gpu: nvgpu: remove use of nvgpu_ctxsw_trace_entry

- Remove the usage of nvgpu_ctxsw_trace_entry splattered
across nvgpu, and replace with a struct defined in common code.
The usage is still inside Linux, but this helps the
subsequent unification efforts, e.g. to unify the fecs trace
path.
- Remove "asm/barrier.h" as "nvgpu/barrier.h" is already
included.

EVLR-3078

Change-Id: Iabfb105b891b0078ed326a8047ef14ebe1888cf2
Signed-off-by: Vaibhav Kachore <vkachore@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1803208
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

4 files changed, 46 insertions, 17 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c b/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c
index 78b753d1..50077a62 100644
--- a/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c
@@ -20,7 +20,6 @@
  * DEALINGS IN THE SOFTWARE.
  */
 
-#include <asm/barrier.h>
 #ifdef CONFIG_DEBUG_FS
 #include <linux/debugfs.h>
 #endif
@@ -259,7 +258,7 @@ static pid_t gk20a_fecs_trace_find_pid(struct gk20a *g, u32 context_ptr)
 static int gk20a_fecs_trace_ring_read(struct gk20a *g, int index)
 {
         int i;
-        struct nvgpu_ctxsw_trace_entry entry = { };
+        struct nvgpu_gpu_ctxsw_trace_entry entry = { };
         struct gk20a_fecs_trace *trace = g->fecs_trace;
         pid_t cur_pid;
         pid_t new_pid;
diff --git a/drivers/gpu/nvgpu/include/nvgpu/ctxsw_trace.h b/drivers/gpu/nvgpu/include/nvgpu/ctxsw_trace.h
index cc6edb49..6e54d1cc 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/ctxsw_trace.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/ctxsw_trace.h
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2017, NVIDIA CORPORATION.  All rights reserved.
+ * Copyright (c) 2017-2018, NVIDIA CORPORATION.  All rights reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -28,7 +28,21 @@
 struct gk20a;
 struct tsg_gk20a;
 struct channel_gk20a;
-struct nvgpu_ctxsw_trace_entry;
+
+/*
+ * The binary format of 'struct nvgpu_gpu_ctxsw_trace_entry' introduced here
+ * should match that of 'struct nvgpu_ctxsw_trace_entry' defined in uapi
+ * header, since this struct is intended to be a mirror copy of the uapi
+ * struct.
+ */
+struct nvgpu_gpu_ctxsw_trace_entry {
+        u8 tag;
+        u8 vmid;
+        u16 seqno;            /* sequence number to detect drops */
+        u32 context_id;       /* context_id as allocated by FECS */
+        u64 pid;              /* 64-bit is max bits of different OS pid */
+        u64 timestamp;        /* 64-bit time */
+};
 
 int gk20a_ctxsw_trace_init(struct gk20a *g);
 
@@ -37,7 +51,7 @@ void gk20a_ctxsw_trace_tsg_reset(struct gk20a *g, struct tsg_gk20a *tsg);
 
 void gk20a_ctxsw_trace_cleanup(struct gk20a *g);
 int gk20a_ctxsw_trace_write(struct gk20a *g,
-                            struct nvgpu_ctxsw_trace_entry *entry);
+                            struct nvgpu_gpu_ctxsw_trace_entry *entry);
 void gk20a_ctxsw_trace_wake_up(struct gk20a *g, int vmid);
 
 #ifdef CONFIG_GK20A_CTXSW_TRACE
diff --git a/drivers/gpu/nvgpu/os/linux/ctxsw_trace.c b/drivers/gpu/nvgpu/os/linux/ctxsw_trace.c
index a335988a..a8f5f815 100644
--- a/drivers/gpu/nvgpu/os/linux/ctxsw_trace.c
+++ b/drivers/gpu/nvgpu/os/linux/ctxsw_trace.c
@@ -20,9 +20,11 @@
 #include <linux/poll.h>
 #include <trace/events/gk20a.h>
 #include <uapi/linux/nvgpu.h>
+#include <nvgpu/ctxsw_trace.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/gr_gk20a.h"
+#include "gk20a/fecs_trace_gk20a.h"
 
 #include <nvgpu/kmem.h>
 #include <nvgpu/log.h>
@@ -43,7 +45,7 @@ struct gk20a_ctxsw_dev {
         struct gk20a *g;
 
         struct nvgpu_ctxsw_ring_header *hdr;
-        struct nvgpu_ctxsw_trace_entry *ents;
+        struct nvgpu_gpu_ctxsw_trace_entry *ents;
         struct nvgpu_ctxsw_trace_filter filter;
         bool write_enabled;
         struct nvgpu_cond readout_wq;
@@ -75,6 +77,17 @@ static inline int ring_len(struct nvgpu_ctxsw_ring_header *hdr)
         return (hdr->write_idx - hdr->read_idx) % hdr->num_ents;
 }
 
+static void nvgpu_set_ctxsw_trace_entry(struct nvgpu_ctxsw_trace_entry *entry_dst,
+        struct nvgpu_gpu_ctxsw_trace_entry *entry_src)
+{
+        entry_dst->tag = entry_src->tag;
+        entry_dst->vmid = entry_src->vmid;
+        entry_dst->seqno = entry_src->seqno;
+        entry_dst->context_id = entry_src->context_id;
+        entry_dst->pid = entry_src->pid;
+        entry_dst->timestamp = entry_src->timestamp;
+}
+
 ssize_t gk20a_ctxsw_dev_read(struct file *filp, char __user *buf, size_t size,
         loff_t *off)
 {
@@ -83,6 +96,7 @@ ssize_t gk20a_ctxsw_dev_read(struct file *filp, char __user *buf, size_t size,
         struct nvgpu_ctxsw_ring_header *hdr = dev->hdr;
         struct nvgpu_ctxsw_trace_entry __user *entry =
                 (struct nvgpu_ctxsw_trace_entry *) buf;
+        struct nvgpu_ctxsw_trace_entry user_entry;
         size_t copied = 0;
         int err;
 
@@ -101,11 +115,12 @@ ssize_t gk20a_ctxsw_dev_read(struct file *filp, char __user *buf, size_t size,
                 nvgpu_mutex_acquire(&dev->write_lock);
         }
 
-        while (size >= sizeof(struct nvgpu_ctxsw_trace_entry)) {
+        while (size >= sizeof(struct nvgpu_gpu_ctxsw_trace_entry)) {
                 if (ring_is_empty(hdr))
                         break;
 
-                if (copy_to_user(entry, &dev->ents[hdr->read_idx],
+                nvgpu_set_ctxsw_trace_entry(&user_entry, &dev->ents[hdr->read_idx]);
+                if (copy_to_user(entry, &user_entry,
                         sizeof(*entry))) {
                         nvgpu_mutex_release(&dev->write_lock);
                         return -EFAULT;
@@ -169,7 +184,7 @@ static int gk20a_ctxsw_dev_alloc_buffer(struct gk20a_ctxsw_dev *dev,
 
 
         dev->hdr = buf;
-        dev->ents = (struct nvgpu_ctxsw_trace_entry *) (dev->hdr + 1);
+        dev->ents = (struct nvgpu_gpu_ctxsw_trace_entry *) (dev->hdr + 1);
         dev->size = size;
         dev->num_ents = dev->hdr->num_ents;
 
@@ -191,8 +206,8 @@ int gk20a_ctxsw_dev_ring_alloc(struct gk20a *g,
         hdr->magic = NVGPU_CTXSW_RING_HEADER_MAGIC;
         hdr->version = NVGPU_CTXSW_RING_HEADER_VERSION;
         hdr->num_ents = (*size - sizeof(struct nvgpu_ctxsw_ring_header))
-                / sizeof(struct nvgpu_ctxsw_trace_entry);
-        hdr->ent_size = sizeof(struct nvgpu_ctxsw_trace_entry);
+                / sizeof(struct nvgpu_gpu_ctxsw_trace_entry);
+        hdr->ent_size = sizeof(struct nvgpu_gpu_ctxsw_trace_entry);
         hdr->drop_count = 0;
         hdr->read_idx = 0;
         hdr->write_idx = 0;
@@ -327,9 +342,9 @@ int gk20a_ctxsw_dev_open(struct inode *inode, struct file *filp)
         n = g->ops.fecs_trace.max_entries(g, &dev->filter);
 
         size = sizeof(struct nvgpu_ctxsw_ring_header) +
-                        n * sizeof(struct nvgpu_ctxsw_trace_entry);
+                        n * sizeof(struct nvgpu_gpu_ctxsw_trace_entry);
         nvgpu_log(g, gpu_dbg_ctxsw, "size=%zu entries=%d ent_size=%zu",
-                size, n, sizeof(struct nvgpu_ctxsw_trace_entry));
+                size, n, sizeof(struct nvgpu_gpu_ctxsw_trace_entry));
 
         err = gk20a_ctxsw_dev_alloc_buffer(dev, size);
         if (!err) {
@@ -583,7 +598,7 @@ void gk20a_ctxsw_trace_cleanup(struct gk20a *g)
 }
 
 int gk20a_ctxsw_trace_write(struct gk20a *g,
-                struct nvgpu_ctxsw_trace_entry *entry)
+                struct nvgpu_gpu_ctxsw_trace_entry *entry)
 {
         struct nvgpu_ctxsw_ring_header *hdr;
         struct gk20a_ctxsw_dev *dev;
@@ -692,7 +707,7 @@ void gk20a_ctxsw_trace_wake_up(struct gk20a *g, int vmid)
 void gk20a_ctxsw_trace_channel_reset(struct gk20a *g, struct channel_gk20a *ch)
 {
 #ifdef CONFIG_GK20A_CTXSW_TRACE
-        struct nvgpu_ctxsw_trace_entry entry = {
+        struct nvgpu_gpu_ctxsw_trace_entry entry = {
                 .vmid = 0,
                 .tag = NVGPU_CTXSW_TAG_ENGINE_RESET,
                 .context_id = 0,
@@ -712,7 +727,7 @@ void gk20a_ctxsw_trace_channel_reset(struct gk20a *g, struct channel_gk20a *ch)
 void gk20a_ctxsw_trace_tsg_reset(struct gk20a *g, struct tsg_gk20a *tsg)
 {
 #ifdef CONFIG_GK20A_CTXSW_TRACE
-        struct nvgpu_ctxsw_trace_entry entry = {
+        struct nvgpu_gpu_ctxsw_trace_entry entry = {
                 .vmid = 0,
                 .tag = NVGPU_CTXSW_TAG_ENGINE_RESET,
                 .context_id = 0,
diff --git a/drivers/gpu/nvgpu/os/linux/vgpu/fecs_trace_vgpu.c b/drivers/gpu/nvgpu/os/linux/vgpu/fecs_trace_vgpu.c
index 6339aef9..255ee2bd 100644
--- a/drivers/gpu/nvgpu/os/linux/vgpu/fecs_trace_vgpu.c
+++ b/drivers/gpu/nvgpu/os/linux/vgpu/fecs_trace_vgpu.c
@@ -26,12 +26,13 @@
 
 #include "gk20a/gk20a.h"
 #include "os/linux/os_linux.h"
+#include "gk20a/fecs_trace_gk20a.h"
 #include "vgpu/fecs_trace_vgpu.h"
 
 struct vgpu_fecs_trace {
         struct tegra_hv_ivm_cookie *cookie;
         struct nvgpu_ctxsw_ring_header *header;
-        struct nvgpu_ctxsw_trace_entry *entries;
+        struct nvgpu_gpu_ctxsw_trace_entry *entries;
         int num_entries;
         bool enabled;
         void *buf;