Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Mon, 4 Nov 2019 19:04:51 -0800
From:   Jaegeuk Kim <jaegeuk@kernel.org>
To:     Chao Yu <yuchao0@huawei.com>
Cc:     linux-kernel@vger.kernel.org,
        linux-f2fs-devel@lists.sourceforge.net
Subject: Re: [PATCH 3/3] Revert "f2fs: use kvmalloc, if kmalloc is failed"
Message-ID: <20191105030451.GA55090@jaegeuk-macbookpro.roam.corp.google.com>
References: <20191101095324.9902-1-yuchao0@huawei.com>
 <20191101095324.9902-3-yuchao0@huawei.com>
 <20191105000249.GA46956@jaegeuk-macbookpro.roam.corp.google.com>
 <40d0df3f-cc55-d31a-474b-76f57d96bd89@huawei.com>
 <20191105023835.GD692@sol.localdomain>
 <00ade77c-5451-4953-0232-89342a029f33@huawei.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <00ade77c-5451-4953-0232-89342a029f33@huawei.com>
User-Agent: Mutt/1.8.2 (2017-04-18)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On 11/05, Chao Yu wrote:
> On 2019/11/5 10:38, Eric Biggers wrote:
> > On Tue, Nov 05, 2019 at 10:17:41AM +0800, Chao Yu wrote:
> >> On 2019/11/5 8:02, Jaegeuk Kim wrote:
> >>> On 11/01, Chao Yu wrote:
> >>>> This reverts commit 5222595d093ebe80329d38d255d14316257afb3e.
> >>>>
> >>>> As discussed with Eric, as kvmalloc() will try kmalloc() first, so
> >>>> when we need allocate large size memory, it'd better to use
> >>>> f2fs_kvmalloc() directly rather than adding additional fallback
> >>>> logic to call kvmalloc() after we failed in f2fs_kmalloc().
> >>>>
> >>>> In order to avoid allocation failure described in original commit,
> >>>> I change to use f2fs_kvmalloc() for .free_nid_bitmap bitmap memory.
> >>>
> >>> Is there any problem in the previous flow?
> >>
> >> No existing problem, however, it's redundant to introduce fallback flow in
> >> f2fs_kmalloc() like vmalloc() did, since we can call f2fs_vmalloc() directly in
> >> places where we need large memory.
> >>
> >> Thanks,
> >>
> > 
> > f2fs_kmalloc() also violated the naming convention used everywhere else in the
> > kernel since it could return both kmalloc and vmalloc memory, not just kmalloc
> > memory.  That's really error-prone since people would naturally assume it's safe
> > to free the *_kmalloc()-ed memory with kfree().
> 
> Agreed.

Then, why not just keeping f2fs_kvmalloc() and replace all f2fs_kmalloc() with
f2fs_kvmalloc()?

f2fs_kvmalloc()
- call kmalloc()
- vmalloc(), if failed

I'd like to keep the allocation behavior first.

Thanks,