From bfecbc2cfba9b06d67d9d249c33d92e570e2fa70 Mon Sep 17 00:00:00 2001 From: Dmitry Baryshkov Date: Tue, 3 Sep 2024 06:22:44 +0300 Subject: drm/msm/dpu: make sure phys resources are properly initialized The commit b954fa6baaca ("drm/msm/dpu: Refactor rm iterator") removed zero-init of the hw_ctl array, but didn't change the error condition, that checked for hw_ctl[i] being NULL. At the same time because of the early returns in case of an error dpu_encoder_phys might be left with the resources assigned in the previous state. Rework assigning of hw_pp / hw_ctl to the dpu_encoder_phys in order to make sure they are always set correctly. Fixes: b954fa6baaca ("drm/msm/dpu: Refactor rm iterator") Suggested-by: Abhinav Kumar Signed-off-by: Dmitry Baryshkov Reviewed-by: Abhinav Kumar Patchwork: https://patchwork.freedesktop.org/patch/612233/ Link: https://lore.kernel.org/r/20240903-dpu-mode-config-width-v6-1-617e1ecc4b7a@linaro.org Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 9 ++++----- 1 file changed, 4 insertions(+), 5 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c index 3b171bf227d1..949ebda2fa82 100644 --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c @@ -1174,21 +1174,20 @@ static void dpu_encoder_virt_atomic_mode_set(struct drm_encoder *drm_enc, for (i = 0; i < dpu_enc->num_phys_encs; i++) { struct dpu_encoder_phys *phys = dpu_enc->phys_encs[i]; - if (!dpu_enc->hw_pp[i]) { + phys->hw_pp = dpu_enc->hw_pp[i]; + if (!phys->hw_pp) { DPU_ERROR_ENC(dpu_enc, "no pp block assigned at idx: %d\n", i); return; } - if (!hw_ctl[i]) { + phys->hw_ctl = i < num_ctl ? to_dpu_hw_ctl(hw_ctl[i]) : NULL; + if (!phys->hw_ctl) { DPU_ERROR_ENC(dpu_enc, "no ctl block assigned at idx: %d\n", i); return; } - phys->hw_pp = dpu_enc->hw_pp[i]; - phys->hw_ctl = to_dpu_hw_ctl(hw_ctl[i]); - phys->cached_mode = crtc_state->adjusted_mode; if (phys->ops.atomic_mode_set) phys->ops.atomic_mode_set(phys, crtc_state, conn_state); -- cgit v1.2.3 From 3ae133b0192b9b0c9f560bbc096887053150195f Mon Sep 17 00:00:00 2001 From: Dmitry Baryshkov Date: Tue, 3 Sep 2024 06:22:45 +0300 Subject: drm/msm/dpu: move CRTC resource assignment to dpu_encoder_virt_atomic_check Historically CRTC resources (LMs and CTLs) were assigned in dpu_crtc_atomic_begin(). The commit 9222cdd27e82 ("drm/msm/dpu: move hw resource tracking to crtc state") simply moved resources to struct dpu_crtc_state, without changing the code sequence. Later on the commit b107603b4ad0 ("drm/msm/dpu: map mixer/ctl hw blocks in encoder modeset") rearanged the code, but still kept the cstate->num_mixers assignment to happen during commit phase. This makes dpu_crtc_state inconsistent between consequent atomic_check() calls. Move CRTC resource assignment to happen at the end of dpu_encoder_virt_atomic_check(). Fixes: b107603b4ad0 ("drm/msm/dpu: map mixer/ctl hw blocks in encoder modeset") Signed-off-by: Dmitry Baryshkov Reviewed-by: Abhinav Kumar Patchwork: https://patchwork.freedesktop.org/patch/612235/ Link: https://lore.kernel.org/r/20240903-dpu-mode-config-width-v6-2-617e1ecc4b7a@linaro.org Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c | 3 -- drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 59 +++++++++++++++++++---------- 2 files changed, 38 insertions(+), 24 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c index 4c1be2f0555f..e81feb0d67f3 100644 --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c @@ -1091,9 +1091,6 @@ static void dpu_crtc_disable(struct drm_crtc *crtc, dpu_core_perf_crtc_update(crtc, 0); - memset(cstate->mixers, 0, sizeof(cstate->mixers)); - cstate->num_mixers = 0; - /* disable clk & bw control until clk & bw properties are set */ cstate->bw_control = false; cstate->bw_split_vote = false; diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c index 949ebda2fa82..bd3698bf0cf7 100644 --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c @@ -624,6 +624,40 @@ static struct msm_display_topology dpu_encoder_get_topology( return topology; } +static void dpu_encoder_assign_crtc_resources(struct dpu_kms *dpu_kms, + struct drm_encoder *drm_enc, + struct dpu_global_state *global_state, + struct drm_crtc_state *crtc_state) +{ + struct dpu_crtc_state *cstate; + struct dpu_hw_blk *hw_ctl[MAX_CHANNELS_PER_ENC]; + struct dpu_hw_blk *hw_lm[MAX_CHANNELS_PER_ENC]; + struct dpu_hw_blk *hw_dspp[MAX_CHANNELS_PER_ENC]; + int num_lm, num_ctl, num_dspp, i; + + cstate = to_dpu_crtc_state(crtc_state); + + memset(cstate->mixers, 0, sizeof(cstate->mixers)); + + num_ctl = dpu_rm_get_assigned_resources(&dpu_kms->rm, global_state, + drm_enc->base.id, DPU_HW_BLK_CTL, hw_ctl, ARRAY_SIZE(hw_ctl)); + num_lm = dpu_rm_get_assigned_resources(&dpu_kms->rm, global_state, + drm_enc->base.id, DPU_HW_BLK_LM, hw_lm, ARRAY_SIZE(hw_lm)); + num_dspp = dpu_rm_get_assigned_resources(&dpu_kms->rm, global_state, + drm_enc->base.id, DPU_HW_BLK_DSPP, hw_dspp, + ARRAY_SIZE(hw_dspp)); + + for (i = 0; i < num_lm; i++) { + int ctl_idx = (i < num_ctl) ? i : (num_ctl-1); + + cstate->mixers[i].hw_lm = to_dpu_hw_mixer(hw_lm[i]); + cstate->mixers[i].lm_ctl = to_dpu_hw_ctl(hw_ctl[ctl_idx]); + cstate->mixers[i].hw_dspp = i < num_dspp ? to_dpu_hw_dspp(hw_dspp[i]) : NULL; + } + + cstate->num_mixers = num_lm; +} + static int dpu_encoder_virt_atomic_check( struct drm_encoder *drm_enc, struct drm_crtc_state *crtc_state, @@ -692,6 +726,9 @@ static int dpu_encoder_virt_atomic_check( if (!crtc_state->active_changed || crtc_state->enable) ret = dpu_rm_reserve(&dpu_kms->rm, global_state, drm_enc, crtc_state, topology); + if (!ret) + dpu_encoder_assign_crtc_resources(dpu_kms, drm_enc, + global_state, crtc_state); } trace_dpu_enc_atomic_check_flags(DRMID(drm_enc), adj_mode->flags); @@ -1093,14 +1130,11 @@ static void dpu_encoder_virt_atomic_mode_set(struct drm_encoder *drm_enc, struct dpu_encoder_virt *dpu_enc; struct msm_drm_private *priv; struct dpu_kms *dpu_kms; - struct dpu_crtc_state *cstate; struct dpu_global_state *global_state; struct dpu_hw_blk *hw_pp[MAX_CHANNELS_PER_ENC]; struct dpu_hw_blk *hw_ctl[MAX_CHANNELS_PER_ENC]; - struct dpu_hw_blk *hw_lm[MAX_CHANNELS_PER_ENC]; - struct dpu_hw_blk *hw_dspp[MAX_CHANNELS_PER_ENC] = { NULL }; struct dpu_hw_blk *hw_dsc[MAX_CHANNELS_PER_ENC]; - int num_lm, num_ctl, num_pp, num_dsc; + int num_ctl, num_pp, num_dsc; unsigned int dsc_mask = 0; int i; @@ -1129,11 +1163,6 @@ static void dpu_encoder_virt_atomic_mode_set(struct drm_encoder *drm_enc, ARRAY_SIZE(hw_pp)); num_ctl = dpu_rm_get_assigned_resources(&dpu_kms->rm, global_state, drm_enc->base.id, DPU_HW_BLK_CTL, hw_ctl, ARRAY_SIZE(hw_ctl)); - num_lm = dpu_rm_get_assigned_resources(&dpu_kms->rm, global_state, - drm_enc->base.id, DPU_HW_BLK_LM, hw_lm, ARRAY_SIZE(hw_lm)); - dpu_rm_get_assigned_resources(&dpu_kms->rm, global_state, - drm_enc->base.id, DPU_HW_BLK_DSPP, hw_dspp, - ARRAY_SIZE(hw_dspp)); for (i = 0; i < MAX_CHANNELS_PER_ENC; i++) dpu_enc->hw_pp[i] = i < num_pp ? to_dpu_hw_pingpong(hw_pp[i]) @@ -1159,18 +1188,6 @@ static void dpu_encoder_virt_atomic_mode_set(struct drm_encoder *drm_enc, dpu_enc->cur_master->hw_cdm = hw_cdm ? to_dpu_hw_cdm(hw_cdm) : NULL; } - cstate = to_dpu_crtc_state(crtc_state); - - for (i = 0; i < num_lm; i++) { - int ctl_idx = (i < num_ctl) ? i : (num_ctl-1); - - cstate->mixers[i].hw_lm = to_dpu_hw_mixer(hw_lm[i]); - cstate->mixers[i].lm_ctl = to_dpu_hw_ctl(hw_ctl[ctl_idx]); - cstate->mixers[i].hw_dspp = to_dpu_hw_dspp(hw_dspp[i]); - } - - cstate->num_mixers = num_lm; - for (i = 0; i < dpu_enc->num_phys_encs; i++) { struct dpu_encoder_phys *phys = dpu_enc->phys_encs[i]; -- cgit v1.2.3 From 3a0851b442d1f63ba42ecfa2506d3176cfabf9d4 Mon Sep 17 00:00:00 2001 From: Dmitry Baryshkov Date: Tue, 3 Sep 2024 06:22:46 +0300 Subject: drm/msm/dpu: check for overflow in _dpu_crtc_setup_lm_bounds() Make _dpu_crtc_setup_lm_bounds() check that CRTC width is not overflowing LM requirements. Rename the function accordingly. Fixes: 25fdd5933e4c ("drm/msm: Add SDM845 DPU support") Reviewed-by: Abhinav Kumar Tested-by: Abhinav Kumar # sc7280 Signed-off-by: Dmitry Baryshkov Patchwork: https://patchwork.freedesktop.org/patch/612237/ Link: https://lore.kernel.org/r/20240903-dpu-mode-config-width-v6-3-617e1ecc4b7a@linaro.org Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c | 17 +++++++++++++---- 1 file changed, 13 insertions(+), 4 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c index e81feb0d67f3..db6c57900781 100644 --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_crtc.c @@ -711,12 +711,13 @@ void dpu_crtc_complete_commit(struct drm_crtc *crtc) _dpu_crtc_complete_flip(crtc); } -static void _dpu_crtc_setup_lm_bounds(struct drm_crtc *crtc, +static int _dpu_crtc_check_and_setup_lm_bounds(struct drm_crtc *crtc, struct drm_crtc_state *state) { struct dpu_crtc_state *cstate = to_dpu_crtc_state(state); struct drm_display_mode *adj_mode = &state->adjusted_mode; u32 crtc_split_width = adj_mode->hdisplay / cstate->num_mixers; + struct dpu_kms *dpu_kms = _dpu_crtc_get_kms(crtc); int i; for (i = 0; i < cstate->num_mixers; i++) { @@ -727,7 +728,12 @@ static void _dpu_crtc_setup_lm_bounds(struct drm_crtc *crtc, r->y2 = adj_mode->vdisplay; trace_dpu_crtc_setup_lm_bounds(DRMID(crtc), i, r); + + if (drm_rect_width(r) > dpu_kms->catalog->caps->max_mixer_width) + return -E2BIG; } + + return 0; } static void _dpu_crtc_get_pcc_coeff(struct drm_crtc_state *state, @@ -803,7 +809,7 @@ static void dpu_crtc_atomic_begin(struct drm_crtc *crtc, DRM_DEBUG_ATOMIC("crtc%d\n", crtc->base.id); - _dpu_crtc_setup_lm_bounds(crtc, crtc->state); + _dpu_crtc_check_and_setup_lm_bounds(crtc, crtc->state); /* encoder will trigger pending mask now */ drm_for_each_encoder_mask(encoder, crtc->dev, crtc->state->encoder_mask) @@ -1189,8 +1195,11 @@ static int dpu_crtc_atomic_check(struct drm_crtc *crtc, if (crtc_state->active_changed) crtc_state->mode_changed = true; - if (cstate->num_mixers) - _dpu_crtc_setup_lm_bounds(crtc, crtc_state); + if (cstate->num_mixers) { + rc = _dpu_crtc_check_and_setup_lm_bounds(crtc, crtc_state); + if (rc) + return rc; + } /* FIXME: move this to dpu_plane_atomic_check? */ drm_atomic_crtc_state_for_each_plane_state(plane, pstate, crtc_state) { -- cgit v1.2.3 From f260ed880c1fbe6bd7fab47708a659e9df2bdbde Mon Sep 17 00:00:00 2001 From: Dmitry Baryshkov Date: Sun, 22 Sep 2024 00:14:48 +0300 Subject: drm/msm/hdmi: drop pll_cmp_to_fdata from hdmi_phy_8998 The pll_cmp_to_fdata() was never used by the working code. Drop it to prevent warnings with W=1 and clang. Reported-by: Jani Nikula Closes: https://lore.kernel.org/dri-devel/3553b1db35665e6ff08592e35eb438a574d1ad65.1725962479.git.jani.nikula@intel.com Signed-off-by: Dmitry Baryshkov Reviewed-by: Jani Nikula Fixes: caedbf17c48d ("drm/msm: add msm8998 hdmi phy/pll support") Reviewed-by: Abhinav Kumar Patchwork: https://patchwork.freedesktop.org/patch/615348/ Link: https://lore.kernel.org/r/20240922-msm-drop-unused-func-v1-1-c5dc083415b8@linaro.org Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/hdmi/hdmi_phy_8998.c | 9 --------- 1 file changed, 9 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/hdmi/hdmi_phy_8998.c b/drivers/gpu/drm/msm/hdmi/hdmi_phy_8998.c index 0e3a2b16a2ce..e6ffaf92d26d 100644 --- a/drivers/gpu/drm/msm/hdmi/hdmi_phy_8998.c +++ b/drivers/gpu/drm/msm/hdmi/hdmi_phy_8998.c @@ -153,15 +153,6 @@ static inline u32 pll_get_pll_cmp(u64 fdata, unsigned long ref_clk) return dividend - 1; } -static inline u64 pll_cmp_to_fdata(u32 pll_cmp, unsigned long ref_clk) -{ - u64 fdata = ((u64)pll_cmp) * ref_clk * 10; - - do_div(fdata, HDMI_PLL_CMP_CNT); - - return fdata; -} - #define HDMI_REF_CLOCK_HZ ((u64)19200000) #define HDMI_MHZ_TO_HZ ((u64)1000000) static int pll_get_post_div(struct hdmi_8998_post_divider *pd, u64 bclk) -- cgit v1.2.3 From 24436a540d16ca6a523b8e5441180001c31b6b35 Mon Sep 17 00:00:00 2001 From: Jonathan Marek Date: Mon, 7 Oct 2024 01:01:48 -0400 Subject: drm/msm/dsi: improve/fix dsc pclk calculation drm_mode_vrefresh() can introduce a large rounding error, avoid it. Fixes: 7c9e4a554d4a ("drm/msm/dsi: Reduce pclk rate for compression") Signed-off-by: Jonathan Marek Reviewed-by: Dmitry Baryshkov Reviewed-by: Abhinav Kumar Patchwork: https://patchwork.freedesktop.org/patch/618432/ Link: https://lore.kernel.org/r/20241007050157.26855-1-jonathan@marek.ca Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/dsi/dsi_host.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/dsi/dsi_host.c b/drivers/gpu/drm/msm/dsi/dsi_host.c index 185d7de0bf37..1205aa398e44 100644 --- a/drivers/gpu/drm/msm/dsi/dsi_host.c +++ b/drivers/gpu/drm/msm/dsi/dsi_host.c @@ -542,7 +542,7 @@ static unsigned long dsi_adjust_pclk_for_compression(const struct drm_display_mo int new_htotal = mode->htotal - mode->hdisplay + new_hdisplay; - return new_htotal * mode->vtotal * drm_mode_vrefresh(mode); + return mult_frac(mode->clock * 1000u, new_htotal, mode->htotal); } static unsigned long dsi_get_pclk_rate(const struct drm_display_mode *mode, -- cgit v1.2.3 From 358b762400bd94db2a14a72dfcef74c7da6bd845 Mon Sep 17 00:00:00 2001 From: Jonathan Marek Date: Mon, 7 Oct 2024 01:01:49 -0400 Subject: drm/msm/dsi: fix 32-bit signed integer extension in pclk_rate calculation When (mode->clock * 1000) is larger than (1<<31), int to unsigned long conversion will sign extend the int to 64 bits and the pclk_rate value will be incorrect. Fix this by making the result of the multiplication unsigned. Note that above (1<<32) would still be broken and require more changes, but its unlikely anyone will need that anytime soon. Fixes: c4d8cfe516dc ("drm/msm/dsi: add implementation for helper functions") Signed-off-by: Jonathan Marek Reviewed-by: Dmitry Baryshkov Reviewed-by: Abhinav Kumar Patchwork: https://patchwork.freedesktop.org/patch/618434/ Link: https://lore.kernel.org/r/20241007050157.26855-2-jonathan@marek.ca Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/dsi/dsi_host.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/dsi/dsi_host.c b/drivers/gpu/drm/msm/dsi/dsi_host.c index 1205aa398e44..a98d24b7cb00 100644 --- a/drivers/gpu/drm/msm/dsi/dsi_host.c +++ b/drivers/gpu/drm/msm/dsi/dsi_host.c @@ -550,7 +550,7 @@ static unsigned long dsi_get_pclk_rate(const struct drm_display_mode *mode, { unsigned long pclk_rate; - pclk_rate = mode->clock * 1000; + pclk_rate = mode->clock * 1000u; if (dsc) pclk_rate = dsi_adjust_pclk_for_compression(mode, dsc); -- cgit v1.2.3 From 293f53263266bc4340d777268ab4328a97f041fa Mon Sep 17 00:00:00 2001 From: Douglas Anderson Date: Mon, 14 Oct 2024 09:36:08 -0700 Subject: drm/msm: Avoid NULL dereference in msm_disp_state_print_regs() If the allocation in msm_disp_state_dump_regs() failed then `block->state` can be NULL. The msm_disp_state_print_regs() function _does_ have code to try to handle it with: if (*reg) dump_addr = *reg; ...but since "dump_addr" is initialized to NULL the above is actually a noop. The code then goes on to dereference `dump_addr`. Make the function print "Registers not stored" when it sees a NULL to solve this. Since we're touching the code, fix msm_disp_state_print_regs() not to pointlessly take a double-pointer and properly mark the pointer as `const`. Fixes: 98659487b845 ("drm/msm: add support to take dpu snapshot") Signed-off-by: Douglas Anderson Reviewed-by: Abhinav Kumar Patchwork: https://patchwork.freedesktop.org/patch/619657/ Link: https://lore.kernel.org/r/20241014093605.1.Ia1217cecec9ef09eb3c6d125360cc6c8574b0e73@changeid Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c b/drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c index add72bbc28b1..bb149281d31f 100644 --- a/drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c +++ b/drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c @@ -48,20 +48,21 @@ static void msm_disp_state_dump_regs(u32 **reg, u32 aligned_len, void __iomem *b } } -static void msm_disp_state_print_regs(u32 **reg, u32 len, void __iomem *base_addr, - struct drm_printer *p) +static void msm_disp_state_print_regs(const u32 *dump_addr, u32 len, + void __iomem *base_addr, struct drm_printer *p) { int i; - u32 *dump_addr = NULL; void __iomem *addr; u32 num_rows; + if (!dump_addr) { + drm_printf(p, "Registers not stored\n"); + return; + } + addr = base_addr; num_rows = len / REG_DUMP_ALIGN; - if (*reg) - dump_addr = *reg; - for (i = 0; i < num_rows; i++) { drm_printf(p, "0x%lx : %08x %08x %08x %08x\n", (unsigned long)(addr - base_addr), @@ -89,7 +90,7 @@ void msm_disp_state_print(struct msm_disp_state *state, struct drm_printer *p) list_for_each_entry_safe(block, tmp, &state->blocks, node) { drm_printf(p, "====================%s================\n", block->name); - msm_disp_state_print_regs(&block->state, block->size, block->base_addr, p); + msm_disp_state_print_regs(block->state, block->size, block->base_addr, p); } drm_printf(p, "===================dpu drm state================\n"); -- cgit v1.2.3 From e4a45582db1b792c57bdb52c45958264f7fcfbdc Mon Sep 17 00:00:00 2001 From: Douglas Anderson Date: Mon, 14 Oct 2024 09:36:09 -0700 Subject: drm/msm: Allocate memory for disp snapshot with kvzalloc() With the "drm/msm: add a display mmu fault handler" series [1] we saw issues in the field where memory allocation was failing when allocating space for registers in msm_disp_state_dump_regs(). Specifically we were seeing an order 5 allocation fail. It's not surprising that order 5 allocations will sometimes fail after the system has been up and running for a while. There's no need here for contiguous memory. Change the allocation to kvzalloc() which should make it much less likely to fail. [1] https://lore.kernel.org/r/20240628214848.4075651-1-quic_abhinavk@quicinc.com/ Fixes: 98659487b845 ("drm/msm: add support to take dpu snapshot") Signed-off-by: Douglas Anderson Reviewed-by: Abhinav Kumar Patchwork: https://patchwork.freedesktop.org/patch/619658/ Link: https://lore.kernel.org/r/20241014093605.2.I72441365ffe91f3dceb17db0a8ec976af8139590@changeid Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c b/drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c index bb149281d31f..4d55e3cf570f 100644 --- a/drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c +++ b/drivers/gpu/drm/msm/disp/msm_disp_snapshot_util.c @@ -26,7 +26,7 @@ static void msm_disp_state_dump_regs(u32 **reg, u32 aligned_len, void __iomem *b end_addr = base_addr + aligned_len; if (!(*reg)) - *reg = kzalloc(len_padded, GFP_KERNEL); + *reg = kvzalloc(len_padded, GFP_KERNEL); if (*reg) dump_addr = *reg; @@ -162,7 +162,7 @@ void msm_disp_state_free(void *data) list_for_each_entry_safe(block, tmp, &disp_state->blocks, node) { list_del(&block->node); - kfree(block->state); + kvfree(block->state); kfree(block); } -- cgit v1.2.3 From 40dad89cb86ce824f2080441b2a6b7aedf695329 Mon Sep 17 00:00:00 2001 From: Jessica Zhang Date: Wed, 9 Oct 2024 20:41:13 -0700 Subject: drm/msm/dpu: Don't always set merge_3d pending flush Don't set the merge_3d pending flush bits if the mode_3d is BLEND_3D_NONE. Always flushing merge_3d can cause timeout issues when there are multiple commits with concurrent writeback enabled. This is because the video phys enc waits for the hw_ctl flush register to be completely cleared [1] in its wait_for_commit_done(), but the WB encoder always sets the merge_3d pending flush during each commit regardless of if the merge_3d is actually active. This means that the hw_ctl flush register will never be 0 when there are multiple CWB commits and the video phys enc will hit vblank timeout errors after the first CWB commit. [1] commit fe9df3f50c39 ("drm/msm/dpu: add real wait_for_commit_done()") Fixes: 3e79527a33a8 ("drm/msm/dpu: enable merge_3d support on sm8150/sm8250") Fixes: d7d0e73f7de3 ("drm/msm/dpu: introduce the dpu_encoder_phys_* for writeback") Signed-off-by: Jessica Zhang Reviewed-by: Dmitry Baryshkov Patchwork: https://patchwork.freedesktop.org/patch/619092/ Link: https://lore.kernel.org/r/20241009-mode3d-fix-v1-1-c0258354fadc@quicinc.com Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c | 5 ++++- drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_wb.c | 5 ++++- 2 files changed, 8 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c index ba8878d21cf0..8864ace938e0 100644 --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c @@ -440,10 +440,12 @@ static void dpu_encoder_phys_vid_enable(struct dpu_encoder_phys *phys_enc) struct dpu_hw_ctl *ctl; const struct msm_format *fmt; u32 fmt_fourcc; + u32 mode_3d; ctl = phys_enc->hw_ctl; fmt_fourcc = dpu_encoder_get_drm_fmt(phys_enc); fmt = mdp_get_format(&phys_enc->dpu_kms->base, fmt_fourcc, 0); + mode_3d = dpu_encoder_helper_get_3d_blend_mode(phys_enc); DPU_DEBUG_VIDENC(phys_enc, "\n"); @@ -466,7 +468,8 @@ static void dpu_encoder_phys_vid_enable(struct dpu_encoder_phys *phys_enc) goto skip_flush; ctl->ops.update_pending_flush_intf(ctl, phys_enc->hw_intf->idx); - if (ctl->ops.update_pending_flush_merge_3d && phys_enc->hw_pp->merge_3d) + if (mode_3d && ctl->ops.update_pending_flush_merge_3d && + phys_enc->hw_pp->merge_3d) ctl->ops.update_pending_flush_merge_3d(ctl, phys_enc->hw_pp->merge_3d->idx); if (ctl->ops.update_pending_flush_cdm && phys_enc->hw_cdm) diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_wb.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_wb.c index 882c717859ce..07035ab77b79 100644 --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_wb.c +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_wb.c @@ -275,6 +275,7 @@ static void _dpu_encoder_phys_wb_update_flush(struct dpu_encoder_phys *phys_enc) struct dpu_hw_pingpong *hw_pp; struct dpu_hw_cdm *hw_cdm; u32 pending_flush = 0; + u32 mode_3d; if (!phys_enc) return; @@ -283,6 +284,7 @@ static void _dpu_encoder_phys_wb_update_flush(struct dpu_encoder_phys *phys_enc) hw_pp = phys_enc->hw_pp; hw_ctl = phys_enc->hw_ctl; hw_cdm = phys_enc->hw_cdm; + mode_3d = dpu_encoder_helper_get_3d_blend_mode(phys_enc); DPU_DEBUG("[wb:%d]\n", hw_wb->idx - WB_0); @@ -294,7 +296,8 @@ static void _dpu_encoder_phys_wb_update_flush(struct dpu_encoder_phys *phys_enc) if (hw_ctl->ops.update_pending_flush_wb) hw_ctl->ops.update_pending_flush_wb(hw_ctl, hw_wb->idx); - if (hw_ctl->ops.update_pending_flush_merge_3d && hw_pp && hw_pp->merge_3d) + if (mode_3d && hw_ctl->ops.update_pending_flush_merge_3d && + hw_pp && hw_pp->merge_3d) hw_ctl->ops.update_pending_flush_merge_3d(hw_ctl, hw_pp->merge_3d->idx); -- cgit v1.2.3 From f87f3b80abaf7949e638dd17dfdc267066eb52d5 Mon Sep 17 00:00:00 2001 From: Jessica Zhang Date: Wed, 9 Oct 2024 20:46:19 -0700 Subject: drm/msm/dpu: don't always program merge_3d block Only program the merge_3d block for the video phys encoder when the 3d blend mode is not NONE Fixes: 3e79527a33a8 ("drm/msm/dpu: enable merge_3d support on sm8150/sm8250") Suggested-by: Abhinav Kumar Signed-off-by: Jessica Zhang Patchwork: https://patchwork.freedesktop.org/patch/619095/ Link: https://lore.kernel.org/r/20241009-merge3d-fix-v1-1-0d0b6f5c244e@quicinc.com Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c index 8864ace938e0..d8a2edebfe8c 100644 --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder_phys_vid.c @@ -302,7 +302,7 @@ static void dpu_encoder_phys_vid_setup_timing_engine( intf_cfg.stream_sel = 0; /* Don't care value for video mode */ intf_cfg.mode_3d = dpu_encoder_helper_get_3d_blend_mode(phys_enc); intf_cfg.dsc = dpu_encoder_helper_get_dsc(phys_enc); - if (phys_enc->hw_pp->merge_3d) + if (intf_cfg.mode_3d && phys_enc->hw_pp->merge_3d) intf_cfg.merge_3d = phys_enc->hw_pp->merge_3d->idx; spin_lock_irqsave(phys_enc->enc_spinlock, lock_flags); -- cgit v1.2.3 From 77ad507dbb7ec1ecd60fc081d03616960ef596fd Mon Sep 17 00:00:00 2001 From: Rob Clark Date: Tue, 15 Oct 2024 15:13:34 -0700 Subject: drm/msm/a6xx+: Insert a fence wait before SMMU table update The CP_SMMU_TABLE_UPDATE _should_ be waiting for idle, but on some devices (x1-85, possibly others), it seems to pass that barrier while there are still things in the event completion FIFO waiting to be written back to memory. Work around that by adding a fence wait before context switch. The CP_EVENT_WRITE that writes the fence is the last write from a submit, so seeing this value hit memory is a reliable indication that it is safe to proceed with the context switch. v2: Only emit CP_WAIT_TIMESTAMP on a7xx, as it is not supported on a6xx. Conversely, I've not been able to reproduce this issue on a6xx, so hopefully it is limited to a7xx, or perhaps just certain a7xx devices. Fixes: af66706accdf ("drm/msm/a6xx: Add skeleton A7xx support") Closes: https://gitlab.freedesktop.org/drm/msm/-/issues/63 Signed-off-by: Rob Clark Reviewed-by: Akhil P Oommen Signed-off-by: Abhinav Kumar --- drivers/gpu/drm/msm/adreno/a6xx_gpu.c | 16 +++++++++++++--- 1 file changed, 13 insertions(+), 3 deletions(-) (limited to 'drivers') diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c index 06cab2c6fd66..702b8d4b3497 100644 --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu.c +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu.c @@ -101,9 +101,10 @@ static void get_stats_counter(struct msm_ringbuffer *ring, u32 counter, } static void a6xx_set_pagetable(struct a6xx_gpu *a6xx_gpu, - struct msm_ringbuffer *ring, struct msm_file_private *ctx) + struct msm_ringbuffer *ring, struct msm_gem_submit *submit) { bool sysprof = refcount_read(&a6xx_gpu->base.base.sysprof_active) > 1; + struct msm_file_private *ctx = submit->queue->ctx; struct adreno_gpu *adreno_gpu = &a6xx_gpu->base; phys_addr_t ttbr; u32 asid; @@ -115,6 +116,15 @@ static void a6xx_set_pagetable(struct a6xx_gpu *a6xx_gpu, if (msm_iommu_pagetable_params(ctx->aspace->mmu, &ttbr, &asid)) return; + if (adreno_gpu->info->family >= ADRENO_7XX_GEN1) { + /* Wait for previous submit to complete before continuing: */ + OUT_PKT7(ring, CP_WAIT_TIMESTAMP, 4); + OUT_RING(ring, 0); + OUT_RING(ring, lower_32_bits(rbmemptr(ring, fence))); + OUT_RING(ring, upper_32_bits(rbmemptr(ring, fence))); + OUT_RING(ring, submit->seqno - 1); + } + if (!sysprof) { if (!adreno_is_a7xx(adreno_gpu)) { /* Turn off protected mode to write to special registers */ @@ -193,7 +203,7 @@ static void a6xx_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit) struct msm_ringbuffer *ring = submit->ring; unsigned int i, ibs = 0; - a6xx_set_pagetable(a6xx_gpu, ring, submit->queue->ctx); + a6xx_set_pagetable(a6xx_gpu, ring, submit); get_stats_counter(ring, REG_A6XX_RBBM_PERFCTR_CP(0), rbmemptr_stats(ring, index, cpcycles_start)); @@ -283,7 +293,7 @@ static void a7xx_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit) OUT_PKT7(ring, CP_THREAD_CONTROL, 1); OUT_RING(ring, CP_THREAD_CONTROL_0_SYNC_THREADS | CP_SET_THREAD_BR); - a6xx_set_pagetable(a6xx_gpu, ring, submit->queue->ctx); + a6xx_set_pagetable(a6xx_gpu, ring, submit); get_stats_counter(ring, REG_A7XX_RBBM_PERFCTR_CP(0), rbmemptr_stats(ring, index, cpcycles_start)); -- cgit v1.2.3