When the client sends a regular blocking accept request, the backend is
expected to return only when the accept is completed, simulating a
blocking behavior, or return an error.
Specifically, on EAGAIN from inet_accept, the backend shouldn't return
"EAGAIN" to the client. Instead, it should simply continue the wait.
Otherwise, the client will send another accept request, which will cause
another EAGAIN to be sent back, which is a waste of resources and not
conforming to the expected behavior. Change the behavior by turning the
"goto error" into a return.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Signed-off-by: Juergen Gross <jgross@suse.com>
Pull iov_iter updates from Al Viro:
- bio_{map,copy}_user_iov() series; those are cleanups - fixes from the
same pile went into mainline (and stable) in late September.
- fs/iomap.c iov_iter-related fixes
- new primitive - iov_iter_for_each_range(), which applies a function
to kernel-mapped segments of an iov_iter.
Usable for kvec and bvec ones, the latter does kmap()/kunmap() around
the callback. _Not_ usable for iovec- or pipe-backed iov_iter; the
latter is not hard to fix if the need ever appears, the former is by
design.
Another related primitive will have to wait for the next cycle - it
passes page + offset + size instead of pointer + size, and that one
will be usable for everything _except_ kvec. Unfortunately, that one
didn't get exposure in -next yet, so...
- a bit more lustre iov_iter work, including a use case for
iov_iter_for_each_range() (checksum calculation)
- vhost/scsi leak fix in failure exit
- misc cleanups and detritectomy...
* 'work.iov_iter' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (21 commits)
iomap_dio_actor(): fix iov_iter bugs
switch ksocknal_lib_recv_...() to use of iov_iter_for_each_range()
lustre: switch struct ksock_conn to iov_iter
vhost/scsi: switch to iov_iter_get_pages()
fix a page leak in vhost_scsi_iov_to_sgl() error recovery
new primitive: iov_iter_for_each_range()
lnet_return_rx_credits_locked: don't abuse list_entry
xen: don't open-code iov_iter_kvec()
orangefs: remove detritus from struct orangefs_kiocb_s
kill iov_shorten()
bio_alloc_map_data(): do bmd->iter setup right there
bio_copy_user_iov(): saner bio size calculation
bio_map_user_iov(): get rid of copying iov_iter
bio_copy_from_iter(): get rid of copying iov_iter
move more stuff down into bio_copy_user_iov()
blk_rq_map_user_iov(): move iov_iter_advance() down
bio_map_user_iov(): get rid of the iov_for_each()
bio_map_user_iov(): move alignment check into the main loop
don't rely upon subsequent bio_add_pc_page() calls failing
... and with iov_iter_get_pages_alloc() it becomes even simpler
...
Since commit ba1029c9cb ("modpost: detect modules without a
MODULE_LICENSE") modules without said macro will generate
WARNING: modpost: missing MODULE_LICENSE() in <filename>
While at it, also add module description and attribution.
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
Acked-by: Stefano Stabellini <sstabellini@kernel.org>
__WARN() is an internal helper that is only available on
some architectures, but causes a build error e.g. on ARM64
in some configurations:
drivers/xen/pvcalls-back.c: In function 'set_backend_state':
drivers/xen/pvcalls-back.c:1097:5: error: implicit declaration of function '__WARN' [-Werror=implicit-function-declaration]
Unfortunately, there is no equivalent of BUG() that takes no
arguments, but WARN_ON(1) is commonly used in other drivers
and works on all configurations.
Fixes: 7160378206b2 ("xen/pvcalls: xenbus state handling")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Stefano Stabellini <sstabellini@kernel.org>
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
When the other end notifies us that there is data to be written
(pvcalls_back_conn_event), increment the io and write counters, and
schedule the ioworker.
Implement the write function called by ioworker by reading the data from
the data ring, writing it to the socket by calling inet_sendmsg.
Set out_error on error.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
When an active socket has data available, increment the io and read
counters, and schedule the ioworker.
Implement the read function by reading from the socket, writing the data
to the data ring.
Set in_error on error.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
We have one ioworker per socket. Each ioworker goes through the list of
outstanding read/write requests. Once all requests have been dealt with,
it returns.
We use one atomic counter per socket for "read" operations and one
for "write" operations to keep track of the reads/writes to do.
We also use one atomic counter ("io") per ioworker to keep track of how
many outstanding requests we have in total assigned to the ioworker. The
ioworker finishes when there are none.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Release both active and passive sockets. For active sockets, make sure
to avoid possible conflicts with the ioworker reading/writing to those
sockets concurrently. Set map->release to let the ioworker know
atomically that the socket will be released soon, then wait until the
ioworker finishes (flush_work).
Unmap indexes pages and data rings.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Implement poll on passive sockets by requesting a delayed response with
mappass->reqcopy, and reply back when there is data on the passive
socket.
Poll on active socket is unimplemented as by the spec, as the frontend
should just wait for events and check the indexes on the indexes page.
Only support one outstanding poll (or accept) request for every passive
socket at any given time.
[ boris: fixed long lines ]
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Implement the accept command by calling inet_accept. To avoid blocking
in the kernel, call inet_accept(O_NONBLOCK) from a workqueue, which get
scheduled on sk_data_ready (for a passive socket, it means that there
are connections to accept).
Use the reqcopy field to store the request. Accept the new socket from
the delayed work function, create a new sock_mapping for it, map
the indexes page and data ring, and reply to the other end. Allocate an
ioworker for the socket.
Only support one outstanding blocking accept request for every socket at
any time.
Add a field to sock_mapping to remember the passive socket from which an
active socket was created.
[ boris: fixed whitespaces ]
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Allocate a socket. Track the allocated passive sockets with a new data
structure named sockpass_mapping. It contains an unbound workqueue to
schedule delayed work for the accept and poll commands. It also has a
reqcopy field to be used to store a copy of a request for delayed work.
Reads/writes to it are protected by a lock (the "copy_lock" spinlock).
Initialize the workqueue in pvcalls_back_bind.
Implement the bind command with inet_bind.
The pass_sk_data_ready event handler will be added later.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Allocate a socket. Keep track of socket <-> ring mappings with a new data
structure, called sock_mapping. Implement the connect command by calling
inet_stream_connect, and mapping the new indexes page and data ring.
Allocate a workqueue and a work_struct, called ioworker, to perform
reads and writes to the socket.
When an active socket is closed (sk_state_change), set in_error to
-ENOTCONN and notify the other end, as specified by the protocol.
sk_data_ready and pvcalls_back_ioworker will be implemented later.
[ boris: fixed whitespaces ]
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Just reply with success to the other end for now. Delay the allocation
of the actual socket to bind and/or connect.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
When the other end notifies us that there are commands to be read
(pvcalls_back_event), wake up the backend thread to parse the command.
The command ring works like most other Xen rings, so use the usual
ring macros to read and write to it. The functions implementing the
commands are empty stubs for now.
[ boris: fixed whitespaces ]
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Introduce a per-frontend data structure named pvcalls_fedata. It
contains pointers to the command ring, its event channel, a list of
active sockets and a tree of passive sockets (passing sockets need to be
looked up from the id on listen, accept and poll commands, while active
sockets only on release).
It also has an unbound workqueue to schedule the work of parsing and
executing commands on the command ring. socket_lock protects the two
lists. In pvcalls_back_global, keep a list of connected frontends.
[ boris: fixed whitespaces/long lines ]
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Introduce the code to handle xenbus state changes.
Implement the probe function for the pvcalls backend. Write the
supported versions, max-page-order and function-calls nodes to xenstore,
as required by the protocol.
Introduce stub functions for disconnecting/connecting to a frontend.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Keep a list of connected frontends. Use a semaphore to protect list
accesses.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Introduce a xenbus backend for the pvcalls protocol, as defined by
https://xenbits.xen.org/docs/unstable/misc/pvcalls.html.
This patch only adds the stubs, the code will be added by the following
patches.
Signed-off-by: Stefano Stabellini <stefano@aporeto.com>
Reviewed-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
Reviewed-by: Juergen Gross <jgross@suse.com>
CC: boris.ostrovsky@oracle.com
CC: jgross@suse.com
Signed-off-by: Boris Ostrovsky <boris.ostrovsky@oracle.com>
