quoted

quoted

quoted

quoted

+      /*
+       * If the region state is active, mctx must be allocated.
+       * In this case, check whether the region is evicted or
+       * mctx allcation fail.
+       */
+      if (unlikely(!srgn->mctx)) {
+              dev_err(&hpb->sdev_ufs_lu->sdev_dev,
+                      "no mctx in region %d subregion %d.\n",
+                      srgn->rgn_idx, srgn->srgn_idx);
+              return true;
+      }
+
+      if ((srgn_offset + cnt) > bitmap_len)
+              bit_len = bitmap_len - srgn_offset;
+      else
+              bit_len = cnt;
+
+      if (find_next_bit(srgn->mctx->ppn_dirty, bitmap_len,
+                        srgn_offset) < bit_len + srgn_offset)
+              return true;
+

It seems unnecessary to search through bitmap_len
How about searching by transfer size?

if (find_next_bit(srgn->mctx->ppn_dirty,
              bit_len + srgn_offset, srgn_offset) < bit_len + srgn_offset)

Isn't bit_len should be used for size, and not bit_len + srgn_offset ?

then find_next_bit checks from start to bit_len.
find_next_bit stops checking if start is greater than bit_len.
it does not check for dirty as transfer_size.

Right. Size (nbits in _find_next_bit) practically means @end - Confusing...
Either way, Is this tad optimization worth another spin in your opinion?

Worst case, we have to search every time up to the bitmap_len size. 
Even though, when bit_len is 1, only 1 bit needs to be checked
it can increase the effect, depending on the subregion size. 

Thanks
Yohan

Thanks,
Avri

quoted

Thanks
Yohan

quoted

Thanks,
Avri

quoted

Thanks
Yohan