tree 145d5796beeb757c4717236c29ad6d8a003e6a81
parent 87c42e870da871b8a866e07f5421a8711e988890
author Felix Held <felix-coreboot@felixheld.de> 1699642647 +0100
committer Felix Held <felix-coreboot@felixheld.de> 1700056364 +0000

acpi/acpigen: rework acpigen_pop_len for different size PkgLength

Previously acpigen_pop_len always wrote a 3 byte PkgLength to the 3
bytes reserved by acpigen_write_len_f. After this patch acpigen_pop_len
encodes PkgLength in 1-3 bytes depending on the PkgLength. When less
than the 3 bytes that were previously reserved in the corresponding
acpigen_write_len_f call are needed for PkgLength, the payload data will
be moved back by the number of reserved bytes that aren't needed for the
PkgLength.

This fixes the problem that the Windows AML parser doesn't like a 3 byte
PkgLength being used for the size of the buffer containing UTF-16
strings when the length could be encoded in a single PkgLength byte. In
that case, Windows previously ignored the whole SSDT containing this
larger than necessary PkgLength encoding. It should however be noted
that the ACPI 6.4 spec doesn't specify if it's required to always use
the most compact possible encoding of the PkgLength or not. Since iasl
generates the shortest possible PkgLength encoding, it's also a good
idea to make coreboot's acpigen do the same although it's not required
by the specification.

With this patch applied, Windows still boots on Mandolin and the time it
takes to write the tables doesn't change. To measure the times, the log
level in bs_sample_time was increased to BIOS_CRIT and the console log
level was increased to BIOS_CRIT too to only get those times as output.

BS: BS_WRITE_TABLES run times (exec / console): 8 / 0 ms

Signed-off-by: Felix Held <felix-coreboot@felixheld.de>
Change-Id: Ib897b08a05a7cdc52902d51364246c260ea1f206
Reviewed-on: https://review.coreboot.org/c/coreboot/+/79002
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Matt DeVillier <matt.devillier@gmail.com>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Jérémy Compostella <jeremy.compostella@intel.com>