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;
From:   =?utf-8?Q?Lu=C3=ADs_Henriques?= <lhenriques@suse.de>
To:     Xiubo Li <xiubli@redhat.com>
Cc:     Jeff Layton <jlayton@kernel.org>,
        Ilya Dryomov <idryomov@gmail.com>,
        Gregory Farnum <gfarnum@redhat.com>,
        ceph-devel@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [RFC PATCH v3] ceph: prevent a client from exceeding the MDS
 maximum xattr size
References: <20220601162939.12278-1-lhenriques@suse.de>
        <b788a7f9-9177-0398-7d21-a19ce7e6c957@redhat.com>
Date:   Thu, 02 Jun 2022 10:26:50 +0100
In-Reply-To: <b788a7f9-9177-0398-7d21-a19ce7e6c957@redhat.com> (Xiubo Li's
        message of "Thu, 2 Jun 2022 10:33:27 +0800")
Message-ID: <87h7534dr9.fsf@brahms.olymp>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Xiubo Li <xiubli@redhat.com> writes:

> On 6/2/22 12:29 AM, Lu=C3=ADs Henriques wrote:
>> The MDS tries to enforce a limit on the total key/values in extended
>> attributes.  However, this limit is enforced only if doing a synchronous
>> operation (MDS_OP_SETXATTR) -- if we're buffering the xattrs, the MDS
>> doesn't have a chance to enforce these limits.
>>
>> This patch adds support for decoding the xattrs maximum size setting tha=
t is
>> distributed in the mdsmap.  Then, when setting an xattr, the kernel clie=
nt
>> will revert to do a synchronous operation if that maximum size is exceed=
ed.
>>
>> While there, fix a dout() that would trigger a printk warning:
>>
>> [   98.718078] ------------[ cut here ]------------
>> [   98.719012] precision 65536 too large
>> [   98.719039] WARNING: CPU: 1 PID: 3755 at lib/vsprintf.c:2703 vsnprint=
f+0x5e3/0x600
>> ...
>>
>> URL: https://tracker.ceph.com/issues/55725
>> Signed-off-by: Lu=C3=ADs Henriques <lhenriques@suse.de>
>> ---
>>   fs/ceph/mdsmap.c            | 27 +++++++++++++++++++++++----
>>   fs/ceph/xattr.c             | 12 ++++++++----
>>   include/linux/ceph/mdsmap.h |  1 +
>>   3 files changed, 32 insertions(+), 8 deletions(-)
>>
>> * Changes since v2
>>
>> Well, a lot has changed since v2!  Now the xattr max value setting is
>> obtained through the mdsmap, which needs to be decoded, and the feature
>> that was used in the previous revision was dropped.  The drawback is that
>> the MDS isn't unable to know in advance if a client is aware of this xat=
tr
>> max value.
>>
>> * Changes since v1
>>
>> Added support for new feature bit to get the MDS max_xattr_pairs_size
>> setting.
>>
>> Also note that this patch relies on a patch that hasn't been merged yet
>> ("ceph: use correct index when encoding client supported features"),
>> otherwise the new feature bit won't be correctly encoded.
>>
>> diff --git a/fs/ceph/mdsmap.c b/fs/ceph/mdsmap.c
>> index 30387733765d..36b2bc18ca2a 100644
>> --- a/fs/ceph/mdsmap.c
>> +++ b/fs/ceph/mdsmap.c
>> @@ -13,6 +13,12 @@
>>     #include "super.h"
>>   +/*
>> + * Maximum size of xattrs the MDS can handle per inode by default.  This
>> + * includes the attribute name and 4+4 bytes for the key/value sizes.
>> + */
>> +#define MDS_MAX_XATTR_SIZE (1<<16) /* 64K */
>> +
>>   #define CEPH_MDS_IS_READY(i, ignore_laggy) \
>>   	(m->m_info[i].state > 0 && ignore_laggy ? true : !m->m_info[i].laggy)
>>   @@ -352,12 +358,10 @@ struct ceph_mdsmap *ceph_mdsmap_decode(void **p,=
 void
>> *end, bool msgr2)
>>   		__decode_and_drop_type(p, end, u8, bad_ext);
>>   	}
>>   	if (mdsmap_ev >=3D 8) {
>> -		u32 name_len;
>>   		/* enabled */
>>   		ceph_decode_8_safe(p, end, m->m_enabled, bad_ext);
>> -		ceph_decode_32_safe(p, end, name_len, bad_ext);
>> -		ceph_decode_need(p, end, name_len, bad_ext);
>> -		*p +=3D name_len;
>> +		/* fs_name */
>> +		ceph_decode_skip_string(p, end, bad_ext);
>>   	}
>>   	/* damaged */
>>   	if (mdsmap_ev >=3D 9) {
>> @@ -370,6 +374,21 @@ struct ceph_mdsmap *ceph_mdsmap_decode(void **p, vo=
id *end, bool msgr2)
>>   	} else {
>>   		m->m_damaged =3D false;
>>   	}
>> +	if (mdsmap_ev >=3D 17) {
>> +		/* balancer */
>> +		ceph_decode_skip_string(p, end, bad_ext);
>> +		/* standby_count_wanted */
>> +		ceph_decode_skip_32(p, end, bad_ext);
>> +		/* old_max_mds */
>> +		ceph_decode_skip_32(p, end, bad_ext);
>> +		/* min_compat_client */
>> +		ceph_decode_skip_8(p, end, bad_ext);
>
> This is incorrect.
>
> If mdsmap_ev =3D=3D 15 the min_compat_client will be a feature_bitset_t i=
nstead of
> int8_t.

Hmm... can you point me at where that's done in the code?  As usual, I'm
confused with that code and simply can't see that.

Also, if that happens only when mdsmap_ev =3D=3D 15, then there's no problem
because that branch is only taken if it's >=3D 17.

>
>
>> +		/* required_client_features */
>> +		ceph_decode_skip_set(p, end, 64, bad_ext);
>> +		ceph_decode_64_safe(p, end, m->m_max_xattr_size, bad_ext);
>> +	} else {
>> +		m->m_max_xattr_size =3D MDS_MAX_XATTR_SIZE;
>> +	}
>>   bad_ext:
>>   	dout("mdsmap_decode m_enabled: %d, m_damaged: %d, m_num_laggy: %d\n",
>>   	     !!m->m_enabled, !!m->m_damaged, m->m_num_laggy);
>> diff --git a/fs/ceph/xattr.c b/fs/ceph/xattr.c
>> index 8c2dc2c762a4..67f046dac35c 100644
>> --- a/fs/ceph/xattr.c
>> +++ b/fs/ceph/xattr.c
>> @@ -1086,7 +1086,7 @@ static int ceph_sync_setxattr(struct inode *inode,=
 const char *name,
>>   			flags |=3D CEPH_XATTR_REMOVE;
>>   	}
>>   -	dout("setxattr value=3D%.*s\n", (int)size, value);
>> +	dout("setxattr value size: %ld\n", size);
>>     	/* do request */
>>   	req =3D ceph_mdsc_create_request(mdsc, op, USE_AUTH_MDS);
>> @@ -1184,8 +1184,14 @@ int __ceph_setxattr(struct inode *inode, const ch=
ar *name,
>>   	spin_lock(&ci->i_ceph_lock);
>>   retry:
>>   	issued =3D __ceph_caps_issued(ci, NULL);
>> -	if (ci->i_xattrs.version =3D=3D 0 || !(issued & CEPH_CAP_XATTR_EXCL))
>> +	required_blob_size =3D __get_required_blob_size(ci, name_len, val_len);
>> +	if ((ci->i_xattrs.version =3D=3D 0) || !(issued & CEPH_CAP_XATTR_EXCL)=
 ||
>> +	    (required_blob_size >=3D mdsc->mdsmap->m_max_xattr_size)) {
>
> Shouldn't it be '>' instead ?

Ok, I'll fix that.

> We'd better always force to do a sync request with old ceph. Just check i=
f the
> mdsmap_ev < 17. It's not safe to buffer it because it maybe discarded as =
your
> ceph PR does.

Right, that can be done.  So, I can simply set the m_max_xattr_size to '0'
if mdsmap_ev < 17.  Then, this 'if' condition will always be evaluated to
true because required_blob_size will be > 0.  Does that sound OK?

Cheers,
--=20
Lu=C3=ADs


> -- Xiubo
>
>> +		dout("%s do sync setxattr: version: %llu size: %d max: %llu\n",
>> +		     __func__, ci->i_xattrs.version, required_blob_size,
>> +		     mdsc->mdsmap->m_max_xattr_size);
>>   		goto do_sync;
>> +	}
>>     	if (!lock_snap_rwsem && !ci->i_head_snapc) {
>>   		lock_snap_rwsem =3D true;
>> @@ -1201,8 +1207,6 @@ int __ceph_setxattr(struct inode *inode, const cha=
r *name,
>>   	     ceph_cap_string(issued));
>>   	__build_xattrs(inode);
>>   -	required_blob_size =3D __get_required_blob_size(ci, name_len, val_le=
n);
>> -
>>   	if (!ci->i_xattrs.prealloc_blob ||
>>   	    required_blob_size > ci->i_xattrs.prealloc_blob->alloc_len) {
>>   		struct ceph_buffer *blob;
>> diff --git a/include/linux/ceph/mdsmap.h b/include/linux/ceph/mdsmap.h
>> index 523fd0452856..4c3e0648dc27 100644
>> --- a/include/linux/ceph/mdsmap.h
>> +++ b/include/linux/ceph/mdsmap.h
>> @@ -25,6 +25,7 @@ struct ceph_mdsmap {
>>   	u32 m_session_timeout;          /* seconds */
>>   	u32 m_session_autoclose;        /* seconds */
>>   	u64 m_max_file_size;
>> +	u64 m_max_xattr_size;		/* maximum size for xattrs blob */
>>   	u32 m_max_mds;			/* expected up:active mds number */
>>   	u32 m_num_active_mds;		/* actual up:active mds number */
>>   	u32 possible_max_rank;		/* possible max rank index */
>>