gpu: nvgpu: move channel header to common

channel_gk20a is clear from chip specifics and from most dependencies,
so move it under the common directory.

Jira NVGPU-967

Change-Id: I41f2160b96d4ec84064288ecc22bb360e82352df
Signed-off-by: Konsta Holtta <kholtta@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1810578
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

60 files changed, 473 insertions, 464 deletions

diff --git a/drivers/gpu/nvgpu/common/fb/fb_gv11b.c b/drivers/gpu/nvgpu/common/fb/fb_gv11b.c
index c46d3976..b3d8cdb9 100644
--- a/drivers/gpu/nvgpu/common/fb/fb_gv11b.c
+++ b/drivers/gpu/nvgpu/common/fb/fb_gv11b.c
@@ -33,6 +33,7 @@
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
 #include <nvgpu/timers.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/mm_gk20a.h"
diff --git a/drivers/gpu/nvgpu/common/fifo/channel.c b/drivers/gpu/nvgpu/common/fifo/channel.c
index a11d9562..58b62ce7 100644
--- a/drivers/gpu/nvgpu/common/fifo/channel.c
+++ b/drivers/gpu/nvgpu/common/fifo/channel.c
@@ -43,6 +43,7 @@
 #include <nvgpu/os_sched.h>
 #include <nvgpu/log2.h>
 #include <nvgpu/ptimer.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/dbg_gpu_gk20a.h"
diff --git a/drivers/gpu/nvgpu/common/fifo/submit.c b/drivers/gpu/nvgpu/common/fifo/submit.c
index 56512894..1f7a04a2 100644
--- a/drivers/gpu/nvgpu/common/fifo/submit.c
+++ b/drivers/gpu/nvgpu/common/fifo/submit.c
@@ -28,7 +28,6 @@
 #include <nvgpu/hw/gk20a/hw_pbdma_gk20a.h>
 
 #include "gk20a/gk20a.h"
-#include "gk20a/channel_gk20a.h"
 #include "gk20a/fence_gk20a.h"
 #include "gk20a/channel_sync_gk20a.h"
 
diff --git a/drivers/gpu/nvgpu/gk20a/ce2_gk20a.c b/drivers/gpu/nvgpu/gk20a/ce2_gk20a.c
index 4cc6c8ca..436d3205 100644
--- a/drivers/gpu/nvgpu/gk20a/ce2_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/ce2_gk20a.c
@@ -29,6 +29,7 @@
 #include <nvgpu/enabled.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a.h"
 #include "gk20a/fence_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gk20a/ce2_gk20a.h b/drivers/gpu/nvgpu/gk20a/ce2_gk20a.h
index 19a169da..f46d0d9d 100644
--- a/drivers/gpu/nvgpu/gk20a/ce2_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/ce2_gk20a.h
@@ -26,8 +26,8 @@
 #ifndef __CE2_GK20A_H__
 #define __CE2_GK20A_H__
 
-#include "channel_gk20a.h"
-#include "tsg_gk20a.h"
+struct channel_gk20a;
+struct tsg_gk20a;
 
 void gk20a_ce2_isr(struct gk20a *g, u32 inst_id, u32 pri_base);
 u32 gk20a_ce2_nonstall_isr(struct gk20a *g, u32 inst_id, u32 pri_base);
diff --git a/drivers/gpu/nvgpu/gk20a/channel_gk20a.h b/drivers/gpu/nvgpu/gk20a/channel_gk20a.h
deleted file mode 100644
index 51207552..00000000
--- a/drivers/gpu/nvgpu/gk20a/channel_gk20a.h
+++ /dev/null
@@ -1,439 +0,0 @@
-/*
- * GK20A graphics channel
- *
- * Copyright (c) 2011-2018, 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"),
- * to deal in the Software without restriction, including without limitation
- * the rights to use, copy, modify, merge, publish, distribute, sublicense,
- * and/or sell copies of the Software, and to permit persons to whom the
- * Software is furnished to do so, subject to the following conditions:
- *
- * The above copyright notice and this permission notice shall be included in
- * all copies or substantial portions of the Software.
- *
- * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
- * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
- * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
- * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
- * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
- * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
- * DEALINGS IN THE SOFTWARE.
- */
-#ifndef CHANNEL_GK20A_H
-#define CHANNEL_GK20A_H
-
-#include <nvgpu/list.h>
-#include <nvgpu/lock.h>
-#include <nvgpu/timers.h>
-#include <nvgpu/cond.h>
-#include <nvgpu/atomic.h>
-#include <nvgpu/nvgpu_mem.h>
-#include <nvgpu/allocator.h>
-
-struct gk20a;
-struct dbg_session_gk20a;
-struct gk20a_fence;
-struct fifo_profile_gk20a;
-struct gk20a_channel_sync;
-
-/* Flags to be passed to gk20a_channel_alloc_gpfifo() */
-#define NVGPU_GPFIFO_FLAGS_SUPPORT_VPR                  (1U << 0U)
-#define NVGPU_GPFIFO_FLAGS_SUPPORT_DETERMINISTIC        (1U << 1U)
-#define NVGPU_GPFIFO_FLAGS_REPLAYABLE_FAULTS_ENABLE     (1U << 2U)
-#define NVGPU_GPFIFO_FLAGS_USERMODE_SUPPORT             (1U << 3U)
-
-/* Flags to be passed to nvgpu_submit_channel_gpfifo() */
-#define NVGPU_SUBMIT_FLAGS_FENCE_WAIT   (1U << 0U)
-#define NVGPU_SUBMIT_FLAGS_FENCE_GET    (1U << 1U)
-#define NVGPU_SUBMIT_FLAGS_HW_FORMAT    (1U << 2U)
-#define NVGPU_SUBMIT_FLAGS_SYNC_FENCE   (1U << 3U)
-#define NVGPU_SUBMIT_FLAGS_SUPPRESS_WFI (1U << 4U)
-#define NVGPU_SUBMIT_FLAGS_SKIP_BUFFER_REFCOUNTING      (1U << 5U)
-
-/*
- * The binary format of 'struct nvgpu_channel_fence' introduced here
- * should match that of 'struct nvgpu_fence' defined in uapi header, since
- * this struct is intended to be a mirror copy of the uapi struct. This is
- * not a hard requirement though because of nvgpu_get_fence_args conversion
- * function.
- */
-struct nvgpu_channel_fence {
-        u32 id;
-        u32 value;
-};
-
-/*
- * The binary format of 'struct nvgpu_gpfifo_entry' introduced here
- * should match that of 'struct nvgpu_gpfifo' defined in uapi header, since
- * this struct is intended to be a mirror copy of the uapi struct. This is
- * a rigid requirement because there's no conversion function and there are
- * memcpy's present between the user gpfifo (of type nvgpu_gpfifo) and the
- * kern gpfifo (of type nvgpu_gpfifo_entry).
- */
-struct nvgpu_gpfifo_entry {
-        u32 entry0;
-        u32 entry1;
-};
-
-struct gpfifo_desc {
-        struct nvgpu_mem mem;
-        u32 entry_num;
-
-        u32 get;
-        u32 put;
-
-        bool wrap;
-
-        /* if gpfifo lives in vidmem or is forced to go via PRAMIN, first copy
-         * from userspace to pipe and then from pipe to gpu buffer */
-        void *pipe;
-};
-
-struct nvgpu_gpfifo_args {
-        u32 num_entries;
-        u32 num_inflight_jobs;
-        u32 userd_dmabuf_fd;
-        u32 gpfifo_dmabuf_fd;
-        u32 work_submit_token;
-        u32 flags;
-};
-
-struct notification {
-        struct {
-                u32 nanoseconds[2];
-        } timestamp;
-        u32 info32;
-        u16 info16;
-        u16 status;
-};
-
-struct priv_cmd_queue {
-        struct nvgpu_mem mem;
-        u32 size;       /* num of entries in words */
-        u32 put;        /* put for priv cmd queue */
-        u32 get;        /* get for priv cmd queue */
-};
-
-struct priv_cmd_entry {
-        bool valid;
-        struct nvgpu_mem *mem;
-        u32 off;        /* offset in mem, in u32 entries */
-        u64 gva;
-        u32 get;        /* start of entry in queue */
-        u32 size;       /* in words */
-};
-
-struct channel_gk20a_job {
-        struct nvgpu_mapped_buf **mapped_buffers;
-        int num_mapped_buffers;
-        struct gk20a_fence *post_fence;
-        struct priv_cmd_entry *wait_cmd;
-        struct priv_cmd_entry *incr_cmd;
-        struct nvgpu_list_node list;
-};
-
-static inline struct channel_gk20a_job *
-channel_gk20a_job_from_list(struct nvgpu_list_node *node)
-{
-        return (struct channel_gk20a_job *)
-        ((uintptr_t)node - offsetof(struct channel_gk20a_job, list));
-};
-
-struct channel_gk20a_joblist {
-        struct {
-                bool enabled;
-                unsigned int length;
-                unsigned int put;
-                unsigned int get;
-                struct channel_gk20a_job *jobs;
-                struct nvgpu_mutex read_lock;
-        } pre_alloc;
-
-        struct {
-                struct nvgpu_list_node jobs;
-                struct nvgpu_spinlock lock;
-        } dynamic;
-
-        /*
-         * Synchronize abort cleanup (when closing a channel) and job cleanup
-         * (asynchronously from worker) - protect from concurrent access when
-         * job resources are being freed.
-         */
-        struct nvgpu_mutex cleanup_lock;
-};
-
-struct channel_gk20a_timeout {
-        /* lock protects the running timer state */
-        struct nvgpu_raw_spinlock lock;
-        struct nvgpu_timeout timer;
-        bool running;
-        u32 gp_get;
-        u64 pb_get;
-
-        /* lock not needed */
-        u32 limit_ms;
-        bool enabled;
-        bool debug_dump;
-};
-
-/*
- * Track refcount actions, saving their stack traces. This number specifies how
- * many most recent actions are stored in a buffer. Set to 0 to disable. 128
- * should be enough to track moderately hard problems from the start.
- */
-#define GK20A_CHANNEL_REFCOUNT_TRACKING 0
-/* Stack depth for the saved actions. */
-#define GK20A_CHANNEL_REFCOUNT_TRACKING_STACKLEN 8
-
-/*
- * Because the puts and gets are not linked together explicitly (although they
- * should always come in pairs), it's not possible to tell which ref holder to
- * delete from the list when doing a put. So, just store some number of most
- * recent gets and puts in a ring buffer, to obtain a history.
- *
- * These are zeroed when a channel is closed, so a new one starts fresh.
- */
-
-enum channel_gk20a_ref_action_type {
-        channel_gk20a_ref_action_get,
-        channel_gk20a_ref_action_put
-};
-
-#if GK20A_CHANNEL_REFCOUNT_TRACKING
-
-#include <linux/stacktrace.h>
-
-struct channel_gk20a_ref_action {
-        enum channel_gk20a_ref_action_type type;
-        s64 timestamp_ms;
-        /*
-         * Many of these traces will be similar. Simpler to just capture
-         * duplicates than to have a separate database for the entries.
-         */
-        struct stack_trace trace;
-        unsigned long trace_entries[GK20A_CHANNEL_REFCOUNT_TRACKING_STACKLEN];
-};
-#endif
-
-/* this is the priv element of struct nvhost_channel */
-struct channel_gk20a {
-        struct gk20a *g; /* set only when channel is active */
-
-        struct nvgpu_list_node free_chs;
-
-        struct nvgpu_spinlock ref_obtain_lock;
-        nvgpu_atomic_t ref_count;
-        struct nvgpu_cond ref_count_dec_wq;
-#if GK20A_CHANNEL_REFCOUNT_TRACKING
-        /*
-         * Ring buffer for most recent refcount gets and puts. Protected by
-         * ref_actions_lock when getting or putting refs (i.e., adding
-         * entries), and when reading entries.
-         */
-        struct channel_gk20a_ref_action ref_actions[
-                GK20A_CHANNEL_REFCOUNT_TRACKING];
-        size_t ref_actions_put; /* index of next write */
-        struct nvgpu_spinlock ref_actions_lock;
-#endif
-
-        struct nvgpu_semaphore_int *hw_sema;
-
-        nvgpu_atomic_t bound;
-
-        int chid;
-        int tsgid;
-        pid_t pid;
-        pid_t tgid;
-        struct nvgpu_mutex ioctl_lock;
-
-        struct nvgpu_list_node ch_entry; /* channel's entry in TSG */
-
-        struct channel_gk20a_joblist joblist;
-        struct nvgpu_allocator fence_allocator;
-
-        struct vm_gk20a *vm;
-
-        struct gpfifo_desc gpfifo;
-
-        struct nvgpu_mem usermode_userd; /* Used for Usermode Submission */
-        struct nvgpu_mem usermode_gpfifo;
-        struct nvgpu_mem inst_block;
-
-        u64 userd_iova;
-        u64 userd_gpu_va;
-
-        struct priv_cmd_queue priv_cmd_q;
-
-        struct nvgpu_cond notifier_wq;
-        struct nvgpu_cond semaphore_wq;
-
-        /* kernel watchdog to kill stuck jobs */
-        struct channel_gk20a_timeout timeout;
-
-        /* for job cleanup handling in the background worker */
-        struct nvgpu_list_node worker_item;
-
-#if defined(CONFIG_GK20A_CYCLE_STATS)
-        struct {
-                void *cyclestate_buffer;
-                u32 cyclestate_buffer_size;
-                struct nvgpu_mutex cyclestate_buffer_mutex;
-        } cyclestate;
-
-        struct nvgpu_mutex cs_client_mutex;
-        struct gk20a_cs_snapshot_client *cs_client;
-#endif
-        struct nvgpu_mutex dbg_s_lock;
-        struct nvgpu_list_node dbg_s_list;
-
-        struct nvgpu_mutex sync_lock;
-        struct gk20a_channel_sync *sync;
-        struct gk20a_channel_sync *user_sync;
-
-#ifdef CONFIG_TEGRA_GR_VIRTUALIZATION
-        u64 virt_ctx;
-#endif
-
-        struct nvgpu_mem ctx_header;
-
-        /* Any operating system specific data. */
-        void *os_priv;
-
-        u32 obj_class;  /* we support only one obj per channel */
-
-        u32 timeout_accumulated_ms;
-        u32 timeout_gpfifo_get;
-
-        u32 subctx_id;
-        u32 runqueue_sel;
-
-        u32 timeout_ms_max;
-        u32 runlist_id;
-
-        bool mmu_nack_handled;
-        bool has_timedout;
-        bool referenceable;
-        bool vpr;
-        bool deterministic;
-        /* deterministic, but explicitly idle and submits disallowed */
-        bool deterministic_railgate_allowed;
-        bool cde;
-        bool usermode_submit_enabled;
-        bool timeout_debug_dump;
-        bool has_os_fence_framework_support;
-
-        bool is_privileged_channel;
-};
-
-static inline struct channel_gk20a *
-channel_gk20a_from_free_chs(struct nvgpu_list_node *node)
-{
-        return (struct channel_gk20a *)
-                   ((uintptr_t)node - offsetof(struct channel_gk20a, free_chs));
-};
-
-static inline struct channel_gk20a *
-channel_gk20a_from_ch_entry(struct nvgpu_list_node *node)
-{
-        return (struct channel_gk20a *)
-           ((uintptr_t)node - offsetof(struct channel_gk20a, ch_entry));
-};
-
-static inline struct channel_gk20a *
-channel_gk20a_from_worker_item(struct nvgpu_list_node *node)
-{
-        return (struct channel_gk20a *)
-           ((uintptr_t)node - offsetof(struct channel_gk20a, worker_item));
-};
-
-static inline bool gk20a_channel_as_bound(struct channel_gk20a *ch)
-{
-        return !!ch->vm;
-}
-int channel_gk20a_commit_va(struct channel_gk20a *c);
-int gk20a_init_channel_support(struct gk20a *, u32 chid);
-
-/* must be inside gk20a_busy()..gk20a_idle() */
-void gk20a_channel_close(struct channel_gk20a *ch);
-void __gk20a_channel_kill(struct channel_gk20a *ch);
-
-bool gk20a_channel_update_and_check_timeout(struct channel_gk20a *ch,
-                u32 timeout_delta_ms, bool *progress);
-void gk20a_disable_channel(struct channel_gk20a *ch);
-void gk20a_channel_abort(struct channel_gk20a *ch, bool channel_preempt);
-void gk20a_channel_abort_clean_up(struct channel_gk20a *ch);
-void gk20a_channel_semaphore_wakeup(struct gk20a *g, bool post_events);
-int gk20a_channel_alloc_priv_cmdbuf(struct channel_gk20a *c, u32 size,
-                             struct priv_cmd_entry *entry);
-int gk20a_free_priv_cmdbuf(struct channel_gk20a *c, struct priv_cmd_entry *e);
-
-int gk20a_enable_channel_tsg(struct gk20a *g, struct channel_gk20a *ch);
-int gk20a_disable_channel_tsg(struct gk20a *g, struct channel_gk20a *ch);
-
-int gk20a_channel_suspend(struct gk20a *g);
-int gk20a_channel_resume(struct gk20a *g);
-
-void gk20a_channel_deterministic_idle(struct gk20a *g);
-void gk20a_channel_deterministic_unidle(struct gk20a *g);
-
-int nvgpu_channel_worker_init(struct gk20a *g);
-void nvgpu_channel_worker_deinit(struct gk20a *g);
-
-struct channel_gk20a *gk20a_get_channel_from_file(int fd);
-void gk20a_channel_update(struct channel_gk20a *c);
-
-/* returns ch if reference was obtained */
-struct channel_gk20a *__must_check _gk20a_channel_get(struct channel_gk20a *ch,
-                                                      const char *caller);
-#define gk20a_channel_get(ch) _gk20a_channel_get(ch, __func__)
-
-
-void _gk20a_channel_put(struct channel_gk20a *ch, const char *caller);
-#define gk20a_channel_put(ch) _gk20a_channel_put(ch, __func__)
-
-int gk20a_wait_channel_idle(struct channel_gk20a *ch);
-
-/* runlist_id -1 is synonym for ENGINE_GR_GK20A runlist id */
-struct channel_gk20a *gk20a_open_new_channel(struct gk20a *g,
-                s32 runlist_id,
-                bool is_privileged_channel,
-                pid_t pid, pid_t tid);
-
-int gk20a_channel_alloc_gpfifo(struct channel_gk20a *c,
-                struct nvgpu_gpfifo_args *gpfifo_args);
-
-void gk20a_channel_timeout_restart_all_channels(struct gk20a *g);
-
-bool channel_gk20a_is_prealloc_enabled(struct channel_gk20a *c);
-void channel_gk20a_joblist_lock(struct channel_gk20a *c);
-void channel_gk20a_joblist_unlock(struct channel_gk20a *c);
-bool channel_gk20a_joblist_is_empty(struct channel_gk20a *c);
-
-int channel_gk20a_update_runlist(struct channel_gk20a *c, bool add);
-int gk20a_channel_get_timescale_from_timeslice(struct gk20a *g,
-                unsigned int timeslice_period,
-                unsigned int *__timeslice_timeout, unsigned int *__timeslice_scale);
-
-void gk20a_wait_until_counter_is_N(
-        struct channel_gk20a *ch, nvgpu_atomic_t *counter, int wait_value,
-        struct nvgpu_cond *c, const char *caller, const char *counter_name);
-int channel_gk20a_alloc_job(struct channel_gk20a *c,
-                struct channel_gk20a_job **job_out);
-void channel_gk20a_free_job(struct channel_gk20a *c,
-                struct channel_gk20a_job *job);
-u32 nvgpu_get_gp_free_count(struct channel_gk20a *c);
-u32 nvgpu_gp_free_count(struct channel_gk20a *c);
-int gk20a_channel_add_job(struct channel_gk20a *c,
-                                 struct channel_gk20a_job *job,
-                                 bool skip_buffer_refcounting);
-void free_priv_cmdbuf(struct channel_gk20a *c,
-                             struct priv_cmd_entry *e);
-void gk20a_channel_clean_up_jobs(struct channel_gk20a *c,
-                                        bool clean_all);
-
-void gk20a_channel_free_usermode_buffers(struct channel_gk20a *c);
-u32 nvgpu_get_gpfifo_entry_size(void);
-
-#endif /* CHANNEL_GK20A_H */
diff --git a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c
index f6134460..f78df0b5 100644
--- a/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/channel_sync_gk20a.c
@@ -30,6 +30,7 @@
 #include <nvgpu/list.h>
 #include <nvgpu/nvhost.h>
 #include <nvgpu/os_fence.h>
+#include <nvgpu/channel.h>
 
 #include "channel_sync_gk20a.h"
 #include "gk20a.h"
diff --git a/drivers/gpu/nvgpu/gk20a/css_gr_gk20a.c b/drivers/gpu/nvgpu/gk20a/css_gr_gk20a.c
index 4aeeec1c..00d1b196 100644
--- a/drivers/gpu/nvgpu/gk20a/css_gr_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/css_gr_gk20a.c
@@ -33,6 +33,7 @@
 #include <nvgpu/bug.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a.h"
 #include "css_gr_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.c b/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.c
index e7f95245..ef505425 100644
--- a/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/dbg_gpu_gk20a.c
@@ -30,6 +30,7 @@
 #include <nvgpu/bug.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a.h"
 #include "gr_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c b/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c
index 453d5b2f..3134df4d 100644
--- a/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fecs_trace_gk20a.c
@@ -38,6 +38,7 @@
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
 #include <nvgpu/timers.h>
+#include <nvgpu/channel.h>
 
 #include "fecs_trace_gk20a.h"
 #include "gk20a.h"
diff --git a/drivers/gpu/nvgpu/gk20a/fence_gk20a.c b/drivers/gpu/nvgpu/gk20a/fence_gk20a.c
index 0df73278..af421304 100644
--- a/drivers/gpu/nvgpu/gk20a/fence_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fence_gk20a.c
@@ -28,9 +28,9 @@
 #include <nvgpu/nvhost.h>
 #include <nvgpu/barrier.h>
 #include <nvgpu/os_fence.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a.h"
-#include "channel_gk20a.h"
 
 struct gk20a_fence_ops {
         int (*wait)(struct gk20a_fence *, long timeout);
diff --git a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
index 9dfe3083..e1702bd7 100644
--- a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.c
@@ -43,6 +43,7 @@
 #include <nvgpu/ptimer.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a.h"
 #include "mm_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.h b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.h
index 3ff1b088..21922426 100644
--- a/drivers/gpu/nvgpu/gk20a/fifo_gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/fifo_gk20a.h
@@ -24,14 +24,13 @@
 #ifndef FIFO_GK20A_H
 #define FIFO_GK20A_H
 
-#include "channel_gk20a.h"
-#include "tsg_gk20a.h"
-
 #include <nvgpu/kref.h>
 
 struct gk20a_debug_output;
 struct mmu_fault_info;
 struct nvgpu_semaphore;
+struct channel_gk20a;
+struct tsg_gk20a;
 
 enum {
         NVGPU_FIFO_RUNLIST_INTERLEAVE_LEVEL_LOW = 0,
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.h b/drivers/gpu/nvgpu/gk20a/gk20a.h
index 898dfec8..34d08b65 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.h
@@ -47,6 +47,7 @@ struct nvgpu_clk_arb;
 struct nvgpu_gpu_ctxsw_trace_filter;
 #endif
 struct priv_cmd_entry;
+struct nvgpu_gpfifo_args;
 
 #include <nvgpu/lock.h>
 #include <nvgpu/thread.h>
diff --git a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
index 2969743b..7bfc48ad 100644
--- a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
@@ -41,6 +41,7 @@
 #include <nvgpu/ecc.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a.h"
 #include "gr_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gk20a/mm_gk20a.c b/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
index 9fcaebff..5b76f166 100644
--- a/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/mm_gk20a.c
@@ -42,6 +42,7 @@
 #include <nvgpu/sizes.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a.h"
 #include "mm_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gk20a/tsg_gk20a.c b/drivers/gpu/nvgpu/gk20a/tsg_gk20a.c
index 6dc2e282..885ce172 100644
--- a/drivers/gpu/nvgpu/gk20a/tsg_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/tsg_gk20a.c
@@ -23,6 +23,7 @@
 #include <nvgpu/kmem.h>
 #include <nvgpu/log.h>
 #include <nvgpu/os_sched.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a.h"
 #include "tsg_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gm20b/fifo_gm20b.c b/drivers/gpu/nvgpu/gm20b/fifo_gm20b.c
index dd11d2c7..19653541 100644
--- a/drivers/gpu/nvgpu/gm20b/fifo_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/fifo_gm20b.c
@@ -35,6 +35,7 @@
 #include <nvgpu/enabled.h>
 #include <nvgpu/io.h>
 #include <nvgpu/bug.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/hw/gm20b/hw_ccsr_gm20b.h>
 #include <nvgpu/hw/gm20b/hw_ram_gm20b.h>
diff --git a/drivers/gpu/nvgpu/gm20b/gr_gm20b.c b/drivers/gpu/nvgpu/gm20b/gr_gm20b.c
index 101f4211..32e98e53 100644
--- a/drivers/gpu/nvgpu/gm20b/gr_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/gr_gm20b.c
@@ -29,6 +29,7 @@
 #include <nvgpu/fuse.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/gr_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gm20b/hal_gm20b.c b/drivers/gpu/nvgpu/gm20b/hal_gm20b.c
index 512da9df..2f90512e 100644
--- a/drivers/gpu/nvgpu/gm20b/hal_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/hal_gm20b.c
@@ -61,6 +61,7 @@
 #include <nvgpu/enabled.h>
 #include <nvgpu/ptimer.h>
 #include <nvgpu/error_notifier.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/hw/gm20b/hw_proj_gm20b.h>
 #include <nvgpu/hw/gm20b/hw_fifo_gm20b.h>
diff --git a/drivers/gpu/nvgpu/gp106/hal_gp106.c b/drivers/gpu/nvgpu/gp106/hal_gp106.c
index 24c4191d..3408cdfa 100644
--- a/drivers/gpu/nvgpu/gp106/hal_gp106.c
+++ b/drivers/gpu/nvgpu/gp106/hal_gp106.c
@@ -96,6 +96,7 @@
 #include <nvgpu/ctxsw_trace.h>
 #include <nvgpu/error_notifier.h>
 #include <nvgpu/clk_arb.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/hw/gp106/hw_proj_gp106.h>
 #include <nvgpu/hw/gp106/hw_fifo_gp106.h>
diff --git a/drivers/gpu/nvgpu/gp10b/ce_gp10b.h b/drivers/gpu/nvgpu/gp10b/ce_gp10b.h
index fbf5e6f8..525599af 100644
--- a/drivers/gpu/nvgpu/gp10b/ce_gp10b.h
+++ b/drivers/gpu/nvgpu/gp10b/ce_gp10b.h
@@ -24,8 +24,9 @@
 #ifndef __CE_GP10B_H__
 #define __CE_GP10B_H__
 
-#include "gk20a/channel_gk20a.h"
-#include "gk20a/tsg_gk20a.h"
+#include <nvgpu/types.h>
+
+struct gk20a;
 
 void gp10b_ce_isr(struct gk20a *g, u32 inst_id, u32 pri_base);
 u32 gp10b_ce_nonstall_isr(struct gk20a *g, u32 inst_id, u32 pri_base);
diff --git a/drivers/gpu/nvgpu/gp10b/fifo_gp10b.c b/drivers/gpu/nvgpu/gp10b/fifo_gp10b.c
index c10e389c..7bf6e769 100644
--- a/drivers/gpu/nvgpu/gp10b/fifo_gp10b.c
+++ b/drivers/gpu/nvgpu/gp10b/fifo_gp10b.c
@@ -28,6 +28,7 @@
 #include <nvgpu/enabled.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "fifo_gp10b.h"
 
diff --git a/drivers/gpu/nvgpu/gp10b/gr_gp10b.c b/drivers/gpu/nvgpu/gp10b/gr_gp10b.c
index 0aa32f8a..92e0165f 100644
--- a/drivers/gpu/nvgpu/gp10b/gr_gp10b.c
+++ b/drivers/gpu/nvgpu/gp10b/gr_gp10b.c
@@ -32,6 +32,7 @@
 #include <nvgpu/enabled.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/gr_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gp10b/hal_gp10b.c b/drivers/gpu/nvgpu/gp10b/hal_gp10b.c
index 25466e7b..321a89fd 100644
--- a/drivers/gpu/nvgpu/gp10b/hal_gp10b.c
+++ b/drivers/gpu/nvgpu/gp10b/hal_gp10b.c
@@ -78,6 +78,7 @@
 #include <nvgpu/ptimer.h>
 #include <nvgpu/ctxsw_trace.h>
 #include <nvgpu/error_notifier.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/hw/gp10b/hw_proj_gp10b.h>
 #include <nvgpu/hw/gp10b/hw_fifo_gp10b.h>
diff --git a/drivers/gpu/nvgpu/gv100/hal_gv100.c b/drivers/gpu/nvgpu/gv100/hal_gv100.c
index 6a4f874d..1bc5d091 100644
--- a/drivers/gpu/nvgpu/gv100/hal_gv100.c
+++ b/drivers/gpu/nvgpu/gv100/hal_gv100.c
@@ -113,6 +113,7 @@
 #include <nvgpu/ctxsw_trace.h>
 #include <nvgpu/error_notifier.h>
 #include <nvgpu/clk_arb.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/hw/gv100/hw_proj_gv100.h>
 #include <nvgpu/hw/gv100/hw_fifo_gv100.h>
diff --git a/drivers/gpu/nvgpu/gv11b/css_gr_gv11b.c b/drivers/gpu/nvgpu/gv11b/css_gr_gv11b.c
index 17d9e051..4707ab31 100644
--- a/drivers/gpu/nvgpu/gv11b/css_gr_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/css_gr_gv11b.c
@@ -35,6 +35,7 @@
 #include <nvgpu/dma.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/css_gr_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gv11b/fifo_gv11b.c b/drivers/gpu/nvgpu/gv11b/fifo_gv11b.c
index 893835a4..0119c215 100644
--- a/drivers/gpu/nvgpu/gv11b/fifo_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/fifo_gv11b.c
@@ -39,10 +39,10 @@
 #include <nvgpu/ptimer.h>
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/fifo_gk20a.h"
-#include "gk20a/channel_gk20a.h"
 
 #include "gp10b/fifo_gp10b.h"
 
diff --git a/drivers/gpu/nvgpu/gv11b/gr_gv11b.c b/drivers/gpu/nvgpu/gv11b/gr_gv11b.c
index 0fbba3a0..0938e2ed 100644
--- a/drivers/gpu/nvgpu/gv11b/gr_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/gr_gv11b.c
@@ -35,6 +35,7 @@
 #include <nvgpu/io.h>
 #include <nvgpu/utils.h>
 #include <nvgpu/bitops.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/gr_gk20a.h"
diff --git a/drivers/gpu/nvgpu/gv11b/hal_gv11b.c b/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
index 8d3fb86b..e5d7e632 100644
--- a/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
@@ -94,6 +94,7 @@
 #include <nvgpu/ctxsw_trace.h>
 #include <nvgpu/error_notifier.h>
 #include <nvgpu/bug.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/hw/gv11b/hw_proj_gv11b.h>
 #include <nvgpu/hw/gv11b/hw_fifo_gv11b.h>
diff --git a/drivers/gpu/nvgpu/gv11b/subctx_gv11b.c b/drivers/gpu/nvgpu/gv11b/subctx_gv11b.c
index 4e429567..25626cd4 100644
--- a/drivers/gpu/nvgpu/gv11b/subctx_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/subctx_gv11b.c
@@ -30,6 +30,7 @@
 #include <nvgpu/log.h>
 #include <nvgpu/gmmu.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/hw/gv11b/hw_ram_gv11b.h>
 #include <nvgpu/hw/gv11b/hw_ctxsw_prog_gv11b.h>
diff --git a/drivers/gpu/nvgpu/include/nvgpu/channel.h b/drivers/gpu/nvgpu/include/nvgpu/channel.h
index 7434f0e7..6cca843e 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/channel.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/channel.h
@@ -23,13 +23,418 @@
 #ifndef NVGPU_CHANNEL_H
 #define NVGPU_CHANNEL_H
 
-#include <nvgpu/types.h>
+#include <nvgpu/list.h>
+#include <nvgpu/lock.h>
+#include <nvgpu/timers.h>
+#include <nvgpu/cond.h>
+#include <nvgpu/atomic.h>
+#include <nvgpu/nvgpu_mem.h>
+#include <nvgpu/allocator.h>
 
-#include "gk20a/gk20a.h"
-
-struct nvgpu_channel_fence;
+struct gk20a;
+struct dbg_session_gk20a;
 struct gk20a_fence;
 struct fifo_profile_gk20a;
+struct gk20a_channel_sync;
+struct nvgpu_gpfifo_userdata;
+
+/* Flags to be passed to gk20a_channel_alloc_gpfifo() */
+#define NVGPU_GPFIFO_FLAGS_SUPPORT_VPR                  (1U << 0U)
+#define NVGPU_GPFIFO_FLAGS_SUPPORT_DETERMINISTIC        (1U << 1U)
+#define NVGPU_GPFIFO_FLAGS_REPLAYABLE_FAULTS_ENABLE     (1U << 2U)
+#define NVGPU_GPFIFO_FLAGS_USERMODE_SUPPORT             (1U << 3U)
+
+/* Flags to be passed to nvgpu_submit_channel_gpfifo() */
+#define NVGPU_SUBMIT_FLAGS_FENCE_WAIT   (1U << 0U)
+#define NVGPU_SUBMIT_FLAGS_FENCE_GET    (1U << 1U)
+#define NVGPU_SUBMIT_FLAGS_HW_FORMAT    (1U << 2U)
+#define NVGPU_SUBMIT_FLAGS_SYNC_FENCE   (1U << 3U)
+#define NVGPU_SUBMIT_FLAGS_SUPPRESS_WFI (1U << 4U)
+#define NVGPU_SUBMIT_FLAGS_SKIP_BUFFER_REFCOUNTING      (1U << 5U)
+
+/*
+ * The binary format of 'struct nvgpu_channel_fence' introduced here
+ * should match that of 'struct nvgpu_fence' defined in uapi header, since
+ * this struct is intended to be a mirror copy of the uapi struct. This is
+ * not a hard requirement though because of nvgpu_get_fence_args conversion
+ * function.
+ */
+struct nvgpu_channel_fence {
+        u32 id;
+        u32 value;
+};
+
+/*
+ * The binary format of 'struct nvgpu_gpfifo_entry' introduced here
+ * should match that of 'struct nvgpu_gpfifo' defined in uapi header, since
+ * this struct is intended to be a mirror copy of the uapi struct. This is
+ * a rigid requirement because there's no conversion function and there are
+ * memcpy's present between the user gpfifo (of type nvgpu_gpfifo) and the
+ * kern gpfifo (of type nvgpu_gpfifo_entry).
+ */
+struct nvgpu_gpfifo_entry {
+        u32 entry0;
+        u32 entry1;
+};
+
+struct gpfifo_desc {
+        struct nvgpu_mem mem;
+        u32 entry_num;
+
+        u32 get;
+        u32 put;
+
+        bool wrap;
+
+        /* if gpfifo lives in vidmem or is forced to go via PRAMIN, first copy
+         * from userspace to pipe and then from pipe to gpu buffer */
+        void *pipe;
+};
+
+struct nvgpu_gpfifo_args {
+        u32 num_entries;
+        u32 num_inflight_jobs;
+        u32 userd_dmabuf_fd;
+        u32 gpfifo_dmabuf_fd;
+        u32 work_submit_token;
+        u32 flags;
+};
+
+struct notification {
+        struct {
+                u32 nanoseconds[2];
+        } timestamp;
+        u32 info32;
+        u16 info16;
+        u16 status;
+};
+
+struct priv_cmd_queue {
+        struct nvgpu_mem mem;
+        u32 size;       /* num of entries in words */
+        u32 put;        /* put for priv cmd queue */
+        u32 get;        /* get for priv cmd queue */
+};
+
+struct priv_cmd_entry {
+        bool valid;
+        struct nvgpu_mem *mem;
+        u32 off;        /* offset in mem, in u32 entries */
+        u64 gva;
+        u32 get;        /* start of entry in queue */
+        u32 size;       /* in words */
+};
+
+struct channel_gk20a_job {
+        struct nvgpu_mapped_buf **mapped_buffers;
+        int num_mapped_buffers;
+        struct gk20a_fence *post_fence;
+        struct priv_cmd_entry *wait_cmd;
+        struct priv_cmd_entry *incr_cmd;
+        struct nvgpu_list_node list;
+};
+
+static inline struct channel_gk20a_job *
+channel_gk20a_job_from_list(struct nvgpu_list_node *node)
+{
+        return (struct channel_gk20a_job *)
+        ((uintptr_t)node - offsetof(struct channel_gk20a_job, list));
+};
+
+struct channel_gk20a_joblist {
+        struct {
+                bool enabled;
+                unsigned int length;
+                unsigned int put;
+                unsigned int get;
+                struct channel_gk20a_job *jobs;
+                struct nvgpu_mutex read_lock;
+        } pre_alloc;
+
+        struct {
+                struct nvgpu_list_node jobs;
+                struct nvgpu_spinlock lock;
+        } dynamic;
+
+        /*
+         * Synchronize abort cleanup (when closing a channel) and job cleanup
+         * (asynchronously from worker) - protect from concurrent access when
+         * job resources are being freed.
+         */
+        struct nvgpu_mutex cleanup_lock;
+};
+
+struct channel_gk20a_timeout {
+        /* lock protects the running timer state */
+        struct nvgpu_raw_spinlock lock;
+        struct nvgpu_timeout timer;
+        bool running;
+        u32 gp_get;
+        u64 pb_get;
+
+        /* lock not needed */
+        u32 limit_ms;
+        bool enabled;
+        bool debug_dump;
+};
+
+/*
+ * Track refcount actions, saving their stack traces. This number specifies how
+ * many most recent actions are stored in a buffer. Set to 0 to disable. 128
+ * should be enough to track moderately hard problems from the start.
+ */
+#define GK20A_CHANNEL_REFCOUNT_TRACKING 0
+/* Stack depth for the saved actions. */
+#define GK20A_CHANNEL_REFCOUNT_TRACKING_STACKLEN 8
+
+/*
+ * Because the puts and gets are not linked together explicitly (although they
+ * should always come in pairs), it's not possible to tell which ref holder to
+ * delete from the list when doing a put. So, just store some number of most
+ * recent gets and puts in a ring buffer, to obtain a history.
+ *
+ * These are zeroed when a channel is closed, so a new one starts fresh.
+ */
+
+enum channel_gk20a_ref_action_type {
+        channel_gk20a_ref_action_get,
+        channel_gk20a_ref_action_put
+};
+
+#if GK20A_CHANNEL_REFCOUNT_TRACKING
+
+#include <linux/stacktrace.h>
+
+struct channel_gk20a_ref_action {
+        enum channel_gk20a_ref_action_type type;
+        s64 timestamp_ms;
+        /*
+         * Many of these traces will be similar. Simpler to just capture
+         * duplicates than to have a separate database for the entries.
+         */
+        struct stack_trace trace;
+        unsigned long trace_entries[GK20A_CHANNEL_REFCOUNT_TRACKING_STACKLEN];
+};
+#endif
+
+/* this is the priv element of struct nvhost_channel */
+struct channel_gk20a {
+        struct gk20a *g; /* set only when channel is active */
+
+        struct nvgpu_list_node free_chs;
+
+        struct nvgpu_spinlock ref_obtain_lock;
+        nvgpu_atomic_t ref_count;
+        struct nvgpu_cond ref_count_dec_wq;
+#if GK20A_CHANNEL_REFCOUNT_TRACKING
+        /*
+         * Ring buffer for most recent refcount gets and puts. Protected by
+         * ref_actions_lock when getting or putting refs (i.e., adding
+         * entries), and when reading entries.
+         */
+        struct channel_gk20a_ref_action ref_actions[
+                GK20A_CHANNEL_REFCOUNT_TRACKING];
+        size_t ref_actions_put; /* index of next write */
+        struct nvgpu_spinlock ref_actions_lock;
+#endif
+
+        struct nvgpu_semaphore_int *hw_sema;
+
+        nvgpu_atomic_t bound;
+
+        int chid;
+        int tsgid;
+        pid_t pid;
+        pid_t tgid;
+        struct nvgpu_mutex ioctl_lock;
+
+        struct nvgpu_list_node ch_entry; /* channel's entry in TSG */
+
+        struct channel_gk20a_joblist joblist;
+        struct nvgpu_allocator fence_allocator;
+
+        struct vm_gk20a *vm;
+
+        struct gpfifo_desc gpfifo;
+
+        struct nvgpu_mem usermode_userd; /* Used for Usermode Submission */
+        struct nvgpu_mem usermode_gpfifo;
+        struct nvgpu_mem inst_block;
+
+        u64 userd_iova;
+        u64 userd_gpu_va;
+
+        struct priv_cmd_queue priv_cmd_q;
+
+        struct nvgpu_cond notifier_wq;
+        struct nvgpu_cond semaphore_wq;
+
+        /* kernel watchdog to kill stuck jobs */
+        struct channel_gk20a_timeout timeout;
+
+        /* for job cleanup handling in the background worker */
+        struct nvgpu_list_node worker_item;
+
+#if defined(CONFIG_GK20A_CYCLE_STATS)
+        struct {
+                void *cyclestate_buffer;
+                u32 cyclestate_buffer_size;
+                struct nvgpu_mutex cyclestate_buffer_mutex;
+        } cyclestate;
+
+        struct nvgpu_mutex cs_client_mutex;
+        struct gk20a_cs_snapshot_client *cs_client;
+#endif
+        struct nvgpu_mutex dbg_s_lock;
+        struct nvgpu_list_node dbg_s_list;
+
+        struct nvgpu_mutex sync_lock;
+        struct gk20a_channel_sync *sync;
+        struct gk20a_channel_sync *user_sync;
+
+#ifdef CONFIG_TEGRA_GR_VIRTUALIZATION
+        u64 virt_ctx;
+#endif
+
+        struct nvgpu_mem ctx_header;
+
+        /* Any operating system specific data. */
+        void *os_priv;
+
+        u32 obj_class;  /* we support only one obj per channel */
+
+        u32 timeout_accumulated_ms;
+        u32 timeout_gpfifo_get;
+
+        u32 subctx_id;
+        u32 runqueue_sel;
+
+        u32 timeout_ms_max;
+        u32 runlist_id;
+
+        bool mmu_nack_handled;
+        bool has_timedout;
+        bool referenceable;
+        bool vpr;
+        bool deterministic;
+        /* deterministic, but explicitly idle and submits disallowed */
+        bool deterministic_railgate_allowed;
+        bool cde;
+        bool usermode_submit_enabled;
+        bool timeout_debug_dump;
+        bool has_os_fence_framework_support;
+
+        bool is_privileged_channel;
+};
+
+static inline struct channel_gk20a *
+channel_gk20a_from_free_chs(struct nvgpu_list_node *node)
+{
+        return (struct channel_gk20a *)
+                   ((uintptr_t)node - offsetof(struct channel_gk20a, free_chs));
+};
+
+static inline struct channel_gk20a *
+channel_gk20a_from_ch_entry(struct nvgpu_list_node *node)
+{
+        return (struct channel_gk20a *)
+           ((uintptr_t)node - offsetof(struct channel_gk20a, ch_entry));
+};
+
+static inline struct channel_gk20a *
+channel_gk20a_from_worker_item(struct nvgpu_list_node *node)
+{
+        return (struct channel_gk20a *)
+           ((uintptr_t)node - offsetof(struct channel_gk20a, worker_item));
+};
+
+static inline bool gk20a_channel_as_bound(struct channel_gk20a *ch)
+{
+        return !!ch->vm;
+}
+int channel_gk20a_commit_va(struct channel_gk20a *c);
+int gk20a_init_channel_support(struct gk20a *, u32 chid);
+
+/* must be inside gk20a_busy()..gk20a_idle() */
+void gk20a_channel_close(struct channel_gk20a *ch);
+void __gk20a_channel_kill(struct channel_gk20a *ch);
+
+bool gk20a_channel_update_and_check_timeout(struct channel_gk20a *ch,
+                u32 timeout_delta_ms, bool *progress);
+void gk20a_disable_channel(struct channel_gk20a *ch);
+void gk20a_channel_abort(struct channel_gk20a *ch, bool channel_preempt);
+void gk20a_channel_abort_clean_up(struct channel_gk20a *ch);
+void gk20a_channel_semaphore_wakeup(struct gk20a *g, bool post_events);
+int gk20a_channel_alloc_priv_cmdbuf(struct channel_gk20a *c, u32 size,
+                             struct priv_cmd_entry *entry);
+int gk20a_free_priv_cmdbuf(struct channel_gk20a *c, struct priv_cmd_entry *e);
+
+int gk20a_enable_channel_tsg(struct gk20a *g, struct channel_gk20a *ch);
+int gk20a_disable_channel_tsg(struct gk20a *g, struct channel_gk20a *ch);
+
+int gk20a_channel_suspend(struct gk20a *g);
+int gk20a_channel_resume(struct gk20a *g);
+
+void gk20a_channel_deterministic_idle(struct gk20a *g);
+void gk20a_channel_deterministic_unidle(struct gk20a *g);
+
+int nvgpu_channel_worker_init(struct gk20a *g);
+void nvgpu_channel_worker_deinit(struct gk20a *g);
+
+struct channel_gk20a *gk20a_get_channel_from_file(int fd);
+void gk20a_channel_update(struct channel_gk20a *c);
+
+/* returns ch if reference was obtained */
+struct channel_gk20a *__must_check _gk20a_channel_get(struct channel_gk20a *ch,
+                                                      const char *caller);
+#define gk20a_channel_get(ch) _gk20a_channel_get(ch, __func__)
+
+
+void _gk20a_channel_put(struct channel_gk20a *ch, const char *caller);
+#define gk20a_channel_put(ch) _gk20a_channel_put(ch, __func__)
+
+int gk20a_wait_channel_idle(struct channel_gk20a *ch);
+
+/* runlist_id -1 is synonym for ENGINE_GR_GK20A runlist id */
+struct channel_gk20a *gk20a_open_new_channel(struct gk20a *g,
+                s32 runlist_id,
+                bool is_privileged_channel,
+                pid_t pid, pid_t tid);
+
+int gk20a_channel_alloc_gpfifo(struct channel_gk20a *c,
+                struct nvgpu_gpfifo_args *gpfifo_args);
+
+void gk20a_channel_timeout_restart_all_channels(struct gk20a *g);
+
+bool channel_gk20a_is_prealloc_enabled(struct channel_gk20a *c);
+void channel_gk20a_joblist_lock(struct channel_gk20a *c);
+void channel_gk20a_joblist_unlock(struct channel_gk20a *c);
+bool channel_gk20a_joblist_is_empty(struct channel_gk20a *c);
+
+int channel_gk20a_update_runlist(struct channel_gk20a *c, bool add);
+int gk20a_channel_get_timescale_from_timeslice(struct gk20a *g,
+                unsigned int timeslice_period,
+                unsigned int *__timeslice_timeout, unsigned int *__timeslice_scale);
+
+void gk20a_wait_until_counter_is_N(
+        struct channel_gk20a *ch, nvgpu_atomic_t *counter, int wait_value,
+        struct nvgpu_cond *c, const char *caller, const char *counter_name);
+int channel_gk20a_alloc_job(struct channel_gk20a *c,
+                struct channel_gk20a_job **job_out);
+void channel_gk20a_free_job(struct channel_gk20a *c,
+                struct channel_gk20a_job *job);
+u32 nvgpu_get_gp_free_count(struct channel_gk20a *c);
+u32 nvgpu_gp_free_count(struct channel_gk20a *c);
+int gk20a_channel_add_job(struct channel_gk20a *c,
+                                 struct channel_gk20a_job *job,
+                                 bool skip_buffer_refcounting);
+void free_priv_cmdbuf(struct channel_gk20a *c,
+                             struct priv_cmd_entry *e);
+void gk20a_channel_clean_up_jobs(struct channel_gk20a *c,
+                                        bool clean_all);
+
+void gk20a_channel_free_usermode_buffers(struct channel_gk20a *c);
+u32 nvgpu_get_gpfifo_entry_size(void);
 
 int nvgpu_submit_channel_gpfifo_user(struct channel_gk20a *c,
                                 struct nvgpu_gpfifo_userdata userdata,
@@ -47,9 +452,9 @@ int nvgpu_submit_channel_gpfifo_kernel(struct channel_gk20a *c,
                                 struct gk20a_fence **fence_out);
 
 #ifdef CONFIG_DEBUG_FS
-void trace_write_pushbuffers(struct channel_gk20a *c, int count);
+void trace_write_pushbuffers(struct channel_gk20a *c, u32 count);
 #else
-static inline void trace_write_pushbuffers(struct channel_gk20a *c, int count)
+static inline void trace_write_pushbuffers(struct channel_gk20a *c, u32 count)
 {
 }
 #endif
diff --git a/drivers/gpu/nvgpu/os/linux/cde.c b/drivers/gpu/nvgpu/os/linux/cde.c
index 30cd0bff..7aac8e01 100644
--- a/drivers/gpu/nvgpu/os/linux/cde.c
+++ b/drivers/gpu/nvgpu/os/linux/cde.c
@@ -38,7 +38,6 @@
 #include <nvgpu/linux/vm.h>
 
 #include "gk20a/gk20a.h"
-#include "gk20a/channel_gk20a.h"
 #include "gk20a/mm_gk20a.h"
 #include "gk20a/fence_gk20a.h"
 #include "gk20a/gr_gk20a.h"
diff --git a/drivers/gpu/nvgpu/os/linux/ctxsw_trace.c b/drivers/gpu/nvgpu/os/linux/ctxsw_trace.c
index 16b040da..aa5ed250 100644
--- a/drivers/gpu/nvgpu/os/linux/ctxsw_trace.c
+++ b/drivers/gpu/nvgpu/os/linux/ctxsw_trace.c
@@ -21,6 +21,7 @@
 #include <trace/events/gk20a.h>
 #include <uapi/linux/nvgpu.h>
 #include <nvgpu/ctxsw_trace.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/gr_gk20a.h"
diff --git a/drivers/gpu/nvgpu/os/linux/debug_fifo.c b/drivers/gpu/nvgpu/os/linux/debug_fifo.c
index 2b5674c0..0c791a42 100644
--- a/drivers/gpu/nvgpu/os/linux/debug_fifo.c
+++ b/drivers/gpu/nvgpu/os/linux/debug_fifo.c
@@ -20,6 +20,7 @@
 
 #include <nvgpu/sort.h>
 #include <nvgpu/timers.h>
+#include <nvgpu/channel.h>
 
 void __gk20a_fifo_profile_free(struct nvgpu_ref *ref);
 
diff --git a/drivers/gpu/nvgpu/os/linux/ioctl_as.c b/drivers/gpu/nvgpu/os/linux/ioctl_as.c
index 7d1d618e..4e479e8e 100644
--- a/drivers/gpu/nvgpu/os/linux/ioctl_as.c
+++ b/drivers/gpu/nvgpu/os/linux/ioctl_as.c
@@ -24,6 +24,7 @@
 #include <nvgpu/gmmu.h>
 #include <nvgpu/vm_area.h>
 #include <nvgpu/log2.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/linux/vm.h>
 
diff --git a/drivers/gpu/nvgpu/os/linux/ioctl_ctrl.c b/drivers/gpu/nvgpu/os/linux/ioctl_ctrl.c
index 2f013029..19b4286d 100644
--- a/drivers/gpu/nvgpu/os/linux/ioctl_ctrl.c
+++ b/drivers/gpu/nvgpu/os/linux/ioctl_ctrl.c
@@ -31,6 +31,7 @@
 #include <nvgpu/sizes.h>
 #include <nvgpu/list.h>
 #include <nvgpu/clk_arb.h>
+#include <nvgpu/channel.h>
 
 #include "ioctl_ctrl.h"
 #include "ioctl_dbg.h"
diff --git a/drivers/gpu/nvgpu/os/linux/ioctl_dbg.c b/drivers/gpu/nvgpu/os/linux/ioctl_dbg.c
index 3931ab12..e4e7394e 100644
--- a/drivers/gpu/nvgpu/os/linux/ioctl_dbg.c
+++ b/drivers/gpu/nvgpu/os/linux/ioctl_dbg.c
@@ -30,6 +30,7 @@
 #include <nvgpu/atomic.h>
 #include <nvgpu/cond.h>
 #include <nvgpu/utils.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/linux/vm.h>
 
diff --git a/drivers/gpu/nvgpu/os/linux/ioctl_tsg.c b/drivers/gpu/nvgpu/os/linux/ioctl_tsg.c
index 6c68ca58..6dd96d6b 100644
--- a/drivers/gpu/nvgpu/os/linux/ioctl_tsg.c
+++ b/drivers/gpu/nvgpu/os/linux/ioctl_tsg.c
@@ -25,6 +25,7 @@
 #include <nvgpu/kmem.h>
 #include <nvgpu/log.h>
 #include <nvgpu/os_sched.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "gk20a/tsg_gk20a.h"
diff --git a/drivers/gpu/nvgpu/os/linux/linux-channel.c b/drivers/gpu/nvgpu/os/linux/linux-channel.c
index fef44f2b..77c14b56 100644
--- a/drivers/gpu/nvgpu/os/linux/linux-channel.c
+++ b/drivers/gpu/nvgpu/os/linux/linux-channel.c
@@ -18,6 +18,7 @@
 #include <nvgpu/debug.h>
 #include <nvgpu/error_notifier.h>
 #include <nvgpu/os_sched.h>
+#include <nvgpu/channel.h>
 
 /*
  * This is required for nvgpu_vm_find_buf() which is used in the tracing
diff --git a/drivers/gpu/nvgpu/os/linux/module.c b/drivers/gpu/nvgpu/os/linux/module.c
index 55f55f00..85439b88 100644
--- a/drivers/gpu/nvgpu/os/linux/module.c
+++ b/drivers/gpu/nvgpu/os/linux/module.c
@@ -44,6 +44,7 @@
 #include <nvgpu/sim.h>
 #include <nvgpu/clk_arb.h>
 #include <nvgpu/timers.h>
+#include <nvgpu/channel.h>
 
 #include "platform_gk20a.h"
 #include "sysfs.h"
diff --git a/drivers/gpu/nvgpu/os/linux/os_fence_android.c b/drivers/gpu/nvgpu/os/linux/os_fence_android.c
index 9be8c6c0..297b3b20 100644
--- a/drivers/gpu/nvgpu/os/linux/os_fence_android.c
+++ b/drivers/gpu/nvgpu/os/linux/os_fence_android.c
@@ -16,6 +16,7 @@
 #include <nvgpu/types.h>
 #include <nvgpu/os_fence.h>
 #include <nvgpu/linux/os_fence_android.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 
diff --git a/drivers/gpu/nvgpu/os/linux/os_fence_android_sema.c b/drivers/gpu/nvgpu/os/linux/os_fence_android_sema.c
index e7070612..ec3ccf0d 100644
--- a/drivers/gpu/nvgpu/os/linux/os_fence_android_sema.c
+++ b/drivers/gpu/nvgpu/os/linux/os_fence_android_sema.c
@@ -20,6 +20,7 @@
 #include <nvgpu/os_fence.h>
 #include <nvgpu/linux/os_fence_android.h>
 #include <nvgpu/semaphore.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/channel_sync_gk20a.h"
 #include "gk20a/mm_gk20a.h"
diff --git a/drivers/gpu/nvgpu/os/linux/os_fence_android_syncpt.c b/drivers/gpu/nvgpu/os/linux/os_fence_android_syncpt.c
index 103ad243..b15dba19 100644
--- a/drivers/gpu/nvgpu/os/linux/os_fence_android_syncpt.c
+++ b/drivers/gpu/nvgpu/os/linux/os_fence_android_syncpt.c
@@ -22,9 +22,9 @@
 #include <nvgpu/linux/os_fence_android.h>
 #include <nvgpu/nvhost.h>
 #include <nvgpu/atomic.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
-#include "gk20a/channel_gk20a.h"
 #include "gk20a/channel_sync_gk20a.h"
 #include "gk20a/mm_gk20a.h"
 
@@ -120,4 +120,4 @@ int nvgpu_os_fence_syncpt_fdget(struct nvgpu_os_fence *fence_out,
         nvgpu_os_fence_init(fence_out, c->g, &syncpt_ops, fence);
 
         return 0;
-}
\ No newline at end of file
+}
diff --git a/drivers/gpu/nvgpu/os/linux/sync_sema_android.c b/drivers/gpu/nvgpu/os/linux/sync_sema_android.c
index 9598f2df..50465d0c 100644
--- a/drivers/gpu/nvgpu/os/linux/sync_sema_android.c
+++ b/drivers/gpu/nvgpu/os/linux/sync_sema_android.c
@@ -25,6 +25,7 @@
 #include <nvgpu/semaphore.h>
 #include <nvgpu/bug.h>
 #include <nvgpu/kref.h>
+#include <nvgpu/channel.h>
 #include "../linux/channel.h"
 
 #include "../drivers/staging/android/sync.h"
diff --git a/drivers/gpu/nvgpu/os/linux/vgpu/vgpu_linux.c b/drivers/gpu/nvgpu/os/linux/vgpu/vgpu_linux.c
index 33c067c6..5b7a5c07 100644
--- a/drivers/gpu/nvgpu/os/linux/vgpu/vgpu_linux.c
+++ b/drivers/gpu/nvgpu/os/linux/vgpu/vgpu_linux.c
@@ -31,12 +31,12 @@
 #include <nvgpu/ctxsw_trace.h>
 #include <nvgpu/defaults.h>
 #include <nvgpu/ltc.h>
+#include <nvgpu/channel.h>
 
 #include "vgpu_linux.h"
 #include "vgpu/fecs_trace_vgpu.h"
 #include "clk_vgpu.h"
 #include "gk20a/tsg_gk20a.h"
-#include "gk20a/channel_gk20a.h"
 #include "gk20a/regops_gk20a.h"
 #include "gm20b/hal_gm20b.h"
 
diff --git a/drivers/gpu/nvgpu/os/posix/posix-channel.c b/drivers/gpu/nvgpu/os/posix/posix-channel.c
index 05697159..5e10b1ee 100644
--- a/drivers/gpu/nvgpu/os/posix/posix-channel.c
+++ b/drivers/gpu/nvgpu/os/posix/posix-channel.c
@@ -20,7 +20,7 @@
  * DEALINGS IN THE SOFTWARE.
  */
 
-#include "gk20a/channel_gk20a.h"
+#include <nvgpu/channel.h>
 
 u32 nvgpu_get_gpfifo_entry_size(void)
 {
diff --git a/drivers/gpu/nvgpu/vgpu/css_vgpu.c b/drivers/gpu/nvgpu/vgpu/css_vgpu.c
index 40ff8f90..1c97d03a 100644
--- a/drivers/gpu/nvgpu/vgpu/css_vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/css_vgpu.c
@@ -26,9 +26,9 @@
 #include <nvgpu/vgpu/tegra_vgpu.h>
 #include <nvgpu/dt.h>
 #include <nvgpu/bug.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
-#include "gk20a/channel_gk20a.h"
 #include "gk20a/css_gr_gk20a.h"
 
 #include "vgpu/css_vgpu.h"
diff --git a/drivers/gpu/nvgpu/vgpu/dbg_vgpu.c b/drivers/gpu/nvgpu/vgpu/dbg_vgpu.c
index 2bb3b205..d188f540 100644
--- a/drivers/gpu/nvgpu/vgpu/dbg_vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/dbg_vgpu.c
@@ -24,9 +24,9 @@
 #include <nvgpu/vgpu/tegra_vgpu.h>
 #include <nvgpu/vgpu/vgpu.h>
 #include <nvgpu/bug.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
-#include "gk20a/channel_gk20a.h"
 #include "gk20a/dbg_gpu_gk20a.h"
 #include "gk20a/regops_gk20a.h"
 #include "dbg_vgpu.h"
diff --git a/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c b/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c
index 98c7b9ce..aa5abec9 100644
--- a/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/fifo_vgpu.c
@@ -33,6 +33,7 @@
 #include <nvgpu/error_notifier.h>
 #include <nvgpu/vgpu/vgpu_ivc.h>
 #include <nvgpu/vgpu/vgpu.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "fifo_vgpu.h"
diff --git a/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_gr_gp10b.c b/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_gr_gp10b.c
index 86184336..8110902f 100644
--- a/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_gr_gp10b.c
+++ b/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_gr_gp10b.c
@@ -24,6 +24,7 @@
 #include <nvgpu/dma.h>
 #include <nvgpu/bug.h>
 #include <nvgpu/vgpu/vgpu.h>
+#include <nvgpu/channel.h>
 
 #include "vgpu/gm20b/vgpu_gr_gm20b.h"
 
diff --git a/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_hal_gp10b.c b/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_hal_gp10b.c
index 7500e92e..0b6a5cb5 100644
--- a/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_hal_gp10b.c
+++ b/drivers/gpu/nvgpu/vgpu/gp10b/vgpu_hal_gp10b.c
@@ -71,6 +71,7 @@
 #include <nvgpu/enabled.h>
 #include <nvgpu/vgpu/vgpu.h>
 #include <nvgpu/error_notifier.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/hw/gp10b/hw_fifo_gp10b.h>
 #include <nvgpu/hw/gp10b/hw_ram_gp10b.h>
diff --git a/drivers/gpu/nvgpu/vgpu/gr_vgpu.c b/drivers/gpu/nvgpu/vgpu/gr_vgpu.c
index 9ee57fb4..4b1cf212 100644
--- a/drivers/gpu/nvgpu/vgpu/gr_vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/gr_vgpu.c
@@ -29,11 +29,11 @@
 #include <nvgpu/dma.h>
 #include <nvgpu/vgpu/vgpu_ivc.h>
 #include <nvgpu/vgpu/vgpu.h>
+#include <nvgpu/channel.h>
 
 #include "gr_vgpu.h"
 #include "gk20a/gk20a.h"
 #include "gk20a/dbg_gpu_gk20a.h"
-#include "gk20a/channel_gk20a.h"
 #include "gk20a/tsg_gk20a.h"
 #include "gk20a/fecs_trace_gk20a.h"
 
diff --git a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_fifo_gv11b.c b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_fifo_gv11b.c
index 43cff1c0..3042ad12 100644
--- a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_fifo_gv11b.c
+++ b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_fifo_gv11b.c
@@ -25,6 +25,7 @@
 #include <nvgpu/vgpu/vgpu.h>
 #include <nvgpu/nvhost.h>
 #include <nvgpu/vgpu/tegra_vgpu.h>
+#include <nvgpu/channel.h>
 
 #include "gv11b/fifo_gv11b.h"
 #include "vgpu_fifo_gv11b.h"
diff --git a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_hal_gv11b.c b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_hal_gv11b.c
index fb118f16..85835cee 100644
--- a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_hal_gv11b.c
+++ b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_hal_gv11b.c
@@ -85,6 +85,7 @@
 #include <gv11b/gr_gv11b.h>
 
 #include <nvgpu/enabled.h>
+#include <nvgpu/channel.h>
 
 #include "vgpu_gv11b.h"
 #include "vgpu_gr_gv11b.h"
diff --git a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_subctx_gv11b.c b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_subctx_gv11b.c
index b5272ae1..14d10847 100644
--- a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_subctx_gv11b.c
+++ b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_subctx_gv11b.c
@@ -25,6 +25,8 @@
 
 #include <nvgpu/vgpu/vgpu.h>
 #include <nvgpu/vgpu/tegra_vgpu.h>
+#include <nvgpu/channel.h>
+
 #include <nvgpu/hw/gv11b/hw_ctxsw_prog_gv11b.h>
 
 int vgpu_gv11b_alloc_subctx_header(struct channel_gk20a *c)
diff --git a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_tsg_gv11b.c b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_tsg_gv11b.c
index e42bea1f..8c999161 100644
--- a/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_tsg_gv11b.c
+++ b/drivers/gpu/nvgpu/vgpu/gv11b/vgpu_tsg_gv11b.c
@@ -22,6 +22,7 @@
 
 #include <nvgpu/vgpu/tegra_vgpu.h>
 #include <nvgpu/vgpu/vgpu.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "vgpu_tsg_gv11b.h"
diff --git a/drivers/gpu/nvgpu/vgpu/mm_vgpu.c b/drivers/gpu/nvgpu/vgpu/mm_vgpu.c
index 229a9767..837508bb 100644
--- a/drivers/gpu/nvgpu/vgpu/mm_vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/mm_vgpu.c
@@ -27,6 +27,7 @@
 #include <nvgpu/bug.h>
 #include <nvgpu/vm.h>
 #include <nvgpu/vm_area.h>
+#include <nvgpu/channel.h>
 
 #include <nvgpu/vgpu/vm.h>
 #include <nvgpu/vgpu/vgpu.h>
diff --git a/drivers/gpu/nvgpu/vgpu/tsg_vgpu.c b/drivers/gpu/nvgpu/vgpu/tsg_vgpu.c
index 7bb8f671..a81b5022 100644
--- a/drivers/gpu/nvgpu/vgpu/tsg_vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/tsg_vgpu.c
@@ -21,9 +21,9 @@
  */
 
 #include "gk20a/gk20a.h"
-#include "gk20a/channel_gk20a.h"
 #include "gk20a/tsg_gk20a.h"
 #include "fifo_vgpu.h"
+#include <nvgpu/channel.h>
 
 #include <nvgpu/bug.h>
 #include <nvgpu/vgpu/tegra_vgpu.h>
diff --git a/drivers/gpu/nvgpu/vgpu/vgpu.c b/drivers/gpu/nvgpu/vgpu/vgpu.c
index 8e20e4fc..bcd352e7 100644
--- a/drivers/gpu/nvgpu/vgpu/vgpu.c
+++ b/drivers/gpu/nvgpu/vgpu/vgpu.c
@@ -25,6 +25,7 @@
 #include <nvgpu/vgpu/vgpu_ivc.h>
 #include <nvgpu/vgpu/vgpu.h>
 #include <nvgpu/timers.h>
+#include <nvgpu/channel.h>
 
 #include "gk20a/gk20a.h"
 #include "fecs_trace_vgpu.h"