Received-SPF: pass (google.com: domain of linux-kernel+bounces-168525-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Date: Sat, 4 May 2024 15:02:11 +0900
From: Sergey Senozhatsky <senozhatsky@chromium.org>
To: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: Andrew Morton <akpm@linux-foundation.org>,
	Minchan Kim <minchan@kernel.org>, linux-kernel@vger.kernel.org,
	linux-block@vger.kernel.org
Subject: Re: [PATCH 13/14] zram: add dictionary support to zstd backend
Message-ID: <20240504060211.GG14947@google.com>
References: <20240503091823.3616962-1-senozhatsky@chromium.org>
 <20240503091823.3616962-14-senozhatsky@chromium.org>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20240503091823.3616962-14-senozhatsky@chromium.org>

On (24/05/03 18:17), Sergey Senozhatsky wrote:
> This adds support for pre-trained zstd dictionaries [1]
> Dictionary is loaded once (per-config) and then loaded to Cctx
> and Dctx by reference, so we don't allocate extra memory.
> 
> The patch is a little non-trivial, as it seems that noone
> ever attempted to use dictionaries in the linux kernel
> port of zstd.
> 
> It also uses GFP_KERNEL gfp in Cctx customAlloc(). We probably
> would want to do something about it. Either make sure that we
> always (somehow) fully setup all Cctx contexts from non-atomic
> context before we attempt to use them, come up with some sort
> of custom allocator or stop calling zcomp_compress() from atomic
> context.
> 
> [1] https://github.com/facebook/zstd/blob/dev/programs/zstd.1.md#dictionary-builder

JFI
I reworked this patch quite significantly in v2 of the series.
I guess I'll post it soon.