28 Mar, 2018
4 commits
-
It is possible that incoming data arrives before the client driver has
reached a point in the probe method where adequate context for handling
the incoming message has been established.In the event that the client's callback function returns an error the
message will be left on the FIFO and by invoking the receive handler
after the device has been probed the message will be picked off the FIFO
and the callback invoked again.Signed-off-by: Bjorn Andersson
-
For special rpmsg devices without a primary endpoint there is nothing to
announce so don't call the backend announce create function if we didn't
create an endpoint.Signed-off-by: Bjorn Andersson
-
The container_of macros should not use the same name for the parameter
as the member to use for lookup, as this will result in a compilation
error unless the passed parameter has the same name as the member.Signed-off-by: Bjorn Andersson
-
In an effort to pick up channels that are in a funky state we
optimistically tried to open all channels that we found, with the
addition that we failed if the other side did not handshake the opening.But as we're starting the modem a second time all channels are found -
in a "funky" state - and we try to open them. But the modem firmware
requires the IPCRTR to be up in order to initialize. So any channels we
try to open before that will fail and will not be opened again.This takes care of the regression, at the cost of reintroducing the
previous behavior of handling of channels with "funky" states.Reverts commit c12fc4519f60 ("rpmsg: smd: Create device for all channels")
Reported-by: Srinivas Kandagatla
Signed-off-by: Bjorn Andersson
20 Mar, 2018
2 commits
-
Switch the tx_lock to a spinlock we allow clients to use rpmsg_trysend()
from atomic context.In order to allow clients to sleep while waiting for space in the FIFO
we release the lock temporarily around the delay; which should be
replaced by sending a READ_NOTIF and waiting for the remote to signal
us that space has been made available.Tested-by: Srinivas Kandagatla
Signed-off-by: Bjorn Andersson -
By switching the tx_lock to a spinlock we allow clients to use
rpmsg_trysend() from atomic context.The mutex was interruptable as it was previously held for the duration
of some client waiting for available space in the FIFO, but this was
recently changed to only be held temporarily - allowing us to replace it
with a spinlock.Tested-by: Srinivas Kandagatla
Signed-off-by: Bjorn Andersson
18 Mar, 2018
2 commits
-
if device_register() returned an error! Always use put_device()
to give up the reference initialized. unregister device for
other return error.Signed-off-by: Arvind Yadav
Signed-off-by: Bjorn Andersson -
if device_register() returned an error! Always use put_device()
to give up the reference initialized. And unregister device for
other return error.Signed-off-by: Arvind Yadav
Signed-off-by: Bjorn Andersson
12 Feb, 2018
1 commit
-
This is the mindless scripted replacement of kernel use of POLL*
variables as described by Al, done by this script:for V in IN OUT PRI ERR RDNORM RDBAND WRNORM WRBAND HUP RDHUP NVAL MSG; do
L=`git grep -l -w POLL$V | grep -v '^t' | grep -v /um/ | grep -v '^sa' | grep -v '/poll.h$'|grep -v '^D'`
for f in $L; do sed -i "-es/^\([^\"]*\)\(\\)/\\1E\\2/" $f; done
donewith de-mangling cleanups yet to come.
NOTE! On almost all architectures, the EPOLL* constants have the same
values as the POLL* constants do. But they keyword here is "almost".
For various bad reasons they aren't the same, and epoll() doesn't
actually work quite correctly in some cases due to this on Sparc et al.The next patch from Al will sort out the final differences, and we
should be all done.Scripted-by: Al Viro
Signed-off-by: Linus Torvalds
06 Feb, 2018
1 commit
-
Pull rpmsg updates from Bjorn Andersson:
"This fixes a few issues found in the SMD and GLINK drivers and
corrects the handling of SMD channels that are found in an
(previously) unexpected state"* tag 'rpmsg-v4.16' of git://github.com/andersson/remoteproc:
rpmsg: smd: Fix double unlock in __qcom_smd_send()
rpmsg: glink: Fix missing mutex_init() in qcom_glink_alloc_channel()
rpmsg: smd: Don't hold the tx lock during wait
rpmsg: smd: Fail send on a closed channel
rpmsg: smd: Wake up all waiters
rpmsg: smd: Create device for all channels
rpmsg: smd: Perform handshake during open
rpmsg: glink: smem: Ensure ordering during tx
drivers: rpmsg: remove duplicate includes
remoteproc: qcom: Use PTR_ERR_OR_ZERO() in glink prob
19 Jan, 2018
1 commit
-
We're not holding the lock here, so we shouldn't unlock.
Fixes: 178f3f75bb4e ("rpmsg: smd: Don't hold the tx lock during wait")
Signed-off-by: Dan Carpenter
[bjorn: renamed "out" label to further distinguish the two exit paths]
Signed-off-by: Bjorn Andersson
03 Jan, 2018
1 commit
-
qcom_glink_alloc_channel() allocates the mutex but not initialize it.
Use mutex_init() on it to initialize it correctly.This is detected by Coccinelle semantic patch.
Signed-off-by: Wei Yongjun
Signed-off-by: Bjorn Andersson
19 Dec, 2017
6 commits
-
Holding the tx lock while waiting for tx-drain events from the remote
side blocks try_send requests from failing quickly, so temporarily drop
the tx lock while waiting.While this allows try_send to fail quickly it also could allow a
subsequent send to succeed putting a smaller packet in the FIFO while
we're waiting for room for our large packet. But as this lock is per
channel we expect that clients with ordering concerns implements their
own ordering mechanism.Signed-off-by: Bjorn Andersson
-
Move the check for a closed channel out from the tx-full loop to fail
any send request on a non-open channel.Signed-off-by: Bjorn Andersson
-
It's possible to have multiple contexts waiting for new channel events
and with an upcoming change it's possible to have multiple contexts
waiting for a full FIFO. As such we need to wake them all up.Signed-off-by: Bjorn Andersson
-
Rather than selectively creating devices only for the channels that the
remote have moved to "opening" state let's create devices for all
channels found. The driver model will match drivers to the ones we care
about and attempt to open these.The one case where this fails is if the user loads a firmware that lacks
a particular channel of the previous firmware that was running, in which
case we would find the old channel and attempt to probe it. The channel
opening handshake will ensure this will result in a graceful failure.The result of this patch is that we will actively open the RPM channel
even though it's left in a state other than "opening" after the boot
loader's closing of the channel.Tested-by: Will Newton
Reported-by: Jeremy McNicoll
Reported-by: Will Newton
Signed-off-by: Bjorn Andersson -
Validate the the remote side is opening the channel that we've found by
performing a handshake when opening the channel.Tested-by: Will Newton
Signed-off-by: Bjorn Andersson -
Ensure the ordering of the fifo write and the update of the write index,
so that the index is not updated before the data has landed in the fifo.Acked-By: Chris Lew
Reported-by: Arun Kumar Neelakantam
Signed-off-by: Bjorn Andersson
07 Dec, 2017
1 commit
-
These duplicate includes have been found with scripts/checkincludes.pl but
they have been removed manually to avoid removing false positives.Signed-off-by: Pravin Shedge
Signed-off-by: Bjorn Andersson
29 Nov, 2017
1 commit
-
Signed-off-by: Al Viro
18 Nov, 2017
1 commit
-
Pull rpmsg updates from Bjorn Andersson:
- turn RPMSG_VIRTIO into a user selectable config
- fix few bugs in GLINK
- provide the support for specifying initial buffer sizes for GLINK
channels.* tag 'rpmsg-v4.15' of git://github.com/andersson/remoteproc:
rpmsg: glink: The mbox client knows_txdone
rpmsg: glink: Add missing MODULE_LICENSE
rpmsg: glink: Use best fit intent during tx
rpmsg: glink: Add support to preallocate intents
dt-bindings: soc: qcom: Support GLINK intents
rpmsg: glink: Initialize the "intent_req_comp" completion variable
rpmsg: Allow RPMSG_VIRTIO to be enabled via menuconfig or defconfig
16 Nov, 2017
1 commit
-
As the GLINK driver is ticking the txdone of the mailbox channel (to
implement the doorbell) it needs to set knows_txdone.Signed-off-by: Bjorn Andersson
15 Nov, 2017
1 commit
-
The qcom_glink_native driver is missing a MODULE_LICENSE(), correct
this.Fixes: 835764ddd9af ("rpmsg: glink: Move the common glink protocol implementation to glink_native.c")
Cc: stable@vger.kernel.org
Reported-by: Randy Dunlap
Signed-off-by: Bjorn Andersson
02 Nov, 2017
1 commit
-
Many source files in the tree are missing licensing information, which
makes it harder for compliance tools to determine the correct license.By default all files without license information are under the default
license of the kernel, which is GPL version 2.Update the files which contain no license information with the 'GPL-2.0'
SPDX license identifier. The SPDX identifier is a legally binding
shorthand, which can be used instead of the full boiler plate text.This patch is based on work done by Thomas Gleixner and Kate Stewart and
Philippe Ombredanne.How this work was done:
Patches were generated and checked against linux-4.14-rc6 for a subset of
the use cases:
- file had no licensing information it it.
- file was a */uapi/* one with no licensing information in it,
- file was a */uapi/* one with existing licensing information,Further patches will be generated in subsequent months to fix up cases
where non-standard license headers were used, and references to license
had to be inferred by heuristics based on keywords.The analysis to determine which SPDX License Identifier to be applied to
a file was done in a spreadsheet of side by side results from of the
output of two independent scanners (ScanCode & Windriver) producing SPDX
tag:value files created by Philippe Ombredanne. Philippe prepared the
base worksheet, and did an initial spot review of a few 1000 files.The 4.13 kernel was the starting point of the analysis with 60,537 files
assessed. Kate Stewart did a file by file comparison of the scanner
results in the spreadsheet to determine which SPDX license identifier(s)
to be applied to the file. She confirmed any determination that was not
immediately clear with lawyers working with the Linux Foundation.Criteria used to select files for SPDX license identifier tagging was:
- Files considered eligible had to be source code files.
- Make and config files were included as candidates if they contained >5
lines of source
- File already had some variant of a license header in it (even if
Reviewed-by: Philippe Ombredanne
Reviewed-by: Thomas Gleixner
Signed-off-by: Greg Kroah-Hartman
31 Oct, 2017
3 commits
-
Intents can vary in size, try to find the best fitting remote intent
instead of first fit when sending a message to the remote proc.Signed-off-by: Chris Lew
Signed-off-by: Bjorn Andersson -
The base intents prequeued during channel creation may not satisfy a
channel's throughput requirement. Add support for intents dt-binding to
allow channels to specify the size and amount of intents to prequeue
during endpoint announcement.Signed-off-by: Chris Lew
[bjorn: Altered how defaults are expressed]
Signed-off-by: Bjorn Andersson -
The "intent_req_comp" variable is used without initialization which
results in NULL pointer dereference in qcom_glink_request_intent().we need to initialize the completion variable before using it.
Fixes: 27b9c5b66b23 ("rpmsg: glink: Request for intents when unavailable")
Signed-off-by: Arun Kumar Neelakantam
Signed-off-by: Bjorn Andersson
26 Oct, 2017
1 commit
-
Currently, RPMSG_VIRTIO can only be enabled if some other kconfig
option selects it. This does not allow it to be enabled for
virtualized systems where Virtio RPMSG is available over Virtio
MMIO or PCI transport.This patch updates RPMSG_VIRTIO kconfig option so that we can
enable the VirtIO RPMSG driver via menuconfig or defconfig. The
patch also removes "select RPMSG_VIRTIO" from various remoteproc
kconfig options because it is now user selectable.Signed-off-by: Anup Patel
Signed-off-by: Bjorn Andersson
11 Oct, 2017
2 commits
-
We need to free "intent" and "intent->data" on a couple error paths.
Fixes: 933b45da5d1d ("rpmsg: glink: Add support for TX intents")
Acked-by: Sricharan R
Signed-off-by: Dan Carpenter
Signed-off-by: Bjorn Andersson -
If qcom_glink_tx() fails, then we need to unlock before returning the
error code.Fixes: 27b9c5b66b23 ("rpmsg: glink: Request for intents when unavailable")
Acked-by: Sricharan R
Signed-off-by: Dan Carpenter
Signed-off-by: Bjorn Andersson
05 Sep, 2017
2 commits
-
The new switch cases for RPM_CMD_RX_DONE, RPM_CMD_RX_DONE_W_REUSE,
RPM_CMD_RX_INTENT_REQ_ACK, RPM_CMD_INTENT and RPM_CMD_RX_INTENT_REQ from
4 recent commits are not setting ret and so a later non-zero check on ret
is testing on a garbage value in ret. Fix this by initializing ret to zero.Detected by CoverityScan CID#1455249 ("Uninitialized scalar variable")
Fixes: 933b45da5d1d ("rpmsg: glink: Add support for TX intents)
Fixes: dacbb35e930f ("glink: Receive and store the remote intent buffers")
Fixes: 27b9c5b66b23 ("rpmsg: glink: Request for intents when unavailable")
Fixes: 88c6060f5a7f ("rpmsg: glink: Handle remote rx done command")
Signed-off-by: Colin Ian King
Signed-off-by: Bjorn Andersson -
In the case where glink->intentless is true and the call
to qcom_glink_tx fails then we have a condition where ret is
non-zero and intent is null, causing a null pointer deference
when setting intent->in_use to false. Add an extra check to
only dereference intent if intent is non-null.Detected by: CoverityScan CID#1455247 ("Explicit null dereferenced")
Fixes: 88c6060f5a7f ("rpmsg: glink: Handle remote rx done command")
Signed-off-by: Colin Ian King
Signed-off-by: Bjorn Andersson
01 Sep, 2017
1 commit
-
The common code needs to export the probe and remove symbols in order
for the SMEM and RPM drivers to access them when compiled as a module.Signed-off-by: Bjorn Andersson
31 Aug, 2017
1 commit
-
The idr_lock should be released in the case that we don't find the given
channel.Fixes: 44f6df922a26 ("rpmsg: glink: Fix idr_lock from mutex to spinlock")
Reported-by: Julia Lawall
Signed-off-by: Bjorn Andersson
30 Aug, 2017
6 commits
-
Once the remote side sends a rx done ack, check for the intent reuse
information from it and suitably discard or reuse the remote passed
intent buffers.Acked-by: Arun Kumar Neelakantam
Signed-off-by: Sricharan R
Signed-off-by: Bjorn Andersson -
While sending data, we search for suitable sized intent to map and
simply fail if a intent is not found. Instead request for a intent of
required size and wait till one is alloted.Acked-by: Arun Kumar Neelakantam
Signed-off-by: Sricharan R
Signed-off-by: Bjorn Andersson -
While sending data, use the remote intent id buffer of suitable size
that was passed by remote previously.Acked-by: Arun Kumar Neelakantam
Signed-off-by: Sricharan R
Signed-off-by: Bjorn Andersson -
Just like we allocating and sending intent ids to remote, remote side
allocates and sends us the intents as well. So save the intent ids and
use it later while sending data targeting the appropriate intents based
on the size.Acked-by: Arun Kumar Neelakantam
Signed-off-by: Sricharan R
Signed-off-by: Bjorn Andersson -
Preallocate local intent buffers and pass the intent ids to the remote.
This way there are some default intents available for the remote to
start sending data without having to wait by sending intent requests. Do
this by adding the rpmsg announce_create ops, which gets called right
after the rpmsg device gets probed.Acked-by: Arun Kumar Neelakantam
Signed-off-by: Sricharan R
Signed-off-by: Bjorn Andersson -
Send RX data receive ack to remote and also inform that local intent
buffer is used and freed. This informs the remote to request for next
set of intent buffers before doing a send operation.Acked-by: Arun Kumar Neelakantam
Signed-off-by: Sricharan R
Signed-off-by: Bjorn Andersson
