[PATCH v2] drm/xe/xe_guc_ads: Add nonpriv registers to write list
Lucas De Marchi
lucas.demarchi at intel.com
Fri Nov 15 00:30:47 UTC 2024
On Fri, Nov 01, 2024 at 10:23:50PM +0000, Jonathan Cavitt wrote:
>When performing a guc_mmio_regset_write, we add all the registers in the
>reg_sr list to the save/restore list, but do not do the same for the
>nonpriv registers. Add them in.
>
>v2:
>- Add all NONPRIV registers to avoid undefined behavior (Harrison)
>- s/whitelist/nonpriv
>
>Closes: https://gitlab.freedesktop.org/drm/xe/kernel/issues/2249
>Signed-off-by: Jonathan Cavitt <jonathan.cavitt at intel.com>
>CC: Lucas de Marchi <lucas.demarchi at intel.com>
>CC: Matt Roper <matthew.d.roper at intel.com>
>CC: John Harrison <john.c.harrison at intel.com>
>CC: Umesh Nerlige Ramappa <umesh.nerlige.ramappa at intel.com>
>CC: Ashutosh Dixit <ashutosh.dixit at intel.com>
>---
> drivers/gpu/drm/xe/xe_guc_ads.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
>diff --git a/drivers/gpu/drm/xe/xe_guc_ads.c b/drivers/gpu/drm/xe/xe_guc_ads.c
>index 943146e5b460..b0afb89d9d90 100644
>--- a/drivers/gpu/drm/xe/xe_guc_ads.c
>+++ b/drivers/gpu/drm/xe/xe_guc_ads.c
>@@ -243,6 +243,8 @@ static size_t calculate_regset_size(struct xe_gt *gt)
> xa_for_each(&hwe->reg_sr.xa, sr_idx, sr_entry)
> count++;
>
>+ count += RING_MAX_NONPRIV_SLOTS * XE_NUM_HW_ENGINES;
>+
> count += ADS_REGSET_EXTRA_MAX * XE_NUM_HW_ENGINES;
>
> if (XE_WA(gt, 1607983814))
>@@ -727,6 +729,11 @@ static unsigned int guc_mmio_regset_write(struct xe_guc_ads *ads,
> xa_for_each(&hwe->reg_sr.xa, idx, entry)
> guc_mmio_regset_write_one(ads, regset_map, entry->reg, count++);
>
>+ for (i = 0; i < RING_MAX_NONPRIV_SLOTS; i++)
>+ guc_mmio_regset_write_one(ads, regset_map,
>+ RING_FORCE_TO_NONPRIV(hwe->mmio_base, i),
>+ count++);
I guess this fixes the issue. I'm still wondering why we are changing it
in the ads though.
Ideally we'd do:
a) switch the order in hw_engine_init() so it does:
xe_reg_sr_apply_whitelist(hwe);
xe_reg_sr_apply_mmio(&hwe->reg_sr, gt);
b) change xe_reg_sr_apply_whitelist() so instead of writting the
mmio slots by itself, it writes them to hwe->reg_sr
I **think** it would also work and stop handling the nonpriv slots as
snow flakes. They would also show up in debugfs/.../register-save-restore
file.
Can you give it a try? If it doesn't work, feel free to
add my
Reviewed-by: Lucas De Marchi <lucas.demarchi at intel.com>
and we can try to polish it later.
thanks
Lucas De Marchi
>+
> for (e = extra_regs; e < extra_regs + ARRAY_SIZE(extra_regs); e++) {
> if (e->skip)
> continue;
>--
>2.43.0
>
More information about the Intel-xe
mailing list