[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Y2st0b00KHTK0nrI@google.com>
Date: Wed, 9 Nov 2022 13:34:25 +0900
From: Sergey Senozhatsky <senozhatsky@...omium.org>
To: Minchan Kim <minchan@...nel.org>
Cc: Andrew Morton <akpm@...ux-foundation.org>,
Nitin Gupta <ngupta@...are.org>, linux-kernel@...r.kernel.org,
linux-mm@...ck.org, Sergey Senozhatsky <senozhatsky@...omium.org>
Subject: Re: [PATCHv4 7/9] zram: add pages_per_pool_page device attribute
On (22/10/31 14:41), Sergey Senozhatsky wrote:
[..]
> zram->mem_pool = zs_create_pool(zram->disk->disk_name,
> - ZS_DEFAULT_PAGES_PER_ZSPAGE);
> + zram->pages_per_pool_page);
> if (!zram->mem_pool) {
> vfree(zram->table);
> return false;
> @@ -2174,6 +2213,7 @@ static DEVICE_ATTR_RW(writeback_limit_enable);
> static DEVICE_ATTR_RW(recomp_algorithm);
> static DEVICE_ATTR_WO(recompress);
> #endif
> +static DEVICE_ATTR_RW(pages_per_pool_page);
May be we can have a more generic "allocator_tunables" device attribute,
which will support named parameters instead. E.g.
pool_page_len_limit=INT
And more in the future.
Having hard times coming up with good names here. max_pages_per_zspage
is too low level and exposes zsmalloc internals, while in theory zram
can use different allocators at some point, and those allocators can
have different tunables.
Powered by blists - more mailing lists