From: "Luís Henriques" <lhenriques@suse.de>
To: Xiubo Li <xiubli@redhat.com>
Cc: idryomov@gmail.com, ceph-devel@vger.kernel.org,
jlayton@kernel.org, vshankar@redhat.com, mchangir@redhat.com
Subject: Re: [PATCH v16 25/68] ceph: make d_revalidate call fscrypt revalidator for encrypted dentries
Date: Wed, 08 Mar 2023 09:29:24 +0000 [thread overview]
Message-ID: <87jzzr8ubv.fsf@suse.de> (raw)
In-Reply-To: <72e7b6cc-ba6b-796e-2ff6-1e8ff2ac7eee@redhat.com> (Xiubo Li's message of "Wed, 8 Mar 2023 09:50:40 +0800")
Xiubo Li <xiubli@redhat.com> writes:
> On 08/03/2023 02:53, Luís Henriques wrote:
>> xiubli@redhat.com writes:
>>
>>> From: Jeff Layton <jlayton@kernel.org>
>>>
>>> If we have a dentry which represents a no-key name, then we need to test
>>> whether the parent directory's encryption key has since been added. Do
>>> that before we test anything else about the dentry.
>>>
>>> Reviewed-by: Xiubo Li <xiubli@redhat.com>
>>> Signed-off-by: Jeff Layton <jlayton@kernel.org>
>>> ---
>>> fs/ceph/dir.c | 8 ++++++--
>>> 1 file changed, 6 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/fs/ceph/dir.c b/fs/ceph/dir.c
>>> index d3c2853bb0f1..5ead9f59e693 100644
>>> --- a/fs/ceph/dir.c
>>> +++ b/fs/ceph/dir.c
>>> @@ -1770,6 +1770,10 @@ static int ceph_d_revalidate(struct dentry *dentry, unsigned int flags)
>>> struct inode *dir, *inode;
>>> struct ceph_mds_client *mdsc;
>>> + valid = fscrypt_d_revalidate(dentry, flags);
>>> + if (valid <= 0)
>>> + return valid;
>>> +
>> This patch has confused me in the past, and today I found myself
>> scratching my head again looking at it.
>>
>> So, I've started seeing generic/123 test failing when running it with
>> test_dummy_encryption. I was almost sure that this test used to run fine
>> before, but I couldn't find any evidence (somehow I lost my old testing
>> logs...).
>>
>> Anyway, the test is quite simple:
>>
>> 1. Creates a directory with write permissions for root only
>> 2. Writes into a file in that directory
>> 3. Uses 'su' to try to modify that file as a different user, and
>> gets -EPERM
>>
>> All these steps run fine, and the test should pass. *However*, in the
>> test cleanup function, a simple 'rm -rf <dir>' will fail with -ENOTEMPTY.
>> 'strace' shows that calling unlinkat() to remove the file got a '-ENOENT'
>> and then -ENOTEMPTY for the directory.
>>
>> Some digging allowed me to figure out that running commands with 'su' will
>> drop caches (I see 'su (874): drop_caches: 2' in the log). And this is
>> how I ended up looking at this patch. fscrypt_d_revalidate() will return
>> '0' if the parent directory does has a key (fscrypt_has_encryption_key()).
>> Can we really say here that the dentry is *not* valid in that case? Or
>> should that '<= 0' be a '< 0'?
>>
>> (But again, this patch has confused me before...)
>
> Luis,
>
> Could you reproduce it with the latest testing branch ?
Yes, I'm seeing this with the latest code.
> I never seen the generic/123 failure yet. And just now I ran the test for many
> times locally it worked fine.
That's odd. With 'test_dummy_encryption' mount option I can reproduce it
every time.
> From the generic/123 test code it will never touch the key while testing, that
> means the dentries under the test dir will always have the keyed name. And then
> the 'fscrypt_d_revalidate()' should return 1 always.
>
> Only when we remove the key will it trigger evicting the inodes and then when we
> add the key back will the 'fscrypt_d_revalidate()' return 0 by checking the
> 'fscrypt_has_encryption_key()'.
>
> As I remembered we have one or more fixes about this those days, not sure
> whether you were hitting those bugs we have already fixed ?
Yeah, I remember now, and I guess there's yet another one here!
I'll look closer into this and see if I can find out something else. I'm
definitely seeing 'fscrypt_d_revalidate()' returning 0, so probably the
bug is in the error paths, when the 'fsgqa' user tries to write into the
file.
Thanks for your feedback, Xiubo.
Cheers,
--
Luís
next prev parent reply other threads:[~2023-03-08 9:29 UTC|newest]
Thread overview: 84+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-02-27 3:27 [PATCH v16 00/68] ceph+fscrypt: full support xiubli
2023-02-27 3:27 ` [PATCH v16 01/68] libceph: add spinlock around osd->o_requests xiubli
2023-02-27 3:27 ` [PATCH v16 02/68] libceph: define struct ceph_sparse_extent and add some helpers xiubli
2023-02-27 3:27 ` [PATCH v16 03/68] libceph: add sparse read support to msgr2 crc state machine xiubli
2023-02-27 3:27 ` [PATCH v16 04/68] libceph: add sparse read support to OSD client xiubli
2023-02-27 3:27 ` [PATCH v16 05/68] libceph: support sparse reads on msgr2 secure codepath xiubli
2023-02-27 3:27 ` [PATCH v16 06/68] libceph: add sparse read support to msgr1 xiubli
2023-02-27 3:27 ` [PATCH v16 07/68] ceph: add new mount option to enable sparse reads xiubli
2023-02-27 3:27 ` [PATCH v16 08/68] ceph: preallocate inode for ops that may create one xiubli
2023-02-27 3:27 ` [PATCH v16 09/68] ceph: make ceph_msdc_build_path use ref-walk xiubli
2023-02-27 3:27 ` [PATCH v16 10/68] libceph: add new iov_iter-based ceph_msg_data_type and ceph_osd_data_type xiubli
2023-02-27 3:27 ` [PATCH v16 11/68] ceph: use osd_req_op_extent_osd_iter for netfs reads xiubli
2023-02-27 3:27 ` [PATCH v16 12/68] ceph: fscrypt_auth handling for ceph xiubli
2023-02-27 3:27 ` [PATCH v16 13/68] ceph: ensure that we accept a new context from MDS for new inodes xiubli
2023-02-27 3:27 ` [PATCH v16 14/68] ceph: add support for fscrypt_auth/fscrypt_file to cap messages xiubli
2023-02-27 3:27 ` [PATCH v16 15/68] ceph: implement -o test_dummy_encryption mount option xiubli
2023-02-27 3:27 ` [PATCH v16 16/68] ceph: decode alternate_name in lease info xiubli
2023-02-27 3:27 ` [PATCH v16 17/68] ceph: add fscrypt ioctls xiubli
2023-02-27 3:27 ` [PATCH v16 18/68] ceph: make the ioctl cmd more readable in debug log xiubli
2023-02-27 3:27 ` [PATCH v16 19/68] ceph: add base64 endcoding routines for encrypted names xiubli
2023-02-27 3:27 ` [PATCH v16 20/68] ceph: add encrypted fname handling to ceph_mdsc_build_path xiubli
2023-02-27 3:27 ` [PATCH v16 21/68] ceph: send altname in MClientRequest xiubli
2023-02-27 3:27 ` [PATCH v16 22/68] ceph: encode encrypted name in dentry release xiubli
2023-02-27 3:27 ` [PATCH v16 23/68] ceph: properly set DCACHE_NOKEY_NAME flag in lookup xiubli
2023-02-27 3:27 ` [PATCH v16 24/68] ceph: set DCACHE_NOKEY_NAME in atomic open xiubli
2023-02-27 3:27 ` [PATCH v16 25/68] ceph: make d_revalidate call fscrypt revalidator for encrypted dentries xiubli
2023-03-07 18:53 ` Luís Henriques
2023-03-08 1:50 ` Xiubo Li
2023-03-08 9:29 ` Luís Henriques [this message]
2023-03-08 10:42 ` Xiubo Li
2023-03-08 17:14 ` Luís Henriques
2023-03-08 17:54 ` Jeff Layton
2023-03-08 18:30 ` Luís Henriques
2023-03-08 19:32 ` Jeff Layton
2023-03-09 9:52 ` Luís Henriques
2023-03-09 7:06 ` Xiubo Li
2023-03-09 9:55 ` Luís Henriques
2023-03-09 11:41 ` Xiubo Li
2023-02-27 3:27 ` [PATCH v16 26/68] ceph: add helpers for converting names for userland presentation xiubli
2023-02-27 3:27 ` [PATCH v16 27/68] ceph: fix base64 encoded name's length check in ceph_fname_to_usr() xiubli
2023-02-27 3:27 ` [PATCH v16 28/68] ceph: add fscrypt support to ceph_fill_trace xiubli
2023-02-27 3:27 ` [PATCH v16 29/68] ceph: pass the request to parse_reply_info_readdir() xiubli
2023-02-27 3:27 ` [PATCH v16 30/68] ceph: add ceph_encode_encrypted_dname() helper xiubli
2023-02-27 3:27 ` [PATCH v16 31/68] ceph: add support to readdir for encrypted filenames xiubli
2023-02-27 3:27 ` [PATCH v16 32/68] ceph: create symlinks with encrypted and base64-encoded targets xiubli
2023-02-27 3:27 ` [PATCH v16 33/68] ceph: make ceph_get_name decrypt filenames xiubli
2023-02-27 3:27 ` [PATCH v16 34/68] ceph: add a new ceph.fscrypt.auth vxattr xiubli
2023-02-27 3:27 ` [PATCH v16 35/68] ceph: add some fscrypt guardrails xiubli
2023-02-27 3:27 ` [PATCH v16 36/68] ceph: allow encrypting a directory while not having Ax caps xiubli
2023-02-27 3:27 ` [PATCH v16 37/68] ceph: mark directory as non-complete after loading key xiubli
2023-02-27 3:27 ` [PATCH v16 38/68] ceph: don't allow changing layout on encrypted files/directories xiubli
2023-02-27 3:27 ` [PATCH v16 39/68] libceph: add CEPH_OSD_OP_ASSERT_VER support xiubli
2023-02-27 3:27 ` [PATCH v16 40/68] ceph: size handling for encrypted inodes in cap updates xiubli
2023-02-27 3:27 ` [PATCH v16 41/68] ceph: fscrypt_file field handling in MClientRequest messages xiubli
2023-02-27 3:27 ` [PATCH v16 42/68] ceph: get file size from fscrypt_file when present in inode traces xiubli
2023-02-27 3:27 ` [PATCH v16 43/68] ceph: handle fscrypt fields in cap messages from MDS xiubli
2023-02-27 3:27 ` [PATCH v16 44/68] ceph: update WARN_ON message to pr_warn xiubli
2023-02-27 3:27 ` [PATCH v16 45/68] ceph: add __ceph_get_caps helper support xiubli
2023-02-27 3:27 ` [PATCH v16 46/68] ceph: add __ceph_sync_read " xiubli
2023-02-27 3:27 ` [PATCH v16 47/68] ceph: add object version support for sync read xiubli
2023-02-27 3:27 ` [PATCH v16 48/68] ceph: add infrastructure for file encryption and decryption xiubli
2023-02-27 3:27 ` [PATCH v16 49/68] ceph: add truncate size handling support for fscrypt xiubli
2023-02-27 3:27 ` [PATCH v16 50/68] libceph: allow ceph_osdc_new_request to accept a multi-op read xiubli
2023-02-27 3:27 ` [PATCH v16 51/68] ceph: disable fallocate for encrypted inodes xiubli
2023-02-27 3:27 ` [PATCH v16 52/68] ceph: disable copy offload on " xiubli
2023-02-27 3:27 ` [PATCH v16 53/68] ceph: don't use special DIO path for " xiubli
2023-02-27 3:27 ` [PATCH v16 54/68] ceph: align data in pages in ceph_sync_write xiubli
2023-02-27 3:28 ` [PATCH v16 55/68] ceph: add read/modify/write to ceph_sync_write xiubli
2023-02-27 3:28 ` [PATCH v16 56/68] ceph: plumb in decryption during sync reads xiubli
2023-02-27 3:28 ` [PATCH v16 57/68] ceph: add fscrypt decryption support to ceph_netfs_issue_op xiubli
2023-02-27 3:28 ` [PATCH v16 58/68] ceph: set i_blkbits to crypto block size for encrypted inodes xiubli
2023-02-27 3:28 ` [PATCH v16 59/68] ceph: add encryption support to writepage xiubli
2023-02-27 3:28 ` [PATCH v16 60/68] ceph: fscrypt support for writepages xiubli
2023-02-27 3:28 ` [PATCH v16 61/68] ceph: invalidate pages when doing direct/sync writes xiubli
2023-02-27 3:28 ` [PATCH v16 62/68] ceph: add support for encrypted snapshot names xiubli
2023-02-27 3:28 ` [PATCH v16 63/68] ceph: add support for handling " xiubli
2023-02-27 3:28 ` [PATCH v16 64/68] ceph: update documentation regarding snapshot naming limitations xiubli
2023-02-27 3:28 ` [PATCH v16 65/68] ceph: prevent snapshots to be created in encrypted locked directories xiubli
2023-02-27 3:28 ` [PATCH v16 66/68] ceph: report STATX_ATTR_ENCRYPTED on encrypted inodes xiubli
2023-02-27 3:28 ` [PATCH v16 67/68] libceph: defer removing the req from osdc just after req->r_callback xiubli
2023-02-27 3:28 ` [PATCH v16 68/68] ceph: drop the messages from MDS when unmounting xiubli
2023-02-27 9:27 ` [PATCH v16 00/68] ceph+fscrypt: full support Luís Henriques
2023-02-27 9:58 ` Xiubo Li
2023-02-27 10:30 ` Luís Henriques
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=87jzzr8ubv.fsf@suse.de \
--to=lhenriques@suse.de \
--cc=ceph-devel@vger.kernel.org \
--cc=idryomov@gmail.com \
--cc=jlayton@kernel.org \
--cc=mchangir@redhat.com \
--cc=vshankar@redhat.com \
--cc=xiubli@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).