media: atomisp: hmm_bo: Simplify alloc_private_pages()
authorHans de Goede <hdegoede@redhat.com>
Sat, 13 Aug 2022 15:38:45 +0000 (17:38 +0200)
committerMauro Carvalho Chehab <mchehab@kernel.org>
Sat, 24 Sep 2022 07:40:10 +0000 (09:40 +0200)
Since lack_mem starts initialized to true, alloc_private_pages() will
always set order to HMM_MIN_ORDER aka 0 / will always alloc 1 page at
a time.

So all the magic to decrease order if allocs fail is not necessary
and can be removed.

Reviewed-by: Andy Shevchenko <andy.shevchenko@gmail.com>
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Signed-off-by: Mauro Carvalho Chehab <mchehab@kernel.org>
drivers/staging/media/atomisp/include/hmm/hmm_bo.h
drivers/staging/media/atomisp/pci/hmm/hmm_bo.c

index 385e22f..901dc37 100644 (file)
@@ -65,9 +65,6 @@
 #define        check_bo_null_return_void(bo)   \
        check_null_return_void(bo, "NULL hmm buffer object.\n")
 
-#define        HMM_MAX_ORDER           3
-#define        HMM_MIN_ORDER           0
-
 #define        ISP_VM_START    0x0
 #define        ISP_VM_SIZE     (0x7FFFFFFF)    /* 2G address space */
 #define        ISP_PTR_NULL    NULL
index f504941..2753142 100644 (file)
 #include "hmm/hmm_common.h"
 #include "hmm/hmm_bo.h"
 
-static unsigned int order_to_nr(unsigned int order)
-{
-       return 1U << order;
-}
-
-static unsigned int nr_to_order_bottom(unsigned int nr)
-{
-       return fls(nr) - 1;
-}
-
 static int __bo_init(struct hmm_bo_device *bdev, struct hmm_buffer_object *bo,
                     unsigned int pgnr)
 {
@@ -653,13 +643,10 @@ static void free_private_bo_pages(struct hmm_buffer_object *bo,
 static int alloc_private_pages(struct hmm_buffer_object *bo)
 {
        int ret;
-       unsigned int pgnr, order, blk_pgnr, alloc_pgnr;
+       unsigned int pgnr, blk_pgnr, alloc_pgnr;
        struct page *pages;
        gfp_t gfp = GFP_NOWAIT | __GFP_NOWARN; /* REVISIT: need __GFP_FS too? */
        int i, j;
-       int failure_number = 0;
-       bool reduce_order = false;
-       bool lack_mem = true;
 
        pgnr = bo->pgnr;
 
@@ -667,58 +654,17 @@ static int alloc_private_pages(struct hmm_buffer_object *bo)
        alloc_pgnr = 0;
 
        while (pgnr) {
-               order = nr_to_order_bottom(pgnr);
-               /*
-                * if be short of memory, we will set order to 0
-                * everytime.
-                */
-               if (lack_mem)
-                       order = HMM_MIN_ORDER;
-               else if (order > HMM_MAX_ORDER)
-                       order = HMM_MAX_ORDER;
-retry:
-               /*
-                * When order > HMM_MIN_ORDER, for performance reasons we don't
-                * want alloc_pages() to sleep. In case it fails and fallbacks
-                * to HMM_MIN_ORDER or in case the requested order is originally
-                * the minimum value, we can allow alloc_pages() to sleep for
-                * robustness purpose.
-                *
-                * REVISIT: why __GFP_FS is necessary?
-                */
-               if (order == HMM_MIN_ORDER) {
-                       gfp &= ~GFP_NOWAIT;
-                       gfp |= __GFP_RECLAIM | __GFP_FS;
-               }
+               gfp &= ~GFP_NOWAIT;
+               gfp |= __GFP_RECLAIM | __GFP_FS;
 
-               pages = alloc_pages(gfp, order);
+               pages = alloc_pages(gfp, 0); // alloc 1 page
                if (unlikely(!pages)) {
-                       /*
-                        * in low memory case, if allocation page fails,
-                        * we turn to try if order=0 allocation could
-                        * succeed. if order=0 fails too, that means there is
-                        * no memory left.
-                        */
-                       if (order == HMM_MIN_ORDER) {
-                               dev_err(atomisp_dev,
-                                       "%s: cannot allocate pages\n",
-                                       __func__);
-                               goto cleanup;
-                       }
-                       order = HMM_MIN_ORDER;
-                       failure_number++;
-                       reduce_order = true;
-                       /*
-                        * if fail two times continuously, we think be short
-                        * of memory now.
-                        */
-                       if (failure_number == 2) {
-                               lack_mem = true;
-                               failure_number = 0;
-                       }
-                       goto retry;
+                       dev_err(atomisp_dev,
+                               "%s: cannot allocate pages\n",
+                               __func__);
+                       goto cleanup;
                } else {
-                       blk_pgnr = order_to_nr(order);
+                       blk_pgnr = 1;
 
                        /*
                         * set memory to uncacheable -- UC_MINUS
@@ -728,7 +674,7 @@ retry:
                                dev_err(atomisp_dev,
                                        "set page uncacheablefailed.\n");
 
-                               __free_pages(pages, order);
+                               __free_pages(pages, 0);
 
                                goto cleanup;
                        }
@@ -738,15 +684,6 @@ retry:
                        }
 
                        pgnr -= blk_pgnr;
-
-                       /*
-                        * if order is not reduced this time, clear
-                        * failure_number.
-                        */
-                       if (reduce_order)
-                               reduce_order = false;
-                       else
-                               failure_number = 0;
                }
        }