linux-kernel - [PATCH v3 0/9] cifs: Use iov_iters down to the network transport

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 27 May 2022 11:43:55 +0100
From:   David Howells <dhowells@...hat.com>
To:     Steve French <smfrench@...il.com>
Cc:     linux-cifs@...r.kernel.org,
        Shyam Prasad N <nspmangalore@...il.com>,
        Rohith Surabattula <rohiths.msft@...il.com>,
        Steve French <sfrench@...ba.org>, dhowells@...hat.com,
        Shyam Prasad N <nspmangalore@...il.com>,
        Rohith Surabattula <rohiths.msft@...il.com>,
        Jeff Layton <jlayton@...nel.org>,
        Al Viro <viro@...iv.linux.org.uk>, linux-cifs@...r.kernel.org,
        linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: [PATCH v3 0/9] cifs: Use iov_iters down to the network transport


Hi Steve,

Here's a third version of a subset of my cifs-experimental branch.  It
alters the cifs driver to pass iov_iters down to the lowest layers where
they can be passed to the network transport.

ver #3)
- I fixed some things in the RDMA interface after running xfstests.  It now
  correctly chops up requests so that they'll actually fit within the
  available SGE slots, both for short and bulk requests.

- I also fixed both the cifs_readahead and cifs_writepages algorithms, so
  those now work properly - and I found and fixed a memory leak too.

cifs over softRoCE RDMA got as far as generic/138 before the RXE driver
  deadlocked itself.  I've posted a bug report for that.  Lockdep spotted
  the bug right away on mount.

cifs over softIWarp RDMA got as far as generic/006 before triggering a UAF
in the reconnect code, spotted by KASAN.  Both this and the above fail on
the vanilla kernel also.

Beyond the 6 main patches, there are two patches which add some trace
points (if they look of interest to you, I can properly fill out the commit
log) and the patch to make softiwarp work - all three of which can be
dropped.

I've pushed the patches here also:

	https://git.kernel.org/pub/scm/linux/kernel/git/dhowells/linux-fs.git/log/?h=cifs-for-sfrench

David
---
David Howells (8):
      iov_iter: Add a function to extract an iter's buffers to a bvec iter
      iov_iter: Add a general purpose iteration function
      cifs: Add some helper functions
      cifs: Add a function to read into an iter from a socket
      cifs: Change the I/O paths to use an iterator rather than a page list
      cifs: Remove unused code
      cifs: Trace writedata page wrangling
      cifs: Add some RDMA send tracepoints

Namjae Jeon (1):
      cifs, ksmbd: Fix MAX_SGE count for softiwarp


 fs/cifs/cifsencrypt.c     |   40 +-
 fs/cifs/cifsfs.h          |    6 +
 fs/cifs/cifsglob.h        |   29 +-
 fs/cifs/cifsproto.h       |   11 +-
 fs/cifs/cifssmb.c         |  276 +++++--
 fs/cifs/connect.c         |   16 +
 fs/cifs/file.c            | 1434 ++++++++++++++-----------------------
 fs/cifs/fscache.c         |   22 +-
 fs/cifs/fscache.h         |   10 +-
 fs/cifs/misc.c            |  109 ---
 fs/cifs/smb2ops.c         |  367 +++++-----
 fs/cifs/smb2pdu.c         |   44 +-
 fs/cifs/smbdirect.c       |  335 ++++-----
 fs/cifs/smbdirect.h       |    6 +-
 fs/cifs/trace.h           |  161 +++++
 fs/cifs/transport.c       |   37 +-
 fs/ksmbd/transport_rdma.c |    2 +-
 include/linux/uio.h       |    8 +
 lib/iov_iter.c            |  133 ++++
 19 files changed, 1470 insertions(+), 1576 deletions(-)