| Message ID | 20250507074620.3162747-1-wangziang.ok@bytedance.com |
|---|---|
| State | New |
| Headers | show |
| Series | lib: utils: hsm: Do not fail on EALREADY in rpmi-hsm fixup. | expand |
Hi all, I would like to hear your feedback on this. It is common to have harts grouped by different mailboxes on a multi-socket system. The current HSM implementation can actually suffice this case, with the only minor change in this patch. We have tested it on both QEMU virt machine and our RISC-V project. Best Regards, Ziang > From: "Ziang Wang"<wangziang.ok@bytedance.com> > Date: Wed, May 7, 2025, 15:46 > Subject: [PATCH] lib: utils: hsm: Do not fail on EALREADY in rpmi-hsm fixup. > To: <opensbi@lists.infradead.org> > Cc: "Ziang Wang"<wangziang.ok@bytedance.com> > In case harts are divided into groups that use different > rpmi-hsm channels in different mailboxes, the suspend > state fixup function will return EALREADY on secondary > entry, simply skip on this error. > > Signed-off-by: Ziang Wang <wangziang.ok@bytedance.com> > --- > lib/utils/hsm/fdt_hsm_rpmi.c | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/lib/utils/hsm/fdt_hsm_rpmi.c b/lib/utils/hsm/fdt_hsm_rpmi.c > index cc78436..09d0c91 100644 > --- a/lib/utils/hsm/fdt_hsm_rpmi.c > +++ b/lib/utils/hsm/fdt_hsm_rpmi.c > @@ -330,7 +330,7 @@ skip_suspend_states: > > /* Register HSM fixup callback */ > rc = fdt_register_general_fixup(&rpmi_hsm_fixup); > - if (rc) > + if (rc && rc != SBI_EALREADY) > goto fail_free_susp_state_names; > > /* Register HSM device */ > -- > 2.39.5 >
Oh I think it's better to introduce myself first :D Nice to meet you folks! this is Ziang. I'm an engineer from Bytedance Ltd, currently based in Shanghai, China. We've been constantly interested in the application of RISC-V in server area, plus the combination of OpenSBI and coreboot/Linuxboot as the firmware solution. We are eager to contribute, and are recently experimenting RPMI on our local RISC-V project, got some finding, hence come this tiny patch. Hope my abruptness didn't freak you out :), and this could be the starting point of our contribution. Thanks! Best Regards, Ziang > From: "王子昂"<wangziang.ok@bytedance.com> > Date: Mon, May 12, 2025, 15:30 > Subject: Re: [PATCH] lib: utils: hsm: Do not fail on EALREADY in rpmi-hsm fixup. > To: <opensbi@lists.infradead.org> > Hi all, > I would like to hear your feedback on this. > > It is common to have harts grouped by different mailboxes on a multi-socket system. The current HSM implementation can actually suffice this case, with the only minor change in this patch. We have tested it on both QEMU virt machine and our RISC-V project. > > Best Regards, > Ziang > > > From: "Ziang Wang"<wangziang.ok@bytedance.com> > > Date: Wed, May 7, 2025, 15:46 > > Subject: [PATCH] lib: utils: hsm: Do not fail on EALREADY in rpmi-hsm fixup. > > To: <opensbi@lists.infradead.org> > > Cc: "Ziang Wang"<wangziang.ok@bytedance.com> > > In case harts are divided into groups that use different > > rpmi-hsm channels in different mailboxes, the suspend > > state fixup function will return EALREADY on secondary > > entry, simply skip on this error. > > > > Signed-off-by: Ziang Wang <wangziang.ok@bytedance.com> > > --- > > lib/utils/hsm/fdt_hsm_rpmi.c | 2 +- > > 1 file changed, 1 insertion(+), 1 deletion(-) > > > > diff --git a/lib/utils/hsm/fdt_hsm_rpmi.c b/lib/utils/hsm/fdt_hsm_rpmi.c > > index cc78436..09d0c91 100644 > > --- a/lib/utils/hsm/fdt_hsm_rpmi.c > > +++ b/lib/utils/hsm/fdt_hsm_rpmi.c > > @@ -330,7 +330,7 @@ skip_suspend_states: > > > > /* Register HSM fixup callback */ > > rc = fdt_register_general_fixup(&rpmi_hsm_fixup); > > - if (rc) > > + if (rc && rc != SBI_EALREADY) > > goto fail_free_susp_state_names; > > > > /* Register HSM device */ > > -- > > 2.39.5 > >
diff --git a/lib/utils/hsm/fdt_hsm_rpmi.c b/lib/utils/hsm/fdt_hsm_rpmi.c index cc78436..09d0c91 100644 --- a/lib/utils/hsm/fdt_hsm_rpmi.c +++ b/lib/utils/hsm/fdt_hsm_rpmi.c @@ -330,7 +330,7 @@ skip_suspend_states: /* Register HSM fixup callback */ rc = fdt_register_general_fixup(&rpmi_hsm_fixup); - if (rc) + if (rc && rc != SBI_EALREADY) goto fail_free_susp_state_names; /* Register HSM device */
In case harts are divided into groups that use different rpmi-hsm channels in different mailboxes, the suspend state fixup function will return EALREADY on secondary entry, simply skip on this error. Signed-off-by: Ziang Wang <wangziang.ok@bytedance.com> --- lib/utils/hsm/fdt_hsm_rpmi.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)