Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <1689669277-15291-1-git-send-email-yunlong.xing@unisoc.com>
In-Reply-To: <1689669277-15291-1-git-send-email-yunlong.xing@unisoc.com>
From:   Amir Goldstein <amir73il@gmail.com>
Date:   Tue, 18 Jul 2023 14:37:48 +0300
Message-ID: <CAOQ4uxgLhf8EyticTQUxtzUdVD=UEUTwQnTkujXE1KuaVxNQyQ@mail.gmail.com>
Subject: Re: [PATCH] ovl: fix mount fail because the upper doesn't have space
To:     Yunlong Xing <yunlong.xing@unisoc.com>
Cc:     miklos@szeredi.hu, linux-unionfs@vger.kernel.org,
        linux-kernel@vger.kernel.org, stable@vger.kernel.org,
        gregkh@linuxfoundation.org, sashal@kernel.org,
        yunlongxing23@gmail.com, zhiguo.niu@unisoc.com,
        hongyu.jin@unisoc.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Tue, Jul 18, 2023 at 11:35=E2=80=AFAM Yunlong Xing <yunlong.xing@unisoc.=
com> wrote:
>
> The current ovlfs mount flow:
>
> ovl_fill_super
>  |_ovl_get_workdir
>     |_ovl_make_workdir
>        |_ovl_check_rename_whiteout
>
> In ovl_check_rename_whiteout(), a new file is attempted to create.But if
> the upper doesn't have space to do this, it will return error -ENOSPC,
> causing the mount fail. It means that if the upper is full, the overlayfs
> cannot be mounted.It is not reasonable, so this patch will omit this erro=
r
>  and continue mount flow.
>
> Fixes: cad218ab3320 ("ovl: check if upper fs supports RENAME_WHITEOUT")
> Signed-off-by: Yunlong Xing <yunlong.xing@unisoc.com>
> ---
>  fs/overlayfs/super.c | 9 ++++++---
>  1 file changed, 6 insertions(+), 3 deletions(-)
>
> diff --git a/fs/overlayfs/super.c b/fs/overlayfs/super.c
> index 5b069f1a1e44..2cf41e978cff 100644
> --- a/fs/overlayfs/super.c
> +++ b/fs/overlayfs/super.c
> @@ -744,9 +744,12 @@ static int ovl_make_workdir(struct super_block *sb, =
struct ovl_fs *ofs,
>
>         /* Check if upper/work fs supports RENAME_WHITEOUT */
>         err =3D ovl_check_rename_whiteout(ofs);
> -       if (err < 0)
> -               goto out;
> -
> +       if (err < 0) {
> +               if (err =3D=3D -ENOSPC)
> +                       pr_warn("upper fs check RENAME_WHITEOUT fail due =
to no space.\n");
> +               else
> +                       goto out;
> +       }
>         rename_whiteout =3D err;
>         if (!rename_whiteout)
>                 pr_warn("upper fs does not support RENAME_WHITEOUT.\n");
> --

This assumes that RENAME_WHITEOUT is supported.
I rather assume it is not supported if the check fails.
Like this is shorter (not tested):

--- a/fs/overlayfs/super.c
+++ b/fs/overlayfs/super.c
@@ -744,12 +744,13 @@ static int ovl_make_workdir(struct super_block
*sb, struct ovl_fs *ofs,

        /* Check if upper/work fs supports RENAME_WHITEOUT */
        err =3D ovl_check_rename_whiteout(ofs);
-       if (err < 0)
+       if (err < 0 && err !=3D -ENOSPC)
                goto out;

-       rename_whiteout =3D err;
+       rename_whiteout =3D err > 0;
        if (!rename_whiteout)
-               pr_warn("upper fs does not support RENAME_WHITEOUT.\n");
+               pr_warn("upper fs does not support RENAME_WHITEOUT (%i).\n"=
,
+                       err);

Thanks,
Amir.