Control: tags -1 + confirmed Sorry for not getting back to you earlier.
On 10/10/2016 04:46 AM, Henrique de Moraes Holschuh wrote: > Plase use iucode_tool to generate the early initramfs for Intel > microcode. Not only this can create a smaller initramfs when coupled > with iucode_tool's "--scan-system" functionality, Note that the current code has two modes of operation: generic and specific. The generic mode simply adds _all_ microcode to the image, while the specific mode will select the microcode file for the current running CPU. (It parses /proc/cpuinfo.) Which is basically what --scan-system does. I believe that in terms of size there should be no difference here. But: > it also tweaks the > cpio header of the early-initramfs segment containing the intel > microcode update data to ensure the start of the microcode data payload > will be 16-byte aligned. I didn't know that was required. Thanks! The reason why I implemented this stuff manually is that I really wanted to avoid special-casing the entire pipeline for the initramfs for Intel (vs. AMD). Oh well, now I'll have to do that anyway. > You can concatenate early-initramfs segments together and it just works, Ah, that helps a lot. Thanks for the pointer. > so you could prepend the one created by iucode-tool to the one created > by mktirfs for AMD microcode Only in the "generic" mode, otherwise I'll just concatenate that with the initramfs itself. > (and ACPI table overrides / DT overrides, > if it supports those). Currently it doesn't, but if you can give me a pointer to where these are packaged in Debian, I can also add support for those. (As long as they really are just files to be added to the CPIO archive.) Regards, Christian