OSDN Git Service

drm/i915: Kill the undead i915_gem_batch_pool.c
authorChris Wilson <chris@chris-wilson.co.uk>
Thu, 22 Aug 2019 06:59:17 +0000 (07:59 +0100)
committerChris Wilson <chris@chris-wilson.co.uk>
Thu, 22 Aug 2019 11:16:10 +0000 (12:16 +0100)
You have to cut it off at the neck, otherwise it just reappears in the
next merge, like it did in commit 3f866026f0ce ("Merge drm/drm-next
into drm-intel-next-queued")

References: 3f866026f0ce ("Merge drm/drm-next into drm-intel-next-queued")
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Rodrigo Vivi <rodrigo.vivi@intel.com>
Acked-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190822065917.18988-1-chris@chris-wilson.co.uk
drivers/gpu/drm/i915/i915_gem_batch_pool.c [deleted file]

diff --git a/drivers/gpu/drm/i915/i915_gem_batch_pool.c b/drivers/gpu/drm/i915/i915_gem_batch_pool.c
deleted file mode 100644 (file)
index 8675a60..0000000
+++ /dev/null
@@ -1,132 +0,0 @@
-/*
- * SPDX-License-Identifier: MIT
- *
- * Copyright © 2014-2018 Intel Corporation
- */
-
-#include "i915_gem_batch_pool.h"
-#include "i915_drv.h"
-
-/**
- * DOC: batch pool
- *
- * In order to submit batch buffers as 'secure', the software command parser
- * must ensure that a batch buffer cannot be modified after parsing. It does
- * this by copying the user provided batch buffer contents to a kernel owned
- * buffer from which the hardware will actually execute, and by carefully
- * managing the address space bindings for such buffers.
- *
- * The batch pool framework provides a mechanism for the driver to manage a
- * set of scratch buffers to use for this purpose. The framework can be
- * extended to support other uses cases should they arise.
- */
-
-/**
- * i915_gem_batch_pool_init() - initialize a batch buffer pool
- * @pool: the batch buffer pool
- * @engine: the associated request submission engine
- */
-void i915_gem_batch_pool_init(struct i915_gem_batch_pool *pool,
-                             struct intel_engine_cs *engine)
-{
-       int n;
-
-       pool->engine = engine;
-
-       for (n = 0; n < ARRAY_SIZE(pool->cache_list); n++)
-               INIT_LIST_HEAD(&pool->cache_list[n]);
-}
-
-/**
- * i915_gem_batch_pool_fini() - clean up a batch buffer pool
- * @pool: the pool to clean up
- *
- * Note: Callers must hold the struct_mutex.
- */
-void i915_gem_batch_pool_fini(struct i915_gem_batch_pool *pool)
-{
-       int n;
-
-       lockdep_assert_held(&pool->engine->i915->drm.struct_mutex);
-
-       for (n = 0; n < ARRAY_SIZE(pool->cache_list); n++) {
-               struct drm_i915_gem_object *obj, *next;
-
-               list_for_each_entry_safe(obj, next,
-                                        &pool->cache_list[n],
-                                        batch_pool_link)
-                       i915_gem_object_put(obj);
-
-               INIT_LIST_HEAD(&pool->cache_list[n]);
-       }
-}
-
-/**
- * i915_gem_batch_pool_get() - allocate a buffer from the pool
- * @pool: the batch buffer pool
- * @size: the minimum desired size of the returned buffer
- *
- * Returns an inactive buffer from @pool with at least @size bytes,
- * with the pages pinned. The caller must i915_gem_object_unpin_pages()
- * on the returned object.
- *
- * Note: Callers must hold the struct_mutex
- *
- * Return: the buffer object or an error pointer
- */
-struct drm_i915_gem_object *
-i915_gem_batch_pool_get(struct i915_gem_batch_pool *pool,
-                       size_t size)
-{
-       struct drm_i915_gem_object *obj;
-       struct list_head *list;
-       int n, ret;
-
-       lockdep_assert_held(&pool->engine->i915->drm.struct_mutex);
-
-       /* Compute a power-of-two bucket, but throw everything greater than
-        * 16KiB into the same bucket: i.e. the the buckets hold objects of
-        * (1 page, 2 pages, 4 pages, 8+ pages).
-        */
-       n = fls(size >> PAGE_SHIFT) - 1;
-       if (n >= ARRAY_SIZE(pool->cache_list))
-               n = ARRAY_SIZE(pool->cache_list) - 1;
-       list = &pool->cache_list[n];
-
-       list_for_each_entry(obj, list, batch_pool_link) {
-               struct dma_resv *resv = obj->base.resv;
-
-               /* The batches are strictly LRU ordered */
-               if (!dma_resv_test_signaled_rcu(resv, true))
-                       break;
-
-               /*
-                * The object is now idle, clear the array of shared
-                * fences before we add a new request. Although, we
-                * remain on the same engine, we may be on a different
-                * timeline and so may continually grow the array,
-                * trapping a reference to all the old fences, rather
-                * than replace the existing fence.
-                */
-               if (rcu_access_pointer(resv->fence)) {
-                       dma_resv_lock(resv, NULL);
-                       dma_resv_add_excl_fence(resv, NULL);
-                       dma_resv_unlock(resv);
-               }
-
-               if (obj->base.size >= size)
-                       goto found;
-       }
-
-       obj = i915_gem_object_create_internal(pool->engine->i915, size);
-       if (IS_ERR(obj))
-               return obj;
-
-found:
-       ret = i915_gem_object_pin_pages(obj);
-       if (ret)
-               return ERR_PTR(ret);
-
-       list_move_tail(&obj->batch_pool_link, list);
-       return obj;
-}