diff options
author | Lyude Paul <lyude@redhat.com> | 2019-10-07 14:20:12 -0400 |
---|---|---|
committer | Lyude Paul <lyude@redhat.com> | 2020-07-16 18:16:33 -0400 |
commit | 12885ecbfe62df4358d452080d3b8feef54ec8cb (patch) | |
tree | 922b733a954afad1524029566c9ce97dc9282d86 /drivers | |
parent | 0bc8ffe09771c182e652cb388293049b88797772 (diff) |
drm/nouveau/kms/nvd9-: Add CRC support
This introduces support for CRC readback on gf119+, using the
documentation generously provided to us by Nvidia:
https://github.com/NVIDIA/open-gpu-doc/blob/master/Display-CRC/display-crc.txt
We expose all available CRC sources. SF, SOR, PIOR, and DAC are exposed
through a single set of "outp" sources: outp-active/auto for a CRC of
the scanout region, outp-complete for a CRC of both the scanout and
blanking/sync region combined, and outp-inactive for a CRC of only the
blanking/sync region. For each source, nouveau selects the appropriate
tap point based on the output path in use. We also expose an "rg"
source, which allows for capturing CRCs of the scanout raster before
it's encoded into a video signal in the output path. This tap point is
referred to as the raster generator.
Note that while there's some other neat features that can be used with
CRC capture on nvidia hardware, like capturing from two CRC sources
simultaneously, I couldn't see any usecase for them and did not
implement them.
Nvidia only allows for accessing CRCs through a shared DMA region that
we program through the core EVO/NvDisplay channel which is referred to
as the notifier context. The notifier context is limited to either 255
(for Fermi-Pascal) or 2047 (Volta+) entries to store CRCs in, and
unfortunately the hardware simply drops CRCs and reports an overflow
once all available entries in the notifier context are filled.
Since the DRM CRC API and igt-gpu-tools don't expect there to be a limit
on how many CRCs can be captured, we work around this in nouveau by
allocating two separate notifier contexts for each head instead of one.
We schedule a vblank worker ahead of time so that once we start getting
close to filling up all of the available entries in the notifier
context, we can swap the currently used notifier context out with
another pre-prepared notifier context in a manner similar to page
flipping.
Unfortunately, the hardware only allows us to this by flushing two
separate updates on the core channel: one to release the current
notifier context handle, and one to program the next notifier context's
handle. When the hardware processes the first update, the CRC for the
current frame is lost. However, the second update can be flushed
immediately without waiting for the first to complete so that CRC
generation resumes on the next frame. According to Nvidia's hardware
engineers, there isn't any cleaner way of flipping notifier contexts
that would avoid this.
Since using vblank workers to swap out the notifier context will ensure
we can usually flush both updates to hardware within the timespan of a
single frame, we can also ensure that there will only be exactly one
frame lost between the first and second update being executed by the
hardware. This gives us the guarantee that we're always correctly
matching each CRC entry with it's respective frame even after a context
flip. And since IGT will retrieve the CRC entry for a frame by waiting
until it receives a CRC for any subsequent frames, this doesn't cause an
issue with any tests and is much simpler than trying to change the
current DRM API to accommodate.
In order to facilitate testing of correct handling of this limitation,
we also expose a debugfs interface to manually control the threshold for
when we start trying to flip the notifier context. We will use this in
igt to trigger a context flip for testing purposes without needing to
wait for the notifier to completely fill up. This threshold is reset
to the default value set by nouveau after each capture, and is exposed
in a separate folder within each CRTC's debugfs directory labelled
"nv_crc".
Changes since v1:
* Forgot to finish saving crc.h before saving, whoops. This just adds
some corrections to the empty function declarations that we use if
CONFIG_DEBUG_FS isn't enabled.
Changes since v2:
* Don't check return code from debugfs_create_dir() or
debugfs_create_file() - Greg K-H
Changes since v3:
(no functional changes)
* Fix SPDX license identifiers (checkpatch)
* s/uint32_t/u32/ (checkpatch)
* Fix indenting in switch cases (checkpatch)
Changes since v4:
* Remove unneeded param changes with nv50_head_flush_clr/set
* Rebase
Changes since v5:
* Remove set but unused variable (outp) in nv50_crc_atomic_check() -
Kbuild bot
Signed-off-by: Lyude Paul <lyude@redhat.com>
Reviewed-by: Ben Skeggs <bskeggs@redhat.com>
Acked-by: Dave Airlie <airlied@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20200627194657.156514-10-lyude@redhat.com
Diffstat (limited to 'drivers')
21 files changed, 1325 insertions, 70 deletions
diff --git a/drivers/gpu/drm/nouveau/dispnv04/crtc.c b/drivers/gpu/drm/nouveau/dispnv04/crtc.c index 086f7cf4c23e..4989627b7802 100644 --- a/drivers/gpu/drm/nouveau/dispnv04/crtc.c +++ b/drivers/gpu/drm/nouveau/dispnv04/crtc.c @@ -44,6 +44,9 @@ #include <subdev/bios/pll.h> #include <subdev/clk.h> +#include <nvif/event.h> +#include <nvif/cl0046.h> + static int nv04_crtc_mode_set_base(struct drm_crtc *crtc, int x, int y, struct drm_framebuffer *old_fb); @@ -756,6 +759,7 @@ static void nv_crtc_destroy(struct drm_crtc *crtc) nouveau_bo_unmap(nv_crtc->cursor.nvbo); nouveau_bo_unpin(nv_crtc->cursor.nvbo); nouveau_bo_ref(NULL, &nv_crtc->cursor.nvbo); + nvif_notify_fini(&nv_crtc->vblank); kfree(nv_crtc); } @@ -1297,9 +1301,19 @@ create_primary_plane(struct drm_device *dev) return primary; } +static int nv04_crtc_vblank_handler(struct nvif_notify *notify) +{ + struct nouveau_crtc *nv_crtc = + container_of(notify, struct nouveau_crtc, vblank); + + drm_crtc_handle_vblank(&nv_crtc->base); + return NVIF_NOTIFY_KEEP; +} + int nv04_crtc_create(struct drm_device *dev, int crtc_num) { + struct nouveau_display *disp = nouveau_display(dev); struct nouveau_crtc *nv_crtc; int ret; @@ -1337,5 +1351,14 @@ nv04_crtc_create(struct drm_device *dev, int crtc_num) nv04_cursor_init(nv_crtc); - return 0; + ret = nvif_notify_init(&disp->disp.object, nv04_crtc_vblank_handler, + false, NV04_DISP_NTFY_VBLANK, + &(struct nvif_notify_head_req_v0) { + .head = nv_crtc->index, + }, + sizeof(struct nvif_notify_head_req_v0), + sizeof(struct nvif_notify_head_rep_v0), + &nv_crtc->vblank); + + return ret; } diff --git a/drivers/gpu/drm/nouveau/dispnv50/Kbuild b/drivers/gpu/drm/nouveau/dispnv50/Kbuild index e0c435eae664..6fdddb266fb1 100644 --- a/drivers/gpu/drm/nouveau/dispnv50/Kbuild +++ b/drivers/gpu/drm/nouveau/dispnv50/Kbuild @@ -10,6 +10,10 @@ nouveau-y += dispnv50/core917d.o nouveau-y += dispnv50/corec37d.o nouveau-y += dispnv50/corec57d.o +nouveau-$(CONFIG_DEBUG_FS) += dispnv50/crc.o +nouveau-$(CONFIG_DEBUG_FS) += dispnv50/crc907d.o +nouveau-$(CONFIG_DEBUG_FS) += dispnv50/crcc37d.o + nouveau-y += dispnv50/dac507d.o nouveau-y += dispnv50/dac907d.o diff --git a/drivers/gpu/drm/nouveau/dispnv50/atom.h b/drivers/gpu/drm/nouveau/dispnv50/atom.h index 62faaf60f47a..3d82b3c67dec 100644 --- a/drivers/gpu/drm/nouveau/dispnv50/atom.h +++ b/drivers/gpu/drm/nouveau/dispnv50/atom.h @@ -2,6 +2,9 @@ #define __NV50_KMS_ATOM_H__ #define nv50_atom(p) container_of((p), struct nv50_atom, state) #include <drm/drm_atomic.h> +#include "crc.h" + +struct nouveau_encoder; struct nv50_atom { struct drm_atomic_state state; @@ -115,9 +118,12 @@ struct nv50_head_atom { u8 nhsync:1; u8 nvsync:1; u8 depth:4; + u8 crc_raster:2; u8 bpc; } or; + struct nv50_crc_atom crc; + /* Currently only used for MST */ struct { int pbn; @@ -135,6 +141,7 @@ struct nv50_head_atom { bool ovly:1; bool dither:1; bool procamp:1; + bool crc:1; bool or:1; }; u16 mask; @@ -150,6 +157,19 @@ nv50_head_atom_get(struct drm_atomic_state *state, struct drm_crtc *crtc) return nv50_head_atom(statec); } +static inline struct drm_encoder * +nv50_head_atom_get_encoder(struct nv50_head_atom *atom) +{ + struct drm_encoder *encoder = NULL; + + /* We only ever have a single encoder */ + drm_for_each_encoder_mask(encoder, atom->state.crtc->dev, + atom->state.encoder_mask) + break; + + return encoder; +} + #define nv50_wndw_atom(p) container_of((p), struct nv50_wndw_atom, state) struct nv50_wndw_atom { diff --git a/drivers/gpu/drm/nouveau/dispnv50/core.h b/drivers/gpu/drm/nouveau/dispnv50/core.h index 99157dc94d23..e021cb340569 100644 --- a/drivers/gpu/drm/nouveau/dispnv50/core.h +++ b/drivers/gpu/drm/nouveau/dispnv50/core.h @@ -2,6 +2,7 @@ #define __NV50_KMS_CORE_H__ #include "disp.h" #include "atom.h" +#include "crc.h" #include <nouveau_encoder.h> struct nv50_core { @@ -26,6 +27,9 @@ struct nv50_core_func { } wndw; const struct nv50_head_func *head; +#if IS_ENABLED(CONFIG_DEBUG_FS) + const struct nv50_crc_func *crc; +#endif const struct nv50_outp_func { void (*ctrl)(struct nv50_core *, int or, u32 ctrl, struct nv50_head_atom *); diff --git a/drivers/gpu/drm/nouveau/dispnv50/core907d.c b/drivers/gpu/drm/nouveau/dispnv50/core907d.c index 271629832629..b17c03529c78 100644 --- a/drivers/gpu/drm/nouveau/dispnv50/core907d.c +++ b/drivers/gpu/drm/nouveau/dispnv50/core907d.c @@ -30,6 +30,9 @@ core907d = { .ntfy_wait_done = core507d_ntfy_wait_done, .update = core507d_update, .head = &head907d, +#if IS_ENABLED(CONFIG_DEBUG_FS) + .crc = &crc907d, +#endif .dac = &dac907d, .sor = &sor907d, }; diff --git a/drivers/gpu/drm/nouveau/dispnv50/core917d.c b/drivers/gpu/drm/nouveau/dispnv50/core917d.c index 5cc072d4c30f..66846f372080 100644 --- a/drivers/gpu/drm/nouveau/dispnv50/core917d.c +++ b/drivers/gpu/drm/nouveau/dispnv50/core917d.c @@ -30,6 +30,9 @@ core917d = { .ntfy_wait_done = core507d_ntfy_wait_done, .update = core507d_update, .head = &head917d, +#if IS_ENABLED(CONFIG_DEBUG_FS) + .crc = &crc907d, +#endif .dac = &dac907d, .sor = &sor907d, }; diff --git a/drivers/gpu/drm/nouveau/dispnv50/corec37d.c b/drivers/gpu/drm/nouveau/dispnv50/corec37d.c index e0c8811fb8e4..ec83189a1d48 100644 --- a/drivers/gpu/drm/nouveau/dispnv50/corec37d.c +++ b/drivers/gpu/drm/nouveau/dispnv50/corec37d.c @@ -142,6 +142,9 @@ corec37d = { .wndw.owner = corec37d_wndw_owner, .head = &headc37d, .sor = &sorc37d, +#if IS_ENABLED(CONFIG_DEBUG_FS) + .crc = &crcc37d, +#endif }; int diff --git a/drivers/gpu/drm/nouveau/dispnv50/corec57d.c b/drivers/gpu/drm/nouveau/dispnv50/corec57d.c index 10ba9e9e4ae6..e1c11eba0ce1 100644 --- a/drivers/gpu/drm/nouveau/dispnv50/corec57d.c +++ b/drivers/gpu/drm/nouveau/dispnv50/corec57d.c @@ -52,6 +52,9 @@ corec57d = { .wndw.owner = corec37d_wndw_owner, .head = &headc57d, .sor = &sorc37d, +#if IS_ENABLED(CONFIG_DEBUG_FS) + .crc = &crcc37d, +#endif }; int diff --git a/drivers/gpu/drm/nouveau/dispnv50/crc.c b/drivers/gpu/drm/nouveau/dispnv50/crc.c new file mode 100644 index 000000000000..0b18d9e3a2b9 --- /dev/null +++ b/drivers/gpu/drm/nouveau/dispnv50/crc.c @@ -0,0 +1,714 @@ +// SPDX-License-Identifier: MIT +#include <linux/string.h> +#include <drm/drm_crtc.h> +#include <drm/drm_atomic_helper.h> +#include <drm/drm_vblank.h> +#include <drm/drm_vblank_work.h> + +#include <nvif/class.h> +#include <nvif/cl0002.h> +#include <nvif/timer.h> + +#include "nouveau_drv.h" +#include "core.h" +#include "head.h" +#include "wndw.h" +#include "handles.h" +#include "crc.h" + +static const char * const nv50_crc_sources[] = { + [NV50_CRC_SOURCE_NONE] = "none", + [NV50_CRC_SOURCE_AUTO] = "auto", + [NV50_CRC_SOURCE_RG] = "rg", + [NV50_CRC_SOURCE_OUTP_ACTIVE] = "outp-active", + [NV50_CRC_SOURCE_OUTP_COMPLETE] = "outp-complete", + [NV50_CRC_SOURCE_OUTP_INACTIVE] = "outp-inactive", +}; + +static int nv50_crc_parse_source(const char *buf, enum nv50_crc_source *s) +{ + int i; + + if (!buf) { + *s = NV50_CRC_SOURCE_NONE; + return 0; + } + + i = match_string(nv50_crc_sources, ARRAY_SIZE(nv50_crc_sources), buf); + if (i < 0) + return i; + + *s = i; + return 0; +} + +int +nv50_crc_verify_source(struct drm_crtc *crtc, const char *source_name, + size_t *values_cnt) +{ + struct nouveau_drm *drm = nouveau_drm(crtc->dev); + enum nv50_crc_source source; + + if (nv50_crc_parse_source(source_name, &source) < 0) { + NV_DEBUG(drm, "unknown source %s\n", source_name); + return -EINVAL; + } + + *values_cnt = 1; + return 0; +} + +const char *const *nv50_crc_get_sources(struct drm_crtc *crtc, size_t *count) +{ + *count = ARRAY_SIZE(nv50_crc_sources); + return nv50_crc_sources; +} + +static void +nv50_crc_program_ctx(struct nv50_head *head, + struct nv50_crc_notifier_ctx *ctx) +{ + struct nv50_disp *disp = nv50_disp(head->base.base.dev); + struct nv50_core *core = disp->core; + u32 interlock[NV50_DISP_INTERLOCK__SIZE] = { 0 }; + + core->func->crc->set_ctx(head, ctx); + core->func->update(core, interlock, false); +} + +static void nv50_crc_ctx_flip_work(struct kthread_work *base) +{ + struct drm_vblank_work *work = to_drm_vblank_work(base); + struct nv50_crc *crc = container_of(work, struct nv50_crc, flip_work); + struct nv50_head *head = container_of(crc, struct nv50_head, crc); + struct drm_crtc *crtc = &head->base.base; + struct nv50_disp *disp = nv50_disp(crtc->dev); + u8 new_idx = crc->ctx_idx ^ 1; + + /* + * We don't want to accidentally wait for longer then the vblank, so + * try again for the next vblank if we don't grab the lock + */ + if (!mutex_trylock(&disp->mutex)) { + DRM_DEV_DEBUG_KMS(crtc->dev->dev, + "Lock contended, delaying CRC ctx flip for head-%d\n", + head->base.index); + drm_vblank_work_schedule(work, + drm_crtc_vblank_count(crtc) + 1, + true); + return; + } + + DRM_DEV_DEBUG_KMS(crtc->dev->dev, + "Flipping notifier ctx for head %d (%d -> %d)\n", + drm_crtc_index(crtc), crc->ctx_idx, new_idx); + + nv50_crc_program_ctx(head, NULL); + nv50_crc_program_ctx(head, &crc->ctx[new_idx]); + mutex_unlock(&disp->mutex); + + spin_lock_irq(&crc->lock); + crc->ctx_changed = true; + spin_unlock_irq(&crc->lock); +} + +static inline void nv50_crc_reset_ctx(struct nv50_crc_notifier_ctx *ctx) +{ + memset_io(ctx->mem.object.map.ptr, 0, ctx->mem.object.map.size); +} + +static void +nv50_crc_get_entries(struct nv50_head *head, + const struct nv50_crc_func *func, + enum nv50_crc_source source) +{ + struct drm_crtc *crtc = &head->base.base; + struct nv50_crc *crc = &head->crc; + u32 output_crc; + + while (crc->entry_idx < func->num_entries) { + /* + * While Nvidia's documentation says CRCs are written on each + * subsequent vblank after being enabled, in practice they + * aren't written immediately. + */ + output_crc = func->get_entry(head, &crc->ctx[crc->ctx_idx], + source, crc->entry_idx); + if (!output_crc) + return; + + drm_crtc_add_crc_entry(crtc, true, crc->frame, &output_crc); + crc->frame++; + crc->entry_idx++; + } +} + +void nv50_crc_handle_vblank(struct nv50_head *head) +{ + struct drm_crtc *crtc = &head->base.base; + struct nv50_crc *crc = &head->crc; + const struct nv50_crc_func *func = + nv50_disp(head->base.base.dev)->core->func->crc; + struct nv50_crc_notifier_ctx *ctx; + bool need_reschedule = false; + + if (!func) + return; + + /* + * We don't lose events if we aren't able to report CRCs until the + * next vblank, so only report CRCs if the locks we need aren't + * contended to prevent missing an actual vblank event + */ + if (!spin_trylock(&crc->lock)) + return; + + if (!crc->src) + goto out; + + ctx = &crc->ctx[crc->ctx_idx]; + if (crc->ctx_changed && func->ctx_finished(head, ctx)) { + nv50_crc_get_entries(head, func, crc->src); + + crc->ctx_idx ^= 1; + crc->entry_idx = 0; + crc->ctx_changed = false; + + /* + * Unfortunately when notifier contexts are changed during CRC + * capture, we will inevitably lose the CRC entry for the + * frame where the hardware actually latched onto the first + * UPDATE. According to Nvidia's hardware engineers, there's + * no workaround for this. + * + * Now, we could try to be smart here and calculate the number + * of missed CRCs based on audit timestamps, but those were + * removed starting with volta. Since we always flush our + * updates back-to-back without waiting, we'll just be + * optimistic and assume we always miss exactly one frame. + */ + DRM_DEV_DEBUG_KMS(head->base.base.dev->dev, + "Notifier ctx flip for head-%d finished, lost CRC for frame %llu\n", + head->base.index, crc->frame); + crc->frame++; + + nv50_crc_reset_ctx(ctx); + need_reschedule = true; + } + + nv50_crc_get_entries(head, func, crc->src); + + if (need_reschedule) + drm_vblank_work_schedule(&crc->flip_work, + drm_crtc_vblank_count(crtc) + + crc->flip_threshold + - crc->entry_idx, + true); + +out: + spin_unlock(&crc->lock); +} + +static void nv50_crc_wait_ctx_finished(struct nv50_head *head, + const struct nv50_crc_func *func, + struct nv50_crc_notifier_ctx *ctx) +{ + struct drm_device *dev = head->base.base.dev; + struct nouveau_drm *drm = nouveau_drm(dev); + s64 ret; + + ret = nvif_msec(&drm->client.device, 50, + if (func->ctx_finished(head, ctx)) break;); + if (ret == -ETIMEDOUT) + NV_ERROR(drm, + "CRC notifier ctx for head %d not finished after 50ms\n", + head->base.index); + else if (ret) + NV_ATOMIC(drm, + "CRC notifier ctx for head-%d finished after %lldns\n", + head->base.index, ret); +} + +void nv50_crc_atomic_stop_reporting(struct drm_atomic_state *state) +{ + struct drm_crtc_state *crtc_state; + struct drm_crtc *crtc; + int i; + + for_each_new_crtc_in_state(state, crtc, crtc_state, i) { + struct nv50_head *head = nv50_head(crtc); + struct nv50_head_atom *asyh = nv50_head_atom(crtc_state); + struct nv50_crc *crc = &head->crc; + + if (!asyh->clr.crc) + continue; + + spin_lock_irq(&crc->lock); + crc->src = NV50_CRC_SOURCE_NONE; + spin_unlock_irq(&crc->lock); + + drm_crtc_vblank_put(crtc); + drm_vblank_work_cancel_sync(&crc->flip_work); + + NV_ATOMIC(nouveau_drm(crtc->dev), + "CRC reporting on vblank for head-%d disabled\n", + head->base.index); + + /* CRC generation is still enabled in hw, we'll just report + * any remaining CRC entries ourselves after it gets disabled + * in hardware + */ + } +} + +void nv50_crc_atomic_prepare_notifier_contexts(struct drm_atomic_state *state) +{ + const struct nv50_crc_func *func = + nv50_disp(state->dev)->core->func->crc; + struct drm_crtc_state *new_crtc_state; + struct drm_crtc *crtc; + int i; + + for_each_new_crtc_in_state(state, crtc, new_crtc_state, i) { + struct nv50_head *head = nv50_head(crtc); + struct nv50_head_atom *asyh = nv50_head_atom(new_crtc_state); + struct nv50_crc *crc = &head->crc; + struct nv50_crc_notifier_ctx *ctx = &crc->ctx[crc->ctx_idx]; + int i; + + if (asyh->clr.crc && asyh->crc.src) { + if (crc->ctx_changed) { + nv50_crc_wait_ctx_finished(head, func, ctx); + ctx = &crc->ctx[crc->ctx_idx ^ 1]; + } + nv50_crc_wait_ctx_finished(head, func, ctx); + } + + if (asyh->set.crc) { + crc->entry_idx = 0; + crc->ctx_changed = false; + for (i = 0; i < ARRAY_SIZE(crc->ctx); i++) + nv50_crc_reset_ctx(&crc->ctx[i]); + } + } +} + +void nv50_crc_atomic_start_reporting(struct drm_atomic_state *state) +{ + struct drm_crtc_state *crtc_state; + struct drm_crtc *crtc; + int i; + + for_each_new_crtc_in_state(state, crtc, crtc_state, i) { + struct nv50_head *head = nv50_head(crtc); + struct nv50_head_atom *asyh = nv50_head_atom(crtc_state); + struct nv50_crc *crc = &head->crc; + u64 vbl_count; + + if (!asyh->set.crc) + continue; + + drm_crtc_vblank_get(crtc); + + spin_lock_irq(&crc->lock); + vbl_count = drm_crtc_vblank_count(crtc); + crc->frame = vbl_count; + crc->src = asyh->crc.src; + drm_vblank_work_schedule(&crc->flip_work, + vbl_count + crc->flip_threshold, + true); + spin_unlock_irq(&crc->lock); + + NV_ATOMIC(nouveau_drm(crtc->dev), + "CRC reporting on vblank for head-%d enabled\n", + head->base.index); + } +} + +int nv50_crc_atomic_check(struct nv50_head *head, + struct nv50_head_atom *asyh, + struct nv50_head_atom *armh) +{ + struct drm_atomic_state *state = asyh->state.state; + struct drm_device *dev = head->base.base.dev; + struct nv50_atom *atom = nv50_atom(state); + struct nv50_disp *disp = nv50_disp(dev); + struct drm_encoder *encoder; + struct nv50_outp_atom *outp_atom; + bool changed = armh->crc.src != asyh->crc.src; + + if (!armh->crc.src && !asyh->crc.src) { + asyh->set.crc = false; + asyh->clr.crc = false; + return 0; + } + + /* While we don't care about entry tags, Volta+ hw always needs the + * controlling wndw channel programmed to a wndw that's owned by our + * head + */ + if (asyh->crc.src && disp->disp->object.oclass >= GV100_DISP && + !(BIT(asyh->crc.wndw) & asyh->wndw.owned)) { + if (!asyh->wndw.owned) { + /* TODO: once we support flexible channel ownership, + * we should write some code here to handle attempting + * to "steal" a plane: e.g. take a plane that is + * currently not-visible and owned by another head, + * and reassign it to this head. If we fail to do so, + * we shuld reject the mode outright as CRC capture + * then becomes impossible. + */ + NV_ATOMIC(nouveau_drm(dev), + "No available wndws for CRC readback\n"); + return -EINVAL; + } + asyh->crc.wndw = ffs(asyh->wndw.owned) - 1; + } + + if (drm_atomic_crtc_needs_modeset(&asyh->state) || changed || + armh->crc.wndw != asyh->crc.wndw) { + asyh->clr.crc = armh->crc.src && armh->state.active; + asyh->set.crc = asyh->crc.src && asyh->state.active; + if (changed) + asyh->set.or |= armh->or.crc_raster != + asyh->or.crc_raster; + + /* + * If we're reprogramming our OR, we need to flush the CRC + * disable first + */ + if (asyh->clr.crc) { + encoder = nv50_head_atom_get_encoder(armh); + + list_for_each_entry(outp_atom, &atom->outp, head) { + if (outp_atom->encoder == encoder) { + if (outp_atom->set.mask) + atom->flush_disable = true; + break; + } + } + } + } else { + asyh->set.crc = false; + asyh->clr.crc = false; + } + + return 0; +} + +static enum nv50_crc_source_type +nv50_crc_source_type(struct nouveau_encoder *outp, + enum nv50_crc_source source) +{ + struct dcb_output *dcbe = outp->dcb; + + switch (source) { + case NV50_CRC_SOURCE_NONE: return NV50_CRC_SOURCE_TYPE_NONE; + case NV50_CRC_SOURCE_RG: return NV50_CRC_SOURCE_TYPE_RG; + default: break; + } + + if (dcbe->location != DCB_LOC_ON_CHIP) + return NV50_CRC_SOURCE_TYPE_PIOR; + + switch (dcbe->type) { + case DCB_OUTPUT_DP: return NV50_CRC_SOURCE_TYPE_SF; + case DCB_OUTPUT_ANALOG: return NV50_CRC_SOURCE_TYPE_DAC; + default: return NV50_CRC_SOURCE_TYPE_SOR; + } +} + +void nv50_crc_atomic_set(struct nv50_head *head, + struct nv50_head_atom *asyh) +{ + struct drm_crtc *crtc = &head->base.base; + struct drm_device *dev = crtc->dev; + struct nv50_crc *crc = &head->crc; + const struct nv50_crc_func *func = nv50_disp(dev)->core->func->crc; + struct nouveau_encoder *outp = + nv50_real_outp(nv50_head_atom_get_encoder(asyh)); + + func->set_src(head, outp->or, + nv50_crc_source_type(outp, asyh->crc.src), + &crc->ctx[crc->ctx_idx], asyh->crc.wndw); +} + +void nv50_crc_atomic_clr(struct nv50_head *head) +{ + const struct nv50_crc_func *func = + nv50_disp(head->base.base.dev)->core->func->crc; + + func->set_src(head, 0, NV50_CRC_SOURCE_TYPE_NONE, NULL, 0); +} + +#define NV50_CRC_RASTER_ACTIVE 0 +#define NV50_CRC_RASTER_COMPLETE 1 +#define NV50_CRC_RASTER_INACTIVE 2 + +static inline int +nv50_crc_raster_type(enum nv50_crc_source source) +{ + switch (source) { + case NV50_CRC_SOURCE_NONE: + case NV50_CRC_SOURCE_AUTO: + case NV50_CRC_SOURCE_RG: + case NV50_CRC_SOURCE_OUTP_ACTIVE: + return NV50_CRC_RASTER_ACTIVE; + case NV50_CRC_SOURCE_OUTP_COMPLETE: + return NV50_CRC_RASTER_COMPLETE; + case NV50_CRC_SOURCE_OUTP_INACTIVE: + return NV50_CRC_RASTER_INACTIVE; + } + + return 0; +} + +/* We handle mapping the memory for CRC notifiers ourselves, since each + * notifier needs it's own handle + */ +static inline int +nv50_crc_ctx_init(struct nv50_head *head, struct nvif_mmu *mmu, + struct nv50_crc_notifier_ctx *ctx, size_t len, int idx) +{ + struct nv50_core *core = nv50_disp(head->base.base.dev)->core; + int ret; + + ret = nvif_mem_init_map(mmu, NVIF_MEM_VRAM, len, &ctx->mem); + if (ret) + return ret; + + ret = nvif_object_init(&core->chan.base.user, + NV50_DISP_HANDLE_CRC_CTX(head, idx), + NV_DMA_IN_MEMORY, + &(struct nv_dma_v0) { + .target = NV_DMA_V0_TARGET_VRAM, + .access = NV_DMA_V0_ACCESS_RDWR, + .start = ctx->mem.addr, + .limit = ctx->mem.addr + + ctx->mem.size - 1, + }, sizeof(struct nv_dma_v0), + &ctx->ntfy); + if (ret) + goto fail_fini; + + return 0; + +fail_fini: + nvif_mem_fini(&ctx->mem); + return ret; +} + +static inline void +nv50_crc_ctx_fini(struct nv50_crc_notifier_ctx *ctx) +{ + nvif_object_fini(&ctx->ntfy); + nvif_mem_fini(&ctx->mem); +} + +int nv50_crc_set_source(struct drm_crtc *crtc, const char *source_str) +{ + struct drm_device *dev = crtc->dev; + struct drm_atomic_state *state; + struct drm_modeset_acquire_ctx ctx; + struct nv50_head *head = nv50_head(crtc); + struct nv50_crc *crc = &head->crc; + const struct nv50_crc_func *func = nv50_disp(dev)->core->func->crc; + struct nvif_mmu *mmu = &nouveau_drm(dev)->client.mmu; + struct nv50_head_atom *asyh; + struct drm_crtc_state *crtc_state; + enum nv50_crc_source source; + int ret = 0, ctx_flags = 0, i; + + ret = nv50_crc_parse_source(source_str, &source); + if (ret) + return ret; + + /* + * Since we don't want the user to accidentally interrupt us as we're + * disabling CRCs + */ + if (source) + ctx_flags |= DRM_MODESET_ACQUIRE_INTERRUPTIBLE; + drm_modeset_acquire_init(&ctx, ctx_flags); + + state = drm_atomic_state_alloc(dev); + if (!state) { + ret = -ENOMEM; + goto out_acquire_fini; + } + state->acquire_ctx = &ctx; + + if (source) { + for (i = 0; i < ARRAY_SIZE(head->crc.ctx); i++) { + ret = nv50_crc_ctx_init(head, mmu, &crc->ctx[i], + func->notifier_len, i); + if (ret) + goto out_ctx_fini; + } + } + +retry: + crtc_state = drm_atomic_get_crtc_state(state, &head->base.base); + if (IS_ERR(crtc_state)) { + ret = PTR_ERR(crtc_state); + if (ret == -EDEADLK) + goto deadlock; + else if (ret) + goto out_drop_locks; + } + asyh = nv50_head_atom(crtc_state); + asyh->crc.src = source; + asyh->or.crc_raster = nv50_crc_raster_type(source); + + ret = drm_atomic_commit(state); + if (ret == -EDEADLK) + goto deadlock; + else if (ret) + goto out_drop_locks; + + if (!source) { + /* + * If the user specified a custom flip threshold through + * debugfs, reset it + */ + crc->flip_threshold = func->flip_threshold; + } + +out_drop_locks: + drm_modeset_drop_locks(&ctx); +out_ctx_fini: + if (!source || ret) { + for (i = 0; i < ARRAY_SIZE(crc->ctx); i++) + nv50_crc_ctx_fini(&crc->ctx[i]); + } + drm_atomic_state_put(state); +out_acquire_fini: + drm_modeset_acquire_fini(&ctx); + return ret; + +deadlock: + drm_atomic_state_clear(state); + drm_modeset_backoff(&ctx); + goto retry; +} + +static int +nv50_crc_debugfs_flip_threshold_get(struct seq_file *m, void *data) +{ + struct nv50_head *head = m->private; + struct drm_crtc *crtc = &head->base.base; + struct nv50_crc *crc = &head->crc; + int ret; + + ret = drm_modeset_lock_single_interruptible(&crtc->mutex); + if (ret) + return ret; + + seq_printf(m, "%d\n", crc->flip_threshold); + + drm_modeset_unlock(&crtc->mutex); + return ret; +} + +static int +nv50_crc_debugfs_flip_threshold_open(struct inode *inode, struct file *file) +{ + return single_open(file, nv50_crc_debugfs_flip_threshold_get, + inode->i_private); +} + +static ssize_t +nv50_crc_debugfs_flip_threshold_set(struct file *file, + const char __user *ubuf, size_t len, + loff_t *offp) +{ + struct seq_file *m = file->private_data; + struct nv50_head *head = m->private; + struct nv50_head_atom *armh; + struct drm_crtc *crtc = &head->base.base; + struct nouveau_drm *drm = nouveau_drm(crtc->dev); + struct nv50_crc *crc = &head->crc; + const struct nv50_crc_func *func = + nv50_disp(crtc->dev)->core->func->crc; + int value, ret; + + ret = kstrtoint_from_user(ubuf, len, 10, &value); + if (ret) + return ret; + + if (value > func->flip_threshold) + return -EINVAL; + else if (value == -1) + value = func->flip_threshold; + else if (value < -1) + return -EINVAL; + + ret = drm_modeset_lock_single_interruptible(&crtc->mutex); + if (ret) + return ret; + + armh = nv50_head_atom(crtc->state); + if (armh->crc.src) { + ret = -EBUSY; + goto out; + } + + NV_DEBUG(drm, + "Changing CRC flip threshold for next capture on head-%d to %d\n", + head->base.index, value); + crc->flip_threshold = value; + ret = len; |