed, it will fail the 2nd time too. Maybe deal with >> this case as well, e.g. by clearing ucode_scan at the end of >> microcode_scan_module() when nothing was found? > > See patch 5. It can only become true then because of how the callers > are arranged.
Right, I've meanwhile seen you do it there. That's fine. Yet I think it could also be done earlier (and if I'm not mistaken also ahead of all of the rearrangements you do). Jan
