Merge tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-tracegrafted

Pull tracing fixes from Steven Rostedt:

 - Don't destroy user event fields when removal fails

   User event fields are destroyed before the event is removed from
   visibility. But that can fail leaving the still visible event with no
   fields. Move the destroying of the fields to after the event is
   successfully removed from visibility.

 - Initialize function graph state is fork before calling
   copy_exec_state()

   For non-CLONE_VM forks, copy_exec_state() allocates a new
   task_exec_state. If that allocation fails, ftrace_graph_exit_task()
   will free the tasks ret_stack pointer. Since that pointer is still
   using the parent's ret_stack, it mistakenly frees the parent's
   pointer too.

   Call ftrace_graph_init() on the task first which will NULL out the
   new tasks's ret_stack and if the copy fails, it will not free
   anything.

 - Remove FGRAPH_MAX_INDEX

   The macro FGRAPH_MAX_INDEX was added but never used. Remove it.

 - Save ent_size in function graph printing of nested functions

   The function graph tracer needs to look at the next event to see if
   the next event is the return of the current function entry. If it is,
   it prints a single line:

	ktime_get();

   Otherwise it prints it like a nested function:

	tick_nohz_irq_exit() {
	    ktime_get();
	    kcpustat_irq_exit();
	}

   In order to look at the next event, it must save the current event so
   that it has the information to print from it. It saves the event in
   the iterator descriptor called "ent". What it doesn't save is the
   ent_size of the event which is now used to know if the function graph
   arguments are to be printed. The peek doesn't save the size so the
   size used happens to be that of the size of the last event that was
   seen.

   Save the entry event size in the iterator descriptor so that the
   correct size is used.

 - Fix several errors with freeing data in the histogram code

   The histogram code had a lot of leaked or or incorrect accounting
   when failures happen. Correct them.

 - Fix histogram regression of .percent and .graph modifiers

   Up until 6.3 histogram values could have "percent" or "graph"
   modifiers that changed how they were printed. But a change that added
   restricting histograms values from being strings, stack traces and
   other modifiers inadvertently prevented them from using the percent
   and graph modifiers, which were legal use cases for values.

   Put back the percent and graph modifiers.

 - Fix various typos in the comments

 - Set the trace_clock before initializing a histogram with clock
   argument

   The histogram API allows the user to specific which trace clock to
   use via a "clock=" string. The histogram is set up first before the
   clock is checked. If the passed in clock is not valid, it exits
   without fully fixing up the histogram leaving it on the list and a
   use-after-free can trigger.

   Update the clock argument first and if it fails then exit gracefully
   before the histogram trigger is placed on any lists.

 - Restore :mod: trailer after parsing in ftrace_set_clr_event

   The function ftrace_set_clr_event() modifies the parse string and
   needs to put it back to what was passed in. It searches for ":mod:"
   via a strsep() but fails to put back the first ':' in the string.

   Add back the ':' in the passed in string.

 - Take trace_array reference when opening a tracer options file

   The options files are dynamically created and some tracers add their
   own options. When a tracer adds their own list of options, the
   trace_array holding them has an array to hold the list of options for
   each tracer. This array increases in size via a krealloc(), and the
   new entry gets a newly allocated array to hold the options of the new
   tracer being added.

   The element in each entry of the tracer's option array holds a
   pointer back to the trace_array, a pointer to the tracer it is
   associated to, a pointer to the flags of the option.

   The issue is that these arrays are freed when the trace_array is
   freed when its instance it represents is removed from the instances
   directory. There's a race that an open of one of these options files
   can happen when the instance is being removed.

   Add a new helper function to be called by the open function of the
   options file to iterate all existing trace_arrays under a lock and
   find the one that has the given option element in one of it's tracer
   arrays. If found, then update the associated trace_array's reference
   counter to keep it from being freed. If not found, have the open call
   return -ENODEV.

 - Disable interrupts when acquiring the lock in rb_wake_up_waiters()

   The function rb_wake_up_waiters() assumes it will be called in
   interrupt context and does not disable irqs when taking
   cpu_buffer->reader_lock, which can be called in hard interrupt
   context. The issue is in PREEMPT_RT, this function is called in
   thread context leaving this lock open to a deadlock.

   Take the lock with interrupts disabled.

 - Use rcu_assign_pointer() for tmp_ops filter hash

   The tmp_ops used in update_ftrace_direct_mod() assigns its
   filter_hash field directly, but that field is annotated as __rcu and
   sparse complains. Assign it with rcu_assign_pointer()

 - Fix use-after-free in enable_trigger_private_data_free()

   The trace_event_call is accessed through the event_trigger_data's
   trace_event_file pointer to put the trace_event_call on freeing. The
   issue is that the trace_event_file data may have been freed already
   causing a use-after-free. Add a field to the event_trigger_data that
   points directly to the trace_event_call so that it can decrement its
   reference directly without needing to go through the
   trace_event_file.

 - Fix accounting of buffer data remote headers

   trace_buffer_desc_size() and trace_remote_alloc_buffer() undercount
   the number of pages is needed for the asked for size as it doesn't
   take into account the meta data on each page. Add a helper function
   to do the calculation properly and use that in these functions.

 - Catch nr_page_va overflow in ring_buffer_desc sizing

   The number of pages per remote ring buffer is capped by
   ring_buffer_desc::nr_page_va (32 bits). A buffer_size large enough to
   overflow that field would silently allocate a descriptor smaller than
   what was asked for.

 - Do not resize the subbuf order if any per_cpu buffer is disabled

   The mmapping of ring buffers disables resizing the subbuffers, but it
   is done per-cpu whereas the subbuf size change is done for all the
   per_cpu buffers under the buffer->mutex. It could change the size of
   some while the mapping is happening on others. Have the resize of the
   subbuf order check all the per_cpu buffers under the lock to see if
   any of them is disabled before starting and causing an inconsistency
   between buffers that are being mapped.

* tag 'trace-v7.3-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace: (25 commits)
  ring-buffer: Check resize_disabled before publishing the new subbuf order
  tracing/remotes: Catch nr_page_va overflow in ring_buffer_desc sizing
  tracing/remotes: Account for ring buffer page header in size calculation
  tracing: Don't dereference trace_event_file in deferred trigger free
  ftrace: Use rcu_assign_pointer() for tmp_ops filter hash
  ring-buffer: Acquire the lock with irqsave in rb_wake_up_waiters()
  tracing: Take trace_array reference when opening a tracer options file
  tracing: Fix ring_buffer_read_page_size() kernel-doc
  tracing: Restore :mod: trailer after parsing in ftrace_set_clr_event()
  tracing: Fix memory corruption from a "STACKTRACE" histogram key
  tracing: Fix memory corruption from the histogram stacktrace modifier
  tracing: Undo the registration when enabling the histogram trigger fails
  tracing: Take the reference before publishing the named histogram trigger
  tracing: Set the trace clock before registering the histogram trigger
  tracing: Fix typo "preceeded" in comment
  tracing: Fix typo "availabe" in comment
  tracing: Let histogram values keep the percent and graph modifiers
  tracing: Keep the entry count when the histogram stats allocation fails
  tracing: Free histogram the field rejected for a bad modifier
  tracing: Free histogram the var ref when its initialization fails
  ...

1 files changed, 205 insertions, 0 deletions

diff --git a/Documentation/gpu/amdgpu/userq.rst b/Documentation/gpu/amdgpu/userq.rst
new file mode 100644
index 000000000..88f54393b
--- /dev/null
+++ b/Documentation/gpu/amdgpu/userq.rst
@@ -0,0 +1,205 @@
+.. _amdgpu-userq:
+
+==================
+ User Mode Queues
+==================
+
+Introduction
+============
+
+Similar to the KFD, GPU engine queues move into userspace.  The idea is to let
+user processes manage their submissions to the GPU engines directly, bypassing
+IOCTL calls to the driver to submit work.  This reduces overhead and also allows
+the GPU to submit work to itself.  Applications can set up work graphs of jobs
+across multiple GPU engines without needing trips through the CPU.
+
+UMDs directly interface with firmware via per application shared memory areas.
+The main vehicle for this is queue.  A queue is a ring buffer with a read
+pointer (rptr) and a write pointer (wptr).  The UMD writes IP specific packets
+into the queue and the firmware processes those packets, kicking off work on the
+GPU engines.  The CPU in the application (or another queue or device) updates
+the wptr to tell the firmware how far into the ring buffer to process packets
+and the rtpr provides feedback to the UMD on how far the firmware has progressed
+in executing those packets.  When the wptr and the rptr are equal, the queue is
+idle.
+
+Theory of Operation
+===================
+
+The various engines on modern AMD GPUs support multiple queues per engine with a
+scheduling firmware which handles dynamically scheduling user queues on the
+available hardware queue slots.  When the number of user queues outnumbers the
+available hardware queue slots, the scheduling firmware dynamically maps and
+unmaps queues based on priority and time quanta.  The state of each user queue
+is managed in the kernel driver in an MQD (Memory Queue Descriptor).  This is a
+buffer in GPU accessible memory that stores the state of a user queue.  The
+scheduling firmware uses the MQD to load the queue state into an HQD (Hardware
+Queue Descriptor) when a user queue is mapped.  Each user queue requires a
+number of additional buffers which represent the ring buffer and any metadata
+needed by the engine for runtime operation.  On most engines this consists of
+the ring buffer itself, a rptr buffer (where the firmware will shadow the rptr
+to userspace), a wptr buffer (where the application will write the wptr for the
+firmware to fetch it), and a doorbell.  A doorbell is a piece of one of the
+device's MMIO BARs which can be mapped to specific user queues.  When the
+application writes to the doorbell, it will signal the firmware to take some
+action. Writing to the doorbell wakes the firmware and causes it to fetch the
+wptr and start processing the packets in the queue. Each 4K page of the doorbell
+BAR supports specific offset ranges for specific engines.  The doorbell of a
+queue must be mapped into the aperture aligned to the IP used by the queue
+(e.g., GFX, VCN, SDMA, etc.).  These doorbell apertures are set up via NBIO
+registers.  Doorbells are 32 bit or 64 bit (depending on the engine) chunks of
+the doorbell BAR.  A 4K doorbell page provides 512 64-bit doorbells for up to
+512 user queues.  A subset of each page is reserved for each IP type supported
+on the device.  The user can query the doorbell ranges for each IP via the INFO
+IOCTL.  See the IOCTL Interfaces section for more information.
+
+When an application wants to create a user queue, it allocates the necessary
+buffers for the queue (ring buffer, wptr and rptr, context save areas, etc.).
+These can be separate buffers or all part of one larger buffer.  The application
+would map the buffer(s) into its GPUVM and use the GPU virtual addresses of for
+the areas of memory they want to use for the user queue.  They would also
+allocate a doorbell page for the doorbells used by the user queues.  The
+application would then populate the MQD in the USERQ IOCTL structure with the
+GPU virtual addresses and doorbell index they want to use.  The user can also
+specify the attributes for the user queue (priority, whether the queue is secure
+for protected content, etc.).  The application would then call the USERQ
+CREATE IOCTL to create the queue using the specified MQD details in the IOCTL.
+The kernel driver then validates the MQD provided by the application and
+translates the MQD into the engine specific MQD format for the IP.  The IP
+specific MQD would be allocated and the queue would be added to the run list
+maintained by the scheduling firmware.  Once the queue has been created, the
+application can write packets directly into the queue, update the wptr, and
+write to the doorbell offset to kick off work in the user queue.
+
+When the application is done with the user queue, it would call the USERQ
+FREE IOCTL to destroy it.  The kernel driver would preempt the queue and
+remove it from the scheduling firmware's run list.  Then the IP specific MQD
+would be freed and the user queue state would be cleaned up.
+
+Some engines may require the aggregated doorbell too if the engine does not
+support doorbells from unmapped queues.  The aggregated doorbell is a special
+page of doorbell space which wakes the scheduler.  In cases where the engine may
+be oversubscribed, some queues may not be mapped.  If the doorbell is rung when
+the queue is not mapped, the engine firmware may miss the request.  Some
+scheduling firmware may work around this by polling wptr shadows when the
+hardware is oversubscribed, other engines may support doorbell updates from
+unmapped queues.  In the event that one of these options is not available, the
+kernel driver will map a page of aggregated doorbell space into each GPUVM
+space.  The UMD will then update the doorbell and wptr as normal and then write
+to the aggregated doorbell as well.
+
+Special Packets
+---------------
+
+In order to support legacy implicit synchronization, as well as mixed user and
+kernel queues, we need a synchronization mechanism that is secure.  Because
+kernel queues or memory management tasks depend on kernel fences, we need a way
+for user queues to update memory that the kernel can use for a fence, that can't
+be messed with by a bad actor.  To support this, we've added a protected fence
+packet.  This packet works by writing a monotonically increasing value to
+a memory location that only privileged clients have write access to. User
+queues only have read access.  When this packet is executed, the memory location
+is updated and other queues (kernel or user) can see the results.  The
+user application would submit this packet in their command stream.  The actual
+packet format varies from IP to IP (GFX/Compute, SDMA, VCN, etc.), but the
+behavior is the same.  The packet submission is handled in userspace.  The
+kernel driver sets up the privileged memory used for each user queue when it
+sets the queues up when the application creates them.
+
+
+Memory Management
+=================
+
+It is assumed that all buffers mapped into the GPUVM space for the process are
+valid when engines on the GPU are running.  The kernel driver will only allow
+user queues to run when all buffers are mapped.  If there is a memory event that
+requires buffer migration, the kernel driver will preempt the user queues,
+migrate buffers to where they need to be, update the GPUVM page tables and
+invaldidate the TLB, and then resume the user queues.
+
+Interaction with Kernel Queues
+==============================
+
+Depending on the IP and the scheduling firmware, you can enable kernel queues
+and user queues at the same time, however, you are limited by the HQD slots.
+Kernel queues are always mapped so any work that goes into kernel queues will
+take priority.  This limits the available HQD slots for user queues.
+
+Not all IPs will support user queues on all GPUs.  As such, UMDs will need to
+support both user queues and kernel queues depending on the IP.  For example, a
+GPU may support user queues for GFX, compute, and SDMA, but not for VCN, JPEG,
+and VPE.  UMDs need to support both.  The kernel driver provides a way to
+determine if user queues and kernel queues are supported on a per IP basis.
+UMDs can query this information via the INFO IOCTL and determine whether to use
+kernel queues or user queues for each IP.
+
+Queue Resets
+============
+
+For most engines, queues can be reset individually.  GFX, compute, and SDMA
+queues can be reset individually.  When a hung queue is detected, it can be
+reset either via the scheduling firmware or MMIO.  Since there are no kernel
+fences for most user queues, they will usually only be detected when some other
+event happens; e.g., a memory event which requires migration of buffers.  When
+the queues are preempted, if the queue is hung, the preemption will fail.
+Driver will then look up the queues that failed to preempt and reset them and
+record which queues are hung.
+
+On the UMD side, we will add a USERQ QUERY_STATUS IOCTL to query the queue
+status.  UMD will provide the queue id in the IOCTL and the kernel driver
+will check if it has already recorded the queue as hung (e.g., due to failed
+peemption) and report back the status.
+
+IOCTL Interfaces
+================
+
+GPU virtual addresses used for queues and related data (rptrs, wptrs, context
+save areas, etc.) should be validated by the kernel mode driver to prevent the
+user from specifying invalid GPU virtual addresses.  If the user provides
+invalid GPU virtual addresses or doorbell indicies, the IOCTL should return an
+error message.  These buffers should also be tracked in the kernel driver so
+that if the user attempts to unmap the buffer(s) from the GPUVM, the umap call
+would return an error.
+
+INFO
+----
+There are several new INFO queries related to user queues in order to query the
+size of user queue meta data needed for a user queue (e.g., context save areas
+or shadow buffers), whether kernel or user queues or both are supported
+for each IP type, and the offsets for each IP type in each doorbell page.
+
+USERQ
+-----
+The USERQ IOCTL is used for creating, freeing, and querying the status of user
+queues.  It supports 3 opcodes:
+
+1. CREATE - Create a user queue.  The application provides an MQD-like structure
+   that defines the type of queue and associated metadata and flags for that
+   queue type.  Returns the queue id.
+2. FREE - Free a user queue.
+3. QUERY_STATUS - Query that status of a queue.  Used to check if the queue is
+   healthy or not.  E.g., if the queue has been reset. (WIP)
+
+USERQ_SIGNAL
+------------
+The USERQ_SIGNAL IOCTL is used to provide a list of sync objects to be signaled.
+
+USERQ_WAIT
+----------
+The USERQ_WAIT IOCTL is used to provide a list of sync object to be waited on.
+
+Kernel and User Queues
+======================
+
+In order to properly validate and test performance, we have a driver option to
+select what type of queues are enabled (kernel queues, user queues or both).
+The user_queue driver parameter allows you to enable kernel queues only (0),
+user queues and kernel queues (1), and user queues only (2).  Enabling user
+queues only will free up static queue assignments that would otherwise be used
+by kernel queues for use by the scheduling firmware.  Some kernel queues are
+required for kernel driver operation and they will always be created.  When the
+kernel queues are not enabled, they are not registered with the drm scheduler
+and the CS IOCTL will reject any incoming command submissions which target those
+queue types.  Kernel queues only mirrors the behavior on all existing GPUs.
+Enabling both queues allows for backwards compatibility with old userspace while
+still supporting user queues.