A short read may occur while reading the message footer from the
socket. Later, when the socket is ready for another read, the
messenger invokes all read_partial_*() handlers, including
read_partial_sparse_msg_data(). The expectation is that
read_partial_sparse_msg_data() would bail, allowing the messenger to
invoke read_partial() for the footer and pick up where it left off.
However read_partial_sparse_msg_data() violates that and ends up
calling into the state machine in the OSD client. The sparse-read
state machine assumes that it's a new op and interprets some piece of
the footer as the sparse-read header and returns bogus extents/data
length, etc.
To determine whether read_partial_sparse_msg_data() should bail, let's
reuse cursor->total_resid. Because once it reaches to zero that means
all the extents and data have been successfully received in last read,
else it could break out when partially reading any of the extents and
data. And then osd_sparse_read() could continue where it left off.
[ idryomov: changelog ]
Link: https://tracker.ceph.com/issues/63586
Fixes: d396f89db3 ("libceph: add sparse read support to msgr1")
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
These functions are supposed to behave like other read_partial_*()
handlers: the contract with messenger v1 is that the handler bails if
the area of the message it's responsible for is already processed.
This comes up when handling short reads from the socket.
[ idryomov: changelog ]
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Acked-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Add 2 new fields to ceph_connection_v1_info to track the necessary info
in sparse reads. Skip initializing the cursor for a sparse read.
Break out read_partial_message_section into a wrapper around a new
read_partial_message_chunk function that doesn't zero out the crc first.
Add new helper functions to drive receiving into the destinations
provided by the sparse_read state machine.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Reviewed-and-tested-by: Luís Henriques <lhenriques@suse.de>
Reviewed-by: Milind Changire <mchangir@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Fix the mishandling of MSG_DONTWAIT and also reinstates the per-page
checking of the source pages (which might have come from a DIO write by
userspace) by partially reverting the changes to support MSG_SPLICE_PAGES
and doing things a little differently. In messenger_v1:
(1) The ceph_tcp_sendpage() is resurrected and the callers reverted to use
that.
(2) The callers now pass MSG_MORE unconditionally. Previously, they were
passing in MSG_MORE|MSG_SENDPAGE_NOTLAST and then degrading that to
just MSG_MORE on the last call to ->sendpage().
(3) Make ceph_tcp_sendpage() a wrapper around sendmsg() rather than
sendpage(), setting MSG_SPLICE_PAGES if sendpage_ok() returns true on
the page.
In messenger_v2:
(4) Bring back do_try_sendpage() and make the callers use that.
(5) Make do_try_sendpage() use sendmsg() for both cases and set
MSG_SPLICE_PAGES if sendpage_ok() is set.
Fixes: 40a8c17aa7 ("ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage")
Fixes: fa094ccae1 ("ceph: Use sendmsg(MSG_SPLICE_PAGES) rather than sendpage()")
Reported-by: Ilya Dryomov <idryomov@gmail.com>
Link: https://lore.kernel.org/r/CAOi1vP9vjLfk3W+AJFeexC93jqPaPUn2dD_4NrzxwoZTbYfOnw@mail.gmail.com/
Link: https://lore.kernel.org/r/CAOi1vP_Bn918j24S94MuGyn+Gxk212btw7yWeDrRcW1U8pc_BA@mail.gmail.com/
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Xiubo Li <xiubli@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
Link: https://lore.kernel.org/r/3101881.1687801973@warthog.procyon.org.uk/ # v1
Link: https://lore.kernel.org/r/3111635.1687813501@warthog.procyon.org.uk/ # v2
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Link: https://lore.kernel.org/r/3199652.1687873788@warthog.procyon.org.uk
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Use sendmsg() and MSG_SPLICE_PAGES rather than sendpage in ceph when
transmitting data. For the moment, this can only transmit one page at a
time because of the architecture of net/ceph/, but if
write_partial_message_data() can be given a bvec[] at a time by the
iteration code, this would allow pages to be sent in a batch.
Signed-off-by: David Howells <dhowells@redhat.com>
cc: Ilya Dryomov <idryomov@gmail.com>
cc: Xiubo Li <xiubli@redhat.com>
cc: Jeff Layton <jlayton@kernel.org>
cc: Jens Axboe <axboe@kernel.dk>
cc: Matthew Wilcox <willy@infradead.org>
Link: https://lore.kernel.org/r/20230623225513.2732256-4-dhowells@redhat.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
READ/WRITE proved to be actively confusing - the meanings are
"data destination, as used with read(2)" and "data source, as
used with write(2)", but people keep interpreting those as
"we read data from it" and "we write data to it", i.e. exactly
the wrong way.
Call them ITER_DEST and ITER_SOURCE - at least that is harder
to misinterpret...
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
ceph_msg_data_next is always passed a NULL pointer for this field. Some
of the "next" operations look at it in order to determine the length,
but we can just take the min of the data on the page or cursor->resid.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Both msgr1 and msgr2 in crc mode are zero copy in the sense that
message data is read from the socket directly into the destination
buffer. We assume that the destination buffer is stable (i.e. remains
unchanged while it is being read to) though. Otherwise, CRC errors
ensue:
libceph: read_partial_message 0000000048edf8ad data crc 1063286393 != exp. 228122706
libceph: osd1 (1)192.168.122.1:6843 bad crc/signature
libceph: bad data crc, calculated 57958023, expected 1805382778
libceph: osd2 (2)192.168.122.1:6876 integrity error, bad crc
Introduce rxbounce option to enable use of a bounce buffer when
receiving message data. In particular this is needed if a mapped
image is a Windows VM disk, passed to QEMU. Windows has a system-wide
"dummy" page that may be mapped into the destination buffer (potentially
more than once into the same buffer) by the Windows Memory Manager in
an effort to generate a single large I/O [1][2]. QEMU makes a point of
preserving overlap relationships when cloning I/O vectors, so krbd gets
exposed to this behaviour.
[1] "What Is Really in That MDL?"
https://docs.microsoft.com/en-us/previous-versions/windows/hardware/design/dn614012(v=vs.85)
[2] https://blogs.msmvps.com/kernelmustard/2005/05/04/dummy-pages/
URL: https://bugzilla.redhat.com/show_bug.cgi?id=1973317
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
This line dates back to 2013, but cppcheck complained because commit
2f713615dd ("libceph: move msgr1 protocol implementation to its own
file") moved it. Add parenthesis to silence the warning.
Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
A pure move, no other changes.
Note that ceph_tcp_recv{msg,page}() and ceph_tcp_send{msg,page}()
helpers are also moved. msgr2 will bring its own, more efficient,
variants based on iov_iter. Switching msgr1 to them was considered
but decided against to avoid subtle regressions.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
