gpu: nvgpu: add safe channel id lookup

Add gk20a_channel_from_id() to retrieve a channel, given a raw channel
ID, with a reference taken (or NULL if the channel was dead). This makes
it harder to mistakenly use a channel that's dead and thus uncovers bugs
sooner. Convert code to use the new lookup when applicable; work remains
to convert complex uses where a ref should have been taken but hasn't.

The channel ID is also validated against FIFO_INVAL_CHANNEL_ID; NULL is
returned for such IDs. This is often useful and does not hurt when
unnecessary.

However, this does not prevent the case where a channel would be closed
and reopened again when someone would hold a stale channel number. In
all such conditions the caller should hold a reference already.

The only conditions where a channel can be safely looked up by an id and
used without taking a ref are when initializing or deinitializing the
list of channels.

Jira NVGPU-1460

Change-Id: I0a30968d17c1e0784d315a676bbe69c03a73481c
Signed-off-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1955400
Signed-off-by: Debarshi Dutta <ddutta@nvidia.com>
(cherry picked from commit 7df3d587502c2de997dfbe8ea8ddc114d0a0481e
in dev-kernel)
Reviewed-on: https://git-master.nvidia.com/r/2008515
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

2 files changed, 13 insertions, 12 deletions

diff --git a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
index 9691d51b..f4a4591d 100644
--- a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
@@ -1172,7 +1172,7 @@ gk20a_refch_from_inst_ptr(struct gk20a *g, u64 inst_ptr)
                 struct channel_gk20a *ch;
                 u64 ch_inst_ptr;
 
-                ch = gk20a_channel_get(&f->channel[ci]);
+                ch = gk20a_channel_from_id(g, ci);
                 /* only alive channels are searched */
                 if (!ch) {
                         continue;
@@ -1959,9 +1959,9 @@ void gk20a_fifo_recover_ch(struct gk20a *g, u32 chid, bool verbose, int rc_type)
                 gk20a_fifo_recover(g, engines, chid, false, true, verbose,
                                         rc_type);
         } else {
-                struct channel_gk20a *ch = &g->fifo.channel[chid];
+                struct channel_gk20a *ch = gk20a_channel_from_id(g, chid);
 
-                if (gk20a_channel_get(ch)) {
+                if (ch != NULL) {
                         gk20a_channel_abort(ch, false);
 
                         if (gk20a_fifo_error_ch(g, ch)) {
@@ -2710,9 +2710,9 @@ static void gk20a_fifo_pbdma_fault_rc(struct gk20a *g,
         id = fifo_pbdma_status_id_v(status);
         if (fifo_pbdma_status_id_type_v(status)
                         == fifo_pbdma_status_id_type_chid_v()) {
-                struct channel_gk20a *ch = &f->channel[id];
+                struct channel_gk20a *ch = gk20a_channel_from_id(g, id);
 
-                if (gk20a_channel_get(ch)) {
+                if (ch != NULL) {
                         g->ops.fifo.set_error_notifier(ch, error_notifier);
                         gk20a_fifo_recover_ch(g, id, true, RC_TYPE_PBDMA_FAULT);
                         gk20a_channel_put(ch);
@@ -2924,12 +2924,12 @@ void gk20a_fifo_preempt_timeout_rc(struct gk20a *g, u32 id,
                 gk20a_fifo_recover_tsg(g, id, true,
                                                 RC_TYPE_PREEMPT_TIMEOUT);
         } else {
-                struct channel_gk20a *ch = &g->fifo.channel[id];
+                struct channel_gk20a *ch = gk20a_channel_from_id(g, id);
 
                 nvgpu_err(g,
                         "preempt channel %d timeout", id);
 
-                if (gk20a_channel_get(ch)) {
+                if (ch != NULL) {
                         g->ops.fifo.set_error_notifier(ch,
                                         NVGPU_ERR_NOTIFIER_FIFO_ERROR_IDLE_TIMEOUT);
                         gk20a_fifo_recover_ch(g, id, true,
@@ -4031,8 +4031,8 @@ void gk20a_debug_dump_all_channel_status_ramfc(struct gk20a *g,
         }
 
         for (chid = 0; chid < f->num_channels; chid++) {
-                struct channel_gk20a *ch = &f->channel[chid];
-                if (gk20a_channel_get(ch)) {
+                struct channel_gk20a *ch = gk20a_channel_from_id(g, chid);
+                if (ch != NULL) {
                         ch_state[chid] =
                                 nvgpu_kmalloc(g, sizeof(struct ch_state) +
                                         ram_in_alloc_size_v());
diff --git a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
index 8cb66279..928d80cb 100644
--- a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
@@ -5581,15 +5581,16 @@ static struct channel_gk20a *gk20a_gr_get_channel_from_ctx(
                 if (gr->chid_tlb[i].curr_ctx == curr_ctx) {
                         chid = gr->chid_tlb[i].chid;
                         tsgid = gr->chid_tlb[i].tsgid;
-                        ret = gk20a_channel_get(&f->channel[chid]);
+                        ret = gk20a_channel_from_id(g, chid);
                         goto unlock;
                 }
         }
 
         /* slow path */
         for (chid = 0; chid < f->num_channels; chid++) {
-                struct channel_gk20a *ch = &f->channel[chid];
-                if (gk20a_channel_get(ch) == NULL) {
+                struct channel_gk20a *ch = gk20a_channel_from_id(g, chid);
+
+                if (ch == NULL) {
                         continue;
                 }