On Fri, Jun 12, 2026 at 3:39 PM Guilherme Ivo Bozi <[email protected]> wrote: > > On Fri, Jun 12, 2026 at 4:34 AM Christian König > <[email protected]> wrote: > > > > On 6/11/26 21:49, Guilherme Ivo Bozi wrote: > > > This series converts the GPIO HW translation logic used by > > > multiple DCN generations from large switch statements to > > > static lookup tables with shared helper functions. > > > > Not to block this patch set, but please keep in mind that when some array > > index comes from userspace to use array_index_nospec() to avoid security > > problems caused by speculative execution. > > > > Thanks for the feedback. I reviewed the relevant GPIO and DDC paths. > In all cases, the offsets and indices are derived from DC BIOS tables > or internal DC GPIO service mappings, not userspace-controlled input. >
I don't have a strong opinion on this. @Leo (Sunpeng) Li , @Wentland, Harry , do you have an opinion? It doesn't seem like it really makes much of a difference either way for code readability, etc. Alex > If you still prefer additional hardening for consistency, I can send a > v2 with array_index_nospec(). > > > We already had problems with that in the past when people converted switch > > statements into an array lockup. > > > > I’ll keep this in mind for other paths where array indices may be > influenced by userspace. > > Kind regards, > Guilherme Ivo > > > > Regards, > > Christian. > > > > > > > > The new implementation reduces duplicated translation logic > > > across generations and makes the GPIO mappings easier to > > > maintain and extend. > > > > > > The series introduces generic GPIO translation helpers first, > > > followed by per-generation conversions. > > > > > > The following generations were converted: > > > > > > - dcn10 > > > - dcn20 > > > - dcn21 > > > - dcn30 > > > - dcn315 > > > - dcn32 > > > - dcn401 > > > - dcn42 > > > > > > The dce60, dce80, dce110 and dce120 implementations were > > > left unchanged for now. > > > > > > dce60 and dce80 contain special switch-case handling paths > > > that currently do not map cleanly to the new table-based > > > representation, so they were intentionally excluded from > > > this refactor. > > > > > > No functional changes intended. > > > > > > v1 resend: > > > > > > No changes since original submission. > > > > > > Automated review: > > > https://lore.gitlab.freedesktop.org/drm-ai-reviews/[email protected]/ > > > > > > Guilherme Ivo Bozi (9): > > > drm/amd/display: add GPIO HW translation helpers > > > drm/amd/display: convert dcn10 GPIO translation to lookup tables > > > drm/amd/display: convert dcn20 GPIO translation to lookup tables > > > drm/amd/display: convert dcn21 GPIO translation to lookup tables > > > drm/amd/display: convert dcn30 GPIO translation to lookup tables > > > drm/amd/display: convert dcn315 GPIO translation to lookup tables > > > drm/amd/display: convert dcn32 GPIO translation to lookup tables > > > drm/amd/display: convert dcn401 GPIO translation to lookup tables > > > drm/amd/display: convert dcn42 GPIO translation to lookup tables > > > > > > .../dc/gpio/dcn10/hw_translate_dcn10.c | 484 +++++++----------- > > > .../dc/gpio/dcn20/hw_translate_dcn20.c | 432 ++++++---------- > > > .../dc/gpio/dcn21/hw_translate_dcn21.c | 417 ++++++--------- > > > .../dc/gpio/dcn30/hw_translate_dcn30.c | 432 ++++++---------- > > > .../dc/gpio/dcn315/hw_translate_dcn315.c | 418 ++++++--------- > > > .../dc/gpio/dcn32/hw_translate_dcn32.c | 386 +++++--------- > > > .../dc/gpio/dcn401/hw_translate_dcn401.c | 392 +++++--------- > > > .../dc/gpio/dcn42/hw_translate_dcn42.c | 193 +++---- > > > .../drm/amd/display/dc/gpio/hw_translate.c | 86 ++++ > > > .../drm/amd/display/dc/gpio/hw_translate.h | 21 + > > > .../gpu/drm/amd/display/include/gpio_types.h | 48 ++ > > > 11 files changed, 1274 insertions(+), 2035 deletions(-) > > > > >
