lkml.org 
[lkml]   [2014]   [Jul]   [9]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH 1/1] Fix permission checking by NFS client for open-create with mode 000
From
On Wed, Jul 9, 2014 at 7:06 PM, Trond Myklebust
<trond.myklebust@primarydata.com> wrote:
> On Wed, Jul 9, 2014 at 6:42 PM, Frank Filz <ffilzlnx@mindspring.com> wrote:
>>> On Wed, Jul 9, 2014 at 5:54 PM, Frank S. Filz <ffilzlnx@mindspring.com>
>>> wrote:
>>> > From: "Frank S. Filz" <ffilzlnx@mindspring.com>
>>> >
>>> > The NFS v4 client sends a COMPOUND with an OPEN and an ACCESS.
>>> >
>>> > The ACCESS is required to verify an open for read is actually allowed
>>> > because RFC 3530 indicates OPEN for read only must succeed for an
>>> > execute only file.
>>> >
>>> > The old code expected to have read access if the requested access was
>>> > O_RDWR.
>>> >
>>> > We can expect the OPEN to properly permission check as long as the
>>> > open is O_WRONLY or O_RDWR.
>>> >
>>> > Signed-off-by: Frank S. Filz <ffilzlnx@mindspring.com>
>>> > ---
>>> > fs/nfs/nfs4proc.c | 25 ++++++++++++++++++++-----
>>> > 1 file changed, 20 insertions(+), 5 deletions(-)
>>> >
>>> > diff --git a/fs/nfs/nfs4proc.c b/fs/nfs/nfs4proc.c index
>>> > 4bf3d97..9742054 100644
>>> > --- a/fs/nfs/nfs4proc.c
>>> > +++ b/fs/nfs/nfs4proc.c
>>> > @@ -1966,15 +1966,30 @@ static int nfs4_opendata_access(struct
>>> rpc_cred *cred,
>>> > return 0;
>>> >
>>> > mask = 0;
>>> > - /* don't check MAY_WRITE - a newly created file may not have
>>> > - * write mode bits, but POSIX allows the creating process to write.
>>> > - * use openflags to check for exec, because fmode won't
>>> > - * always have FMODE_EXEC set when file open for exec. */
>>> > + /* Don't trust the permission check on OPEN if open for exec or for
>>> > + * read only. Since FMODE_EXEC doesn't go across the wire, the server
>>> > + * has no way to distinguish between an open to read an executable
>>> file
>>> > + * and an open to read a readable file. Write access is properly checked
>>> > + * and permission SHOULD always be granted if the file was created as
>>> a
>>> > + * result of this OPEN, no matter what mode the file was created with.
>>> > + *
>>> > + * NOTE: If the case of a OPEN CREATE READ-ONLY with a mode that
>>> does
>>> > + * not allow read access, this test will produce an incorrect
>>> > + * EACCES error.
>>> > + */
>>> > if (openflags & __FMODE_EXEC) {
>>> > /* ONLY check for exec rights */
>>> > mask = MAY_EXEC;
>>> > - } else if (fmode & FMODE_READ)
>>> > + } else if (!(fmode & FMODE_WRITE)) {
>>> > + /* In case the file was execute only, check for read permission
>>> > + * ONLY if write access was not requested. It is expected that
>>> > + * an OPEN for write will fail if the file is execute only.
>>> > + * Note that if the file was newly created, the fmode SHOULD
>>> > + * include FMODE_WRITE, especially if the file will be created
>>> > + * with a restrictive mode.
>>> > + */
>>> > mask = MAY_READ;
>>> > + }
>>>
>>> This looks wrong. AFAICS it will allow you to open an existing file which has -
>>> wx permissions (i.e. no read permissions) for O_RDWR. That should not be
>>> permitted under POSIX rules.
>>
>> The server permission checks the OPEN, this only affects the subsequent ACCESS.
>>
>> The server will fail the OPEN with NFS4_ERR_ACCESS if the open is for read/write and the file has write-execute permission.
>
> RFC3530bis draft 33 (6.2.1.3.1. Discussion of Mask Attributes) states
> that for both the OPEN and the READ operations, "Servers SHOULD allow
> a user the ability to read the data of the file when only the
> ACE4_EXECUTE access mask bit is allowed". RFC5561 has the same
> language.

Oops. Sorry, the correct sub-sub-sub-sub-....paragraph is this one:

Permission to execute a file.

Servers SHOULD allow a user the ability to read the data of the
file when only the ACE4_EXECUTE access mask bit is allowed.
This is because there is no way to execute a file without
reading the contents. Though a server may treat ACE4_EXECUTE
and ACE4_READ_DATA bits identically when deciding to permit a
READ operation, it SHOULD still allow the two bits to be set
independently in ACLs, and MUST distinguish between them when
replying to ACCESS operations. In particular, servers SHOULD
NOT silently turn on one of the two bits when the other is set,
as that would make it impossible for the client to correctly
enforce the distinction between read and execute permissions.


> To me that translates as saying that the server SHOULD accept an
> OPEN(SHARE_ACCESS_READ|SHARE_ACCESS_WRITE) request in the above
> situation.

Same conclusion, though....

--
Trond Myklebust

Linux NFS client maintainer, PrimaryData

trond.myklebust@primarydata.com


\
 
 \ /
  Last update: 2014-07-10 01:41    [W:0.074 / U:3.004 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site