[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <a9a9343f962a50f508324db6397131982abb12e4.camel@linux.intel.com>
Date: Mon, 17 Mar 2025 10:30:39 -0700
From: Tim Chen <tim.c.chen@...ux.intel.com>
To: Kemeng Shi <shikemeng@...weicloud.com>, akpm@...ux-foundation.org
Cc: kasong@...cent.com, linux-mm@...ck.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH 7/9] mm: swap: free each cluster individually in
swap_entries_put_map_nr()
>
> fallback:
> - for (i = 0; i < nr; i++) {
> - if (data_race(si->swap_map[offset + i])) {
> - count = swap_entry_put(si, swp_entry(type, offset + i));
> - if (count == SWAP_HAS_CACHE)
> - has_cache = true;
> - } else {
> - WARN_ON_ONCE(1);
> - }
> + ci = lock_cluster(si, offset);
Relocking the cluster is unnecessary overhead if we don't release the ci lock
prior to jumping to fallback label for the locked case.
Maybe something like:
diff --git a/mm/swapfile.c b/mm/swapfile.c
index 9cd93a03b25c..41aa841b86d6 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -1504,8 +1504,7 @@ static bool swap_entries_put_map(struct swap_info_struct *si,
ci = lock_cluster(si, offset);
if (!swap_is_last_map(si, offset, nr, &has_cache)) {
- unlock_cluster(ci);
- goto fallback;
+ goto locked_fallback;
}
if (!has_cache)
swap_entries_free(si, ci, entry, nr);
@@ -1518,6 +1517,7 @@ static bool swap_entries_put_map(struct swap_info_struct *si,
fallback:
ci = lock_cluster(si, offset);
+locked_fallback:
for (i = 0; i < nr; i++, entry.val++) {
count = swap_entry_put_locked(si, ci, entry, 1);
if (count == SWAP_HAS_CACHE)
Tim
> + for (i = 0; i < nr; i++, entry.val++) {
> + count = swap_entry_put_locked(si, ci, entry, 1);
> + if (count == SWAP_HAS_CACHE)
> + has_cache = true;
> }
> + unlock_cluster(ci);
> + return has_cache;
> +
> +}
> +
Powered by blists - more mailing lists