On Thu, 21 Mar 2019 at 11:08, Robert Richter <rrichter@xxxxxxxxxxx> wrote: > > On 21.03.19 10:51:34, Ard Biesheuvel wrote: > > On Thu, 21 Mar 2019 at 10:39, Robert Richter <rrichter@xxxxxxxxxxx> wrote: > > > > > > On 20.03.19 23:02:09, Ard Biesheuvel wrote: > > > > On Wed, 20 Mar 2019 at 16:23, Robert Richter <rrichter@xxxxxxxxxxx> wrote: > > > > > > > > > > On 20.03.19 14:16:07, Robert Richter wrote: > > > > > > On 20.03.19 13:05:37, Robert Richter wrote: > > > > > > > @@ -167,6 +167,7 @@ static int __init arm_dmi_init(void) > > > > > > > * itself, depends on dmi_scan_machine() having been called already. > > > > > > > */ > > > > > > > dmi_scan_machine(); > > > > > > > + dmi_memdev_walk(); > > > > > > > if (dmi_available) > > > > > > > dmi_set_dump_stack_arch_desc(); > > > > > > > return 0; > > > > > > > > > > > > After > > > > > > > > > > > > [PATCH] efi/arm: Show SMBIOS bank/device location in cper and > > > > > > ghes error logs > > > > > > > > > > > > wents in for arm/arm64, we can unify the code. See patch below. > > > > > > > > > > V2 with the fix in arm_dmi_init() below. > > > > > > > > > > -Robert > > > > > > > > > > > > > > > -- >8 -- > > > > > From: Robert Richter <rrichter@xxxxxxxxxxx> > > > > > Subject: [PATCH v2] efi: Unify dmi setup code over architectures arm/arm64, > > > > > io64 and x86 > > > > > > > > > > All architectures (arm/arm64, io64 and x86) do the same here, so unify > > > > > the code. > > > > > > > > > > Note: We do not need to call dump_stack_set_arch_desc() in case of > > > > > !dmi_available. Both strings, dmi_ids_string and dump_stack_arch_ > > > > > desc_str are initialized zero and thus nothing would change. > > > > > > > > > > > > > I don't understand the last sentence - we do not need to call > > > > dump_stack_set_arch_desc() when !dmi_available, but we do so anyway, > > > > right? Doesn't that wipe the arch description we set based on the DT > > > > machine name? > > > > > > No, in dmi_setup() we exit early when !dmi_available. So for arm/arm64 > > > nothing changed. But for x86 and ia64 we no longer call dump_stack_ > > > set_arch_desc() in this case. This is ok since both strings, > > > dmi_ids_string and dump_stack_arch_desc_str, are initialized zero and > > > copying one to the other does not change anything. > > > > > > > Ah, of course. Apologies for not reading more carefully. > > > > I'll take this patch via the EFI tree. > > > > It seems to me though that the previous patch makes the memdev_walk() > > call unconditional for ARM, and this change subsequently makes it > > dependent on dmi_available. Should we fix that? > > The first patch has the check in memdev_walk(). So that is looking > good. The check is then moved to dmi_setup() in the 2nd patch. > OK, that works for me. > Thanks for handling this. > Thanks for the contribution.
