Fix a pre-4.19 bug in seq procfs files and add detailed channel print

- The sequence file infrastructure prior to kernel version 4.19
  has a bug in the retry code when the write buffer overflows that
  causes our private iterator state to be corrupted. Work around
  this by tracking some info out-of-band.
- Now supports including detailed channel status information from
  channel RAM when printing the runlist.
- Adds helper function to probe for and return struct gk20a*.

1 files changed, 20 insertions, 15 deletions

diff --git a/runlist.c b/runlist.c
index 8691b51..8dfa1c7 100644
--- a/runlist.c
+++ b/runlist.c
@@ -8,23 +8,20 @@
 // Bus types are global symbols in the kernel
 extern struct bus_type platform_bus_type;
 
-int get_runlist_iter(struct runlist_iter *rl_iter) {
+struct gk20a* get_live_gk20a(void) {
         struct device *dev = NULL;
         struct device *temp_dev;
         struct gk20a *g;
-        struct entry_tsg head;
-        runlist_base_t rl_base;
-        runlist_info_t rl_info;
-        u64 runlist_iova;
+        struct nvgpu_os_linux *l;
         // Get the last device that matches our name
         while ((temp_dev = bus_find_device_by_name(&platform_bus_type, dev, "17000000.gv11b"))) {
                 dev = temp_dev;
                 printk(KERN_INFO "[nvdebug] Found a matching device %s\n", dev_name(dev));
         }
         if (!dev)
-                return -EIO;
+                return NULL;
         g = get_gk20a(dev);
-        // This address seems to not be:
+        // The address pointed to `regs` + NV_PFIFO_RUNLIST_BASE seems to not be:
         // - A GPU address (type is sysmem_coherent)
         // - A physical address (dereferencing after ioremap crashes)
         // - A kernel virtual address (dereferencing segfaults)
@@ -49,15 +46,23 @@ int get_runlist_iter(struct runlist_iter *rl_iter) {
         // BUT, it turns out that it IS JUST A PHYSICAL ADDRESS! It wasn't working
         // before because the GPU had simply gone to sleep and invalidated its
         // register state, so nvgpu_readl() was simply returning garbage.
-
-        printk(KERN_INFO "[nvdebug] Pulling runlist base address from %x\n", NV_PFIFO_RUNLIST_BASE);
-        printk(KERN_INFO "[nvdebug] Using struct gk20a* of %px\n", g);
-        printk(KERN_INFO "[nvdebug] g->name: %s, g->power_on: %d, g->sw_ready: %d, g->is_virtual %d\n",
-                g->name, g->power_on, g->sw_ready, g->is_virtual);
-        struct nvgpu_os_linux *l = container_of(g, struct nvgpu_os_linux, g);
-        printk(KERN_INFO "[nvdebug] l->regs %px, l->regs_saved %px\n", l->regs, l->regs_saved);
+        l = container_of(g, struct nvgpu_os_linux, g);
         if (!l->regs)
-                return -EIO;
+                return NULL;
+        return g;
+}
+
+/* Get runlist head and info (incl. length)
+   @param rl_iter Location at which to store output
+*/
+int get_runlist_iter(struct runlist_iter *rl_iter) {
+        struct entry_tsg head;
+        runlist_base_t rl_base;
+        runlist_info_t rl_info;
+        u64 runlist_iova;
+        struct gk20a *g = get_live_gk20a();
+        if (!g)
+                return -EIO;
         rl_base.raw = nvdebug_readl(g, NV_PFIFO_RUNLIST_BASE);
         rl_info.raw = nvdebug_readl(g, NV_PFIFO_RUNLIST);
         runlist_iova = ((u64)rl_base.ptr) << 12;