dri-devel.lists.freedesktop.org archive mirror
 help / color / mirror / Atom feed
From: Mikko Perttunen <mperttunen@nvidia.com>
To: thierry.reding@gmail.com, jonathanh@nvidia.com, digetx@gmail.com,
	airlied@linux.ie, daniel@ffwll.ch
Cc: linux-tegra@vger.kernel.org, talho@nvidia.com,
	bhuntsman@nvidia.com, dri-devel@lists.freedesktop.org,
	Mikko Perttunen <mperttunen@nvidia.com>
Subject: [PATCH v3 12/20] gpu: host1x: Add support for syncpoint waits in CDMA pushbuffer
Date: Wed,  7 Oct 2020 20:12:30 +0300	[thread overview]
Message-ID: <20201007171238.1795964-13-mperttunen@nvidia.com> (raw)
In-Reply-To: <20201007171238.1795964-1-mperttunen@nvidia.com>

Add support for inserting syncpoint waits in the CDMA pushbuffer.
These waits need to be done in HOST1X class, while gather submitted
by the application execute in engine class.

Support is added by converting the gather list of job into a command
list that can include both gathers and waits. When the job is
submitted, these commands are pushed as the appropriate opcodes
on the CDMA pushbuffer.

Signed-off-by: Mikko Perttunen <mperttunen@nvidia.com>
---
 drivers/gpu/host1x/hw/channel_hw.c | 51 +++++++++++++++--------
 drivers/gpu/host1x/hw/debug_hw.c   |  9 +++-
 drivers/gpu/host1x/job.c           | 67 +++++++++++++++++++++---------
 drivers/gpu/host1x/job.h           | 14 +++++++
 include/linux/host1x.h             |  5 ++-
 5 files changed, 105 insertions(+), 41 deletions(-)

diff --git a/drivers/gpu/host1x/hw/channel_hw.c b/drivers/gpu/host1x/hw/channel_hw.c
index bf21512e5078..d88a32f73f5e 100644
--- a/drivers/gpu/host1x/hw/channel_hw.c
+++ b/drivers/gpu/host1x/hw/channel_hw.c
@@ -55,31 +55,46 @@ static void submit_gathers(struct host1x_job *job)
 #endif
 	unsigned int i;
 
-	for (i = 0; i < job->num_gathers; i++) {
-		struct host1x_job_gather *g = &job->gathers[i];
-		dma_addr_t addr = g->base + g->offset;
-		u32 op2, op3;
+	for (i = 0; i < job->num_cmds; i++) {
+		struct host1x_job_cmd *cmd = &job->cmds[i];
 
-		op2 = lower_32_bits(addr);
-		op3 = upper_32_bits(addr);
+		if (cmd->is_wait) {
+			/* TODO use modern wait */
+			host1x_cdma_push(cdma,
+				 host1x_opcode_setclass(HOST1X_CLASS_HOST1X,
+					host1x_uclass_wait_syncpt_r(), 1),
+				 host1x_class_host_wait_syncpt(cmd->wait.id,
+					cmd->wait.threshold));
+			host1x_cdma_push(
+				cdma, host1x_opcode_setclass(job->class, 0, 0),
+				HOST1X_OPCODE_NOP);
+		} else {
+			struct host1x_job_gather *g = &cmd->gather;
 
-		trace_write_gather(cdma, g->bo, g->offset, g->words);
+			dma_addr_t addr = g->base + g->offset;
+			u32 op2, op3;
 
-		if (op3 != 0) {
+			op2 = lower_32_bits(addr);
+			op3 = upper_32_bits(addr);
+
+			trace_write_gather(cdma, g->bo, g->offset, g->words);
+
+			if (op3 != 0) {
 #if HOST1X_HW >= 6
-			u32 op1 = host1x_opcode_gather_wide(g->words);
-			u32 op4 = HOST1X_OPCODE_NOP;
+				u32 op1 = host1x_opcode_gather_wide(g->words);
+				u32 op4 = HOST1X_OPCODE_NOP;
 
-			host1x_cdma_push_wide(cdma, op1, op2, op3, op4);
+				host1x_cdma_push_wide(cdma, op1, op2, op3, op4);
 #else
-			dev_err(dev, "invalid gather for push buffer %pad\n",
-				&addr);
-			continue;
+				dev_err(dev, "invalid gather for push buffer %pad\n",
+					&addr);
+				continue;
 #endif
-		} else {
-			u32 op1 = host1x_opcode_gather(g->words);
+			} else {
+				u32 op1 = host1x_opcode_gather(g->words);
 
-			host1x_cdma_push(cdma, op1, op2);
+				host1x_cdma_push(cdma, op1, op2);
+			}
 		}
 	}
 }
@@ -126,7 +141,7 @@ static int channel_submit(struct host1x_job *job)
 	struct host1x *host = dev_get_drvdata(ch->dev->parent);
 
 	trace_host1x_channel_submit(dev_name(ch->dev),
-				    job->num_gathers, job->num_relocs,
+				    job->num_cmds, job->num_relocs,
 				    job->syncpt->id, job->syncpt_incrs);
 
 	/* before error checks, return current max */
diff --git a/drivers/gpu/host1x/hw/debug_hw.c b/drivers/gpu/host1x/hw/debug_hw.c
index ceb48229d14b..35952fd5597e 100644
--- a/drivers/gpu/host1x/hw/debug_hw.c
+++ b/drivers/gpu/host1x/hw/debug_hw.c
@@ -208,10 +208,15 @@ static void show_channel_gathers(struct output *o, struct host1x_cdma *cdma)
 				    job->first_get, job->timeout,
 				    job->num_slots, job->num_unpins);
 
-		for (i = 0; i < job->num_gathers; i++) {
-			struct host1x_job_gather *g = &job->gathers[i];
+		for (i = 0; i < job->num_cmds; i++) {
+			struct host1x_job_gather *g;
 			u32 *mapped;
 
+			if (job->cmds[i].is_wait)
+				continue;
+
+			g = &job->cmds[i].gather;
+
 			if (job->gather_copy_mapped)
 				mapped = (u32 *)job->gather_copy_mapped;
 			else
diff --git a/drivers/gpu/host1x/job.c b/drivers/gpu/host1x/job.c
index acf322beb56c..ac4091596811 100644
--- a/drivers/gpu/host1x/job.c
+++ b/drivers/gpu/host1x/job.c
@@ -38,7 +38,7 @@ struct host1x_job *host1x_job_alloc(struct host1x_channel *ch,
 	total = sizeof(struct host1x_job) +
 		(u64)num_relocs * sizeof(struct host1x_reloc) +
 		(u64)num_unpins * sizeof(struct host1x_job_unpin_data) +
-		(u64)num_cmdbufs * sizeof(struct host1x_job_gather) +
+		(u64)num_cmdbufs * sizeof(struct host1x_job_cmd) +
 		(u64)num_unpins * sizeof(dma_addr_t) +
 		(u64)num_unpins * sizeof(u32 *);
 	if (total > ULONG_MAX)
@@ -57,8 +57,8 @@ struct host1x_job *host1x_job_alloc(struct host1x_channel *ch,
 	mem += num_relocs * sizeof(struct host1x_reloc);
 	job->unpins = num_unpins ? mem : NULL;
 	mem += num_unpins * sizeof(struct host1x_job_unpin_data);
-	job->gathers = num_cmdbufs ? mem : NULL;
-	mem += num_cmdbufs * sizeof(struct host1x_job_gather);
+	job->cmds = num_cmdbufs ? mem : NULL;
+	mem += num_cmdbufs * sizeof(struct host1x_job_cmd);
 	job->addr_phys = num_unpins ? mem : NULL;
 
 	job->reloc_addr_phys = job->addr_phys;
@@ -101,22 +101,35 @@ EXPORT_SYMBOL(host1x_job_put);
 void host1x_job_add_gather(struct host1x_job *job, struct host1x_bo *bo,
 			   unsigned int words, unsigned int offset)
 {
-	struct host1x_job_gather *gather = &job->gathers[job->num_gathers];
+	struct host1x_job_gather *gather = &job->cmds[job->num_cmds].gather;
 
 	gather->words = words;
 	gather->bo = bo;
 	gather->offset = offset;
 
-	job->num_gathers++;
+	job->num_cmds++;
 }
 EXPORT_SYMBOL(host1x_job_add_gather);
 
+void host1x_job_add_wait(struct host1x_job *job, u32 id, u32 thresh)
+{
+	struct host1x_job_cmd *cmd = &job->cmds[job->num_cmds];
+
+	cmd->is_wait = true;
+	cmd->wait.id = id;
+	cmd->wait.threshold = thresh;
+
+	job->num_cmds++;
+}
+EXPORT_SYMBOL(host1x_job_add_wait);
+
 static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 {
 	struct host1x_client *client = job->client;
 	struct device *dev = client->dev;
 	struct host1x_job_gather *g;
 	struct iommu_domain *domain;
+	struct sg_table *sgt;
 	unsigned int i;
 	int err;
 
@@ -126,7 +139,6 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 	for (i = 0; i < job->num_relocs; i++) {
 		struct host1x_reloc *reloc = &job->relocs[i];
 		dma_addr_t phys_addr, *phys;
-		struct sg_table *sgt;
 
 		reloc->target.bo = host1x_bo_get(reloc->target.bo);
 		if (!reloc->target.bo) {
@@ -204,17 +216,20 @@ static unsigned int pin_job(struct host1x *host, struct host1x_job *job)
 	if (IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL))
 		return 0;
 
-	for (i = 0; i < job->num_gathers; i++) {
+	for (i = 0; i < job->num_cmds; i++) {
 		size_t gather_size = 0;
 		struct scatterlist *sg;
-		struct sg_table *sgt;
 		dma_addr_t phys_addr;
 		unsigned long shift;
 		struct iova *alloc;
 		dma_addr_t *phys;
 		unsigned int j;
 
-		g = &job->gathers[i];
+		if (job->cmds[i].is_wait)
+			continue;
+
+		g = &job->cmds[i].gather;
+
 		g->bo = host1x_bo_get(g->bo);
 		if (!g->bo) {
 			err = -EINVAL;
@@ -550,8 +565,13 @@ static inline int copy_gathers(struct device *host, struct host1x_job *job,
 	fw.num_relocs = job->num_relocs;
 	fw.class = job->class;
 
-	for (i = 0; i < job->num_gathers; i++) {
-		struct host1x_job_gather *g = &job->gathers[i];
+	for (i = 0; i < job->num_cmds; i++) {
+		struct host1x_job_gather *g;
+
+		if (job->cmds[i].is_wait)
+			continue;
+
+		g = &job->cmds[i].gather;
 
 		size += g->words * sizeof(u32);
 	}
@@ -573,10 +593,14 @@ static inline int copy_gathers(struct device *host, struct host1x_job *job,
 
 	job->gather_copy_size = size;
 
-	for (i = 0; i < job->num_gathers; i++) {
-		struct host1x_job_gather *g = &job->gathers[i];
+	for (i = 0; i < job->num_cmds; i++) {
+		struct host1x_job_gather *g;
 		void *gather;
 
+		if (job->cmds[i].is_wait)
+			continue;
+		g = &job->cmds[i].gather;
+
 		/* Copy the gather */
 		gather = host1x_bo_mmap(g->bo);
 		memcpy(job->gather_copy_mapped + offset, gather + g->offset,
@@ -619,8 +643,12 @@ int host1x_job_pin(struct host1x_job *job, struct device *dev)
 	}
 
 	/* patch gathers */
-	for (i = 0; i < job->num_gathers; i++) {
-		struct host1x_job_gather *g = &job->gathers[i];
+	for (i = 0; i < job->num_cmds; i++) {
+		struct host1x_job_gather *g;
+
+		if (job->cmds[i].is_wait)
+			continue;
+		g = &job->cmds[i].gather;
 
 		/* process each gather mem only once */
 		if (g->handled)
@@ -630,10 +658,11 @@ int host1x_job_pin(struct host1x_job *job, struct device *dev)
 		if (!IS_ENABLED(CONFIG_TEGRA_HOST1X_FIREWALL))
 			g->base = job->gather_addr_phys[i];
 
-		for (j = i + 1; j < job->num_gathers; j++) {
-			if (job->gathers[j].bo == g->bo) {
-				job->gathers[j].handled = true;
-				job->gathers[j].base = g->base;
+		for (j = i + 1; j < job->num_cmds; j++) {
+			if (!job->cmds[j].is_wait &&
+			    job->cmds[j].gather.bo == g->bo) {
+				job->cmds[j].gather.handled = true;
+				job->cmds[j].gather.base = g->base;
 			}
 		}
 
diff --git a/drivers/gpu/host1x/job.h b/drivers/gpu/host1x/job.h
index 94bc2e4ae241..33adfaede842 100644
--- a/drivers/gpu/host1x/job.h
+++ b/drivers/gpu/host1x/job.h
@@ -18,6 +18,20 @@ struct host1x_job_gather {
 	bool handled;
 };
 
+struct host1x_job_wait {
+	u32 id;
+	u32 threshold;
+};
+
+struct host1x_job_cmd {
+	bool is_wait;
+
+	union {
+		struct host1x_job_gather gather;
+		struct host1x_job_wait wait;
+	};
+};
+
 struct host1x_job_unpin_data {
 	struct host1x_bo *bo;
 	struct sg_table *sgt;
diff --git a/include/linux/host1x.h b/include/linux/host1x.h
index d7070fd65833..59296d3346fe 100644
--- a/include/linux/host1x.h
+++ b/include/linux/host1x.h
@@ -215,8 +215,8 @@ struct host1x_job {
 	struct host1x_client *client;
 
 	/* Gathers and their memory */
-	struct host1x_job_gather *gathers;
-	unsigned int num_gathers;
+	struct host1x_job_cmd *cmds;
+	unsigned int num_cmds;
 
 	/* Array of handles to be pinned & unpinned */
 	struct host1x_reloc *relocs;
@@ -275,6 +275,7 @@ struct host1x_job *host1x_job_alloc(struct host1x_channel *ch,
 				    u32 num_cmdbufs, u32 num_relocs);
 void host1x_job_add_gather(struct host1x_job *job, struct host1x_bo *bo,
 			   unsigned int words, unsigned int offset);
+void host1x_job_add_wait(struct host1x_job *job, u32 id, u32 thresh);
 struct host1x_job *host1x_job_get(struct host1x_job *job);
 void host1x_job_put(struct host1x_job *job);
 int host1x_job_pin(struct host1x_job *job, struct device *dev);
-- 
2.28.0

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

  parent reply	other threads:[~2020-10-07 17:17 UTC|newest]

Thread overview: 38+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-10-07 17:12 [PATCH v3 00/20] Host1x/TegraDRM UAPI Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 01/20] gpu: host1x: Use different lock classes for each client Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 02/20] gpu: host1x: Allow syncpoints without associated client Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 03/20] gpu: host1x: Show number of pending waiters in debugfs Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 04/20] gpu: host1x: Remove cancelled waiters immediately Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 05/20] gpu: host1x: Use HW-equivalent syncpoint expiration check Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 06/20] gpu: host1x: Cleanup and refcounting for syncpoints Mikko Perttunen
2020-10-07 22:23   ` kernel test robot
2020-10-07 17:12 ` [PATCH v3 07/20] gpu: host1x: Introduce UAPI header Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 08/20] gpu: host1x: Implement /dev/host1x device node Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 09/20] gpu: host1x: DMA fences and userspace fence creation Mikko Perttunen
2020-10-07 23:13   ` kernel test robot
2020-10-08 11:13   ` kernel test robot
2020-10-07 17:12 ` [PATCH v3 10/20] gpu: host1x: Add no-recovery mode Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 11/20] gpu: host1x: Add job release callback Mikko Perttunen
2020-10-07 17:12 ` Mikko Perttunen [this message]
2020-10-07 17:12 ` [PATCH v3 13/20] gpu: host1x: Reset max value when freeing a syncpoint Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 14/20] gpu: host1x: Reserve VBLANK syncpoints at initialization Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 15/20] drm/tegra: Add new UAPI to header Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 16/20] drm/tegra: Boot VIC during runtime PM resume Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 17/20] drm/tegra: Set resv fields when importing/exporting GEMs Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 18/20] drm/tegra: Allocate per-engine channel in core code Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 19/20] drm/tegra: Implement new UAPI Mikko Perttunen
2020-10-08  3:42   ` kernel test robot
2020-10-19  2:21   ` Dmitry Osipenko
2020-10-19  8:13     ` Mikko Perttunen
2020-10-19 17:27       ` Dmitry Osipenko
2020-10-20  9:18         ` Mikko Perttunen
2020-10-22  4:20           ` Dmitry Osipenko
2020-10-26  9:11             ` Mikko Perttunen
2020-10-27 19:06               ` Dmitry Osipenko
2020-10-28  9:54                 ` Mikko Perttunen
2020-10-30 23:13                   ` Dmitry Osipenko
2020-11-09 14:53                     ` Mikko Perttunen
2020-11-12 18:35                       ` Dmitry Osipenko
2020-10-20 11:40         ` Daniel Vetter
2020-10-20 12:51           ` Mikko Perttunen
2020-10-07 17:12 ` [PATCH v3 20/20] drm/tegra: Add job firewall Mikko Perttunen

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20201007171238.1795964-13-mperttunen@nvidia.com \
    --to=mperttunen@nvidia.com \
    --cc=airlied@linux.ie \
    --cc=bhuntsman@nvidia.com \
    --cc=daniel@ffwll.ch \
    --cc=digetx@gmail.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=jonathanh@nvidia.com \
    --cc=linux-tegra@vger.kernel.org \
    --cc=talho@nvidia.com \
    --cc=thierry.reding@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).