[PATCH] drm/amd/display: mark static functions noinline_for_stack
Nathan Chancellor
nathan at kernel.org
Thu Jan 16 14:40:49 UTC 2025
Hi Tzung-Bi,
First of all, thanks for the patch!
On Thu, Jan 09, 2025 at 05:35:04AM +0000, Tzung-Bi Shih wrote:
> When compiling allmodconfig (CONFIG_WERROR=y) with clang-19, see the
> following errors:
>
> .../display/dc/dml2/display_mode_core.c:6268:13: warning: stack frame size (3128) exceeds limit (3072) in 'dml_prefetch_check' [-Wframe-larger-than]
> .../display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c:7236:13: warning: stack frame size (3256) exceeds limit (3072) in 'dml_core_mode_support' [-Wframe-larger-than]
>
> Mark static functions called by dml_prefetch_check() and
> dml_core_mode_support() noinline_for_stack to avoid them become huge
> functions and thus exceed the frame size limit.
So I fixed these particular instances by hiding them when certain
sanitizers are enabled:
https://gitlab.freedesktop.org/agd5f/linux/-/commit/e4479aecf6581af81bc0908575447878d2a07e01
However, there are still reports of the regular 2048 limit being hit
with certain configurations such as LTO (the issue number is somewhat
funny given the situation):
https://github.com/ClangBuiltLinux/linux/issues/2048
So I think it would be a good idea for the AMD folks to consider
applying this patch as well.
> A way to reproduce:
> $ git checkout next-20250107
> $ mkdir build_dir
> $ export PATH=/tmp/llvm-19.1.6-x86_64/bin:$PATH
> $ make LLVM=1 O=build_dir allmodconfig
> $ make LLVM=1 O=build_dir drivers/gpu/drm/ -j
>
> The way how it chose static functions to mark:
> [0] Unset CONFIG_WERROR in build_dir/.config.
> To get display_mode_core.o without errors.
>
> [1] Get a function list called by dml_prefetch_check().
> $ sed -n '6268,6711p' drivers/gpu/drm/amd/display/dc/dml2/display_mode_core.c \
> | sed -n -r 's/.*\W(\w+)\(.*/\1/p' | sort -u >/tmp/syms
>
> [2] Get the non-inline function list.
> Objdump won't show the symbols if they are inline functions.
>
> $ make LLVM=1 O=build_dir drivers/gpu/drm/ -j
> $ objdump -d build_dir/.../display_mode_core.o | \
> ./scripts/checkstack.pl x86_64 0 | \
> grep -f /tmp/syms | cut -d' ' -f2- >/tmp/orig
>
> [3] Get the full function list.
> Append "-fno-inline" to `CFLAGS_.../display_mode_core.o` in
> drivers/gpu/drm/amd/display/dc/dml2/Makefile.
>
> $ make LLVM=1 O=build_dir drivers/gpu/drm/ -j
> $ objdump -d build_dir/.../display_mode_core.o | \
> ./scripts/checkstack.pl x86_64 0 | \
> grep -f /tmp/syms | cut -d' ' -f2- >/tmp/noinline
>
> [4] Get the inline function list.
> If a symbol only in /tmp/noinline but not in /tmp/orig, it is a good
> candidate to mark noinline.
>
> $ diff /tmp/orig /tmp/noinline
>
> Chosen functions and their stack sizes:
> CalculateBandwidthAvailableForImmediateFlip [display_mode_core.o]:144
> CalculateExtraLatency [display_mode_core.o]:176
> CalculateTWait [display_mode_core.o]:64
> CalculateVActiveBandwithSupport [display_mode_core.o]:112
> set_calculate_prefetch_schedule_params [display_mode_core.o]:48
>
> CheckGlobalPrefetchAdmissibility [dml2_core_dcn4_calcs.o]:544
> calculate_bandwidth_available [dml2_core_dcn4_calcs.o]:320
> calculate_vactive_det_fill_latency [dml2_core_dcn4_calcs.o]:272
> CalculateDCFCLKDeepSleep [dml2_core_dcn4_calcs.o]:208
> CalculateODMMode [dml2_core_dcn4_calcs.o]:208
> CalculateOutputLink [dml2_core_dcn4_calcs.o]:176
>
> Signed-off-by: Tzung-Bi Shih <tzungbi at kernel.org>
> ---
> .../gpu/drm/amd/display/dc/dml2/display_mode_core.c | 12 ++++++------
> .../dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c | 12 ++++++------
> 2 files changed, 12 insertions(+), 12 deletions(-)
>
> diff --git a/drivers/gpu/drm/amd/display/dc/dml2/display_mode_core.c b/drivers/gpu/drm/amd/display/dc/dml2/display_mode_core.c
> index 35bc917631ae..84a2de9a76d4 100644
> --- a/drivers/gpu/drm/amd/display/dc/dml2/display_mode_core.c
> +++ b/drivers/gpu/drm/amd/display/dc/dml2/display_mode_core.c
> @@ -1736,7 +1736,7 @@ static void CalculateBytePerPixelAndBlockSizes(
> #endif
> } // CalculateBytePerPixelAndBlockSizes
>
> -static dml_float_t CalculateTWait(
> +static noinline_for_stack dml_float_t CalculateTWait(
> dml_uint_t PrefetchMode,
> enum dml_use_mall_for_pstate_change_mode UseMALLForPStateChange,
> dml_bool_t SynchronizeDRRDisplaysForUCLKPStateChangeFinal,
> @@ -4458,7 +4458,7 @@ static void CalculateSwathWidth(
> }
> } // CalculateSwathWidth
>
> -static dml_float_t CalculateExtraLatency(
> +static noinline_for_stack dml_float_t CalculateExtraLatency(
> dml_uint_t RoundTripPingLatencyCycles,
> dml_uint_t ReorderingBytes,
> dml_float_t DCFCLK,
> @@ -5915,7 +5915,7 @@ static dml_uint_t DSCDelayRequirement(
> return DSCDelayRequirement_val;
> }
>
> -static dml_bool_t CalculateVActiveBandwithSupport(dml_uint_t NumberOfActiveSurfaces,
> +static noinline_for_stack dml_bool_t CalculateVActiveBandwithSupport(dml_uint_t NumberOfActiveSurfaces,
> dml_float_t ReturnBW,
> dml_bool_t NotUrgentLatencyHiding[],
> dml_float_t ReadBandwidthLuma[],
> @@ -6019,7 +6019,7 @@ static void CalculatePrefetchBandwithSupport(
> #endif
> }
>
> -static dml_float_t CalculateBandwidthAvailableForImmediateFlip(
> +static noinline_for_stack dml_float_t CalculateBandwidthAvailableForImmediateFlip(
> dml_uint_t NumberOfActiveSurfaces,
> dml_float_t ReturnBW,
> dml_float_t ReadBandwidthLuma[],
> @@ -6213,7 +6213,7 @@ static dml_uint_t CalculateMaxVStartup(
> return max_vstartup_lines;
> }
>
> -static void set_calculate_prefetch_schedule_params(struct display_mode_lib_st *mode_lib,
> +static noinline_for_stack void set_calculate_prefetch_schedule_params(struct display_mode_lib_st *mode_lib,
> struct CalculatePrefetchSchedule_params_st *CalculatePrefetchSchedule_params,
> dml_uint_t j,
> dml_uint_t k)
> @@ -6265,7 +6265,7 @@ static void set_calculate_prefetch_schedule_params(struct display_mode_lib_st *m
> CalculatePrefetchSchedule_params->Tno_bw = &mode_lib->ms.Tno_bw[k];
> }
>
> -static void dml_prefetch_check(struct display_mode_lib_st *mode_lib)
> +static noinline_for_stack void dml_prefetch_check(struct display_mode_lib_st *mode_lib)
> {
> struct dml_core_mode_support_locals_st *s = &mode_lib->scratch.dml_core_mode_support_locals;
> struct CalculatePrefetchSchedule_params_st *CalculatePrefetchSchedule_params = &mode_lib->scratch.CalculatePrefetchSchedule_params;
> diff --git a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c
> index b9ec243cf9ba..7fffca67ca9d 100644
> --- a/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c
> +++ b/drivers/gpu/drm/amd/display/dc/dml2/dml21/src/dml2_core/dml2_core_dcn4_calcs.c
> @@ -2778,7 +2778,7 @@ static double dml_get_return_bandwidth_available(
> return return_bw_mbps;
> }
>
> -static void calculate_bandwidth_available(
> +static noinline_for_stack void calculate_bandwidth_available(
> double avg_bandwidth_available_min[dml2_core_internal_soc_state_max],
> double avg_bandwidth_available[dml2_core_internal_soc_state_max][dml2_core_internal_bw_max],
> double urg_bandwidth_available_min[dml2_core_internal_soc_state_max], // min between SDP and DRAM
> @@ -3531,7 +3531,7 @@ static void CalculateUrgentBurstFactor(
>
> }
>
> -static void CalculateDCFCLKDeepSleep(
> +static noinline_for_stack void CalculateDCFCLKDeepSleep(
> const struct dml2_display_cfg *display_cfg,
> unsigned int NumberOfActiveSurfaces,
> unsigned int BytePerPixelY[],
> @@ -4076,7 +4076,7 @@ static bool ValidateODMMode(enum dml2_odm_mode ODMMode,
> return true;
> }
>
> -static void CalculateODMMode(
> +static noinline_for_stack void CalculateODMMode(
> unsigned int MaximumPixelsPerLinePerDSCUnit,
> unsigned int HActive,
> enum dml2_output_format_class OutFormat,
> @@ -4173,7 +4173,7 @@ static void CalculateODMMode(
> #endif
> }
>
> -static void CalculateOutputLink(
> +static noinline_for_stack void CalculateOutputLink(
> struct dml2_core_internal_scratch *s,
> double PHYCLK,
> double PHYCLKD18,
> @@ -5928,7 +5928,7 @@ static double calculate_impacted_Tsw(unsigned int exclude_plane_idx, unsigned in
> }
>
> // a global check against the aggregate effect of the per plane prefetch schedule
> -static bool CheckGlobalPrefetchAdmissibility(struct dml2_core_internal_scratch *scratch,
> +static noinline_for_stack bool CheckGlobalPrefetchAdmissibility(struct dml2_core_internal_scratch *scratch,
> struct dml2_core_calcs_CheckGlobalPrefetchAdmissibility_params *p)
> {
> struct dml2_core_calcs_CheckGlobalPrefetchAdmissibility_locals *s = &scratch->CheckGlobalPrefetchAdmissibility_locals;
> @@ -6941,7 +6941,7 @@ static void calculate_bytes_to_fetch_required_to_hide_latency(
> }
> }
>
> -static void calculate_vactive_det_fill_latency(
> +static noinline_for_stack void calculate_vactive_det_fill_latency(
> const struct dml2_display_cfg *display_cfg,
> unsigned int num_active_planes,
> unsigned int bytes_required_l[],
> --
> 2.47.1.613.gc27f4b7a9f-goog
>
More information about the dri-devel
mailing list