On Mon, Jan 24, 2022 at 09:25:34AM +0100, Heinrich Schuchardt wrote: > On 1/24/22 08:23, AKASHI Takahiro wrote: > > On Fri, Jan 21, 2022 at 05:03:03PM +0100, Heinrich Schuchardt wrote: > > > On 1/21/22 16:20, Simon Glass wrote: > > > > Hi Heinrich, > > > > > > > > On Sun, 16 Jan 2022 at 08:14, Heinrich Schuchardt > > > > <heinrich.schucha...@canonical.com> wrote: > > > > > > > > > > Use "%pS" to print text representations of GUIDs. > > > > > > > > > > Signed-off-by: Heinrich Schuchardt <heinrich.schucha...@canonical.com> > > > > > --- > > > > > cmd/efidebug.c | 160 > > > > > ++--------------------------------------- > > > > > include/efi_api.h | 8 +++ > > > > > include/efi_dt_fixup.h | 4 -- > > > > > include/efi_rng.h | 4 -- > > > > > lib/uuid.c | 116 ++++++++++++++++++++++++++++++ > > > > > 5 files changed, 128 insertions(+), 164 deletions(-) > > > > > > > > > > > > > Does this blow up the image size? These strings only in the debug side > > > > before. > > > > > > It was to avoid image size increase that I added > > > +#ifdef CONFIG_CMD_EFIDEBUG > > > > > > > > > > > Having said that, I would much rather see a string than a guid, which > > > > I consider to be little more than an obfuscation. > > > > > > That was my motivation. When debugging a boot failure I set DEBUG in > > > lib/efi_loader/efi_boottime.c and reading GUIDs in the debug output was > > > not > > > helpful. > > > > But setting DEBUG in efi_boottime.c doesn't lead to CONFIG_CMD_EFIDEBUG > > being on in uuid.c. Do we want to have a more direct CONFIG? > > We could use CONFIG_LOGLEVEL >= LOGL_DEBUG because once that loglevel is > set you can use the log command to log one of the EFI related functions.
I'm not sure the log level is the best choice for controlling messages. Showing a human-readable GUID would also be a help in case of errors. > Overall the EFI debugging requires rethinking: I definitely agree, no specific idea though. -Takahiro Akashi > If you simply add '#define DEFINE 1' to the top of > lib/efi_loader/efi_boottime your screen will be flooded by function > calls related to timer events. Probably those high frequency events > should use LOGL_DEBUG_IO. > > Best regards > > Heinrich > > > > > -Takahiro Akashi > > > > > Best regards > > > > > > Heinrich > > > > > > > > > > > Regards, > > > > Simon >