The important part of qrtr_port_lookup() wrt synchronization is that the
function returns a reference counted struct qrtr_sock, or fail.
As such we need only to ensure that an decrement of the object's
refcount happens inbetween the finding of the object in the idr and
qrtr_port_lookup()'s own increment of the object.
By using RCU and putting a synchronization point after we remove the
mapping from the idr, but before it can be released we achieve this -
with the benefit of not having to hold the mutex in qrtr_port_lookup().
Change-Id: Ia9907a9c97a29e92a92e67afa988207f501a2f48
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Patch-mainline: linux-arm-kernel @ 07/05/19, 23:06
Signed-off-by: Chris Lew <clew@codeaurora.org>
Move operations on the qrtr_nodes radix tree under a separate spinlock
and make the qrtr_nodes tree GFP_ATOMIC, to allow operation from atomic
context in a subsequent patch.
Change-Id: I8f5ce15edf6043e5b27ef280f1c09f4c64ebe3bb
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Patch-mainline: linux-arm-kernel @ 07/05/19, 23:06
Signed-off-by: Chris Lew <clew@codeaurora.org>
In order to prevent overconsumption of resources on the remote side QRTR
implements a flow control mechanism.
The mechanism works by the sender keeping track of the number of
outstanding unconfirmed messages that has been transmitted to a
particular node/port pair.
Upon count reaching a low watermark (L) the confirm_rx bit is set in the
outgoing message and when the count reaching a high watermark (H)
transmission will be blocked upon the reception of a resume_tx message
from the remote, that resets the counter to 0.
This guarantees that there will be at most 2H - L messages in flight.
Values chosen for L and H are 5 and 10 respectively.
Change-Id: Ifa3c0c5821d31e385c34428d0bd4f5b5eec2fb77
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Patch-mainline: linux-arm-kernel @ 07/05/19, 23:06
[clew@codeaurora.org: Fix uninitialized qrtr_tx_lock mutex]
Signed-off-by: Chris Lew <clew@codeaurora.org>
The confirm-rx bit is used to implement a per port flow control, in
order to make sure that no messages are dropped due to resource
exhaustion. Move the resume-tx transmission to recvmsg to only confirm
messages as they are consumed by the application.
Change-Id: Icd990ef9dc46c983c9231fe1e72ba9dcbef5a0c3
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Patch-mainline: linux-arm-kernel @ 07/05/19, 23:06
Signed-off-by: Chris Lew <clew@codeaurora.org>
As the endpoint is unregistered there might still be work pending to
handle incoming messages, which will result in a use after free
scenario. The plan is to remove the rx_worker, but until then (and for
stable@) ensure that the work is stopped before the node is freed.
Change-Id: Ic35d1c0f5e5e4f297c7c05dd09c7c8a4d8b23cf4
Fixes: bdabad3e36 ("net: Add Qualcomm IPC router")
Cc: stable@vger.kernel.org
Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
Git-commit: 73f0c11d11
Git-repo: git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
Signed-off-by: Chris Lew <clew@codeaurora.org>
Support for coherent and non-coherent atomic allocations
in fastmap.
Change-Id: I3628508a4b2af5aa82d56a4a6a6ee4cef98cd47f
Signed-off-by: Sudarshan Rajagopalan <sudaraja@codeaurora.org>
[isaacm@codeaurora.org: Leverage atomic DMA pool framework code]
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Enable a check to see if devices with reserved memory regions allocate from
those regions, when using the DMA API for allocations. WARN_ON if check
doesn't pass.
Change-Id: Ic788dc4638d83cdf8a1e34f60efe678b9639d4b3
Signed-off-by: Chris Goldsworthy <cgoldswo@codeaurora.org>
Inside of dma_alloc_attrs(), verify that devices with a reserved memory
region are actually using the region, in order to prevent wastage of
memory.
Change-Id: I56852a761d3253ec2a3a7db0f6501c73983e55c6
Signed-off-by: Chris Goldsworthy <cgoldswo@codeaurora.org>
Add version check for FnR stats ioctl and error return,
since only IPA4.5+ HW will support this feature.
Change-Id: If5c5021d16bc1ee6af516d9e2200c3e6b5081d27
Signed-off-by: Bojun Pan <bojunp@codeaurora.org>
Add uc event ring to have ipa-uc monitor the
quota and bw-voting threshold to generate the
irq to ipa-driver to inform if quota or and bw
threshold is hit.
Change-Id: I7ce9c1a0048016fc6b6248a1632365f51717422c
Signed-off-by: Skylar Chang <chiaweic@codeaurora.org>
Access ipa3_ctx through ipa_utils.c instead of direct access
which is prohibited.
Change-Id: I1682037cf845d3d8553966551410ff5e8a833f84
Acked-by: Suhas Mallesh <smallesh@qti.qualcomm.com>
Signed-off-by: Michael Adisumarta <madisuma@codeaurora.org>
Adding a function to check for dangling pointer using IPA IDR structure
to avoid use after free error.
Change-Id: I0bd3d733bf10403366abc643f89c3e5c5e6228e9
Acked-by: Suhas Mallesh <smallesh@qti.qualcomm.com>
Signed-off-by: Michael Adisumarta <madisuma@codeaurora.org>
The same QMI transaction data is synchronized using two mutex locks.
Remove ineffective transaction mutex lock because it is always
covered by QMI txn_lock with same purpose.
CRs-Fixed: 2470638
Change-Id: If53e006436c77d572b2d69507e1593e3d57e1cd4
Signed-off-by: Arun Kumar Neelakantam <aneela@codeaurora.org>
Mutex lock not released before function return in failure case.
Add mutex unlock in failure case.
CRs-Fixed: 2463072
Change-Id: I68b23364e8cb1001c0a65b1b831fd3b68894e1a9
Signed-off-by: Arun Kumar Neelakantam <aneela@codeaurora.org>
Add an api for clients to set the send timeout of the underlying qrtr
socket. QMI sets this timeout to 10 seconds by default but this is too
long for some clients in critical paths.
Change-Id: I4c892a2f034c87b6bf2f71664369140910e72383
Signed-off-by: Chris Lew <clew@codeaurora.org>
txn idr entry contains local pointer, which can go out of context
if the calling function exits. Qmi handle can be released from other
thread context(SSR), where iterating over pending transactions list
if txn idr entry is not removed it results in crash in next iteration
as the previous txn idr entry is not valid.
CR-Fixed: 2373541
Change-Id: I0c96a575fb198115532de9977095f2c056742b33
Signed-off-by: Deepak Kumar Singh <deesin@codeaurora.org>
Receiving data on a socket which is getting closed by client at same time
can cause invalid pointer access.
Use sk_callback_lock to sync between data_read callback and socket close.
CRs-Fixed: 2372711
Change-Id: I98951a86a6a53bf196026a342ec4fff726bedf92
Signed-off-by: Arun Kumar Neelakantam <aneela@codeaurora.org>
of_parse_phandle() is used to deduce if a memory region is
associated with a particular device. The call to of_parse_phandle()
internally increments the reference count to the memory region's
of_node, and thus, the caller should reverse that action by
decrementing it via a call to of_node_put().
Change-Id: I80a0ed843312229f277c09fe107e407118a0a473
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Use rate-limited prints when an ION buffer should not be
mapped to userspace. This prevents the kernel logs from being
flooded by failures triggered by continuous userspace requests
to map a buffer that should not be mapped to it.
Change-Id: I5f4a1bc1fab7424588e68a55cb18aa28e969841c
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
This reverts commit b0cbb06dd1.
This commit is no longer needed, since the ION core now allows
for dmabuf ops to be overridable.
Change-Id: I32def4db2d7de97f465a82531367c78eeae8a4a1
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Since the dmabuf portion of the ION core now allows for heaps
to specify their own dmabuf ops for cache maintenance, update
the heaps so that they specify the MSM ION dmabuf ops.
Change-Id: I98d9f2fa3e3ab8288e481c72e18e5d2f14a8a400
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Update the heap IDs as per the standard imposed by the ION
core (i.e.):
SYSTEM_HEAP: (1 << 0)
SYSTEM_CONTIG_HEAP: (1 << 1)
CARVEOUT_HEAPS: [(1 << 2), (1 << 6)]
CHUNK_HEAP: (1 << 7)
CMA_HEAPS: [(1 << 8), (1 << 15)]
VENDOR_HEAPS: [(1 << 16), (1 << 31)]
Since the heap IDs are now bitmasks, as opposed to offsets
within a bitmask, update the userspace ION_HEAP() macro/as
well as the heap ID treatment in the kernel to account for
this as well.
Change-Id: I6b7ca1caf89dda03dfb499a362db863168c2753c
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Add a reference to the MSM ION heap ID header as part of the
dt-bindings headers, so that the device-tree can reference
the heap IDs via the defined macros.
Change-Id: I7fd7e45e8a61908b037a421b08f4ba8b4ddebfbf
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
The ION heap IDs are also needed by the device-tree to
provide the kernel with the heap ID for the heap that
a device-tree entry describes. Currently, these IDs are
hard-coded, which can be error-prone when updating the
heap IDs.
Thus, create a header that can also be accessed by the
device-tree and userspace, and move the heap IDs to that
header. This makes it so that the device-tree can now use
the constants defined in the header, instead of hard-coded
IDs, making updates significantly easier.
Change-Id: I58b1aae9924504f2b20e159b9be8da7a2ee8d248
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
Introduce a new domain for wakeup capable GPIOs. The domain can be
requested using the bus token DOMAIN_BUS_WAKEUP. In the following
patches, we will specify PDC as the wakeup-parent for the TLMM GPIO
irqchip. Requesting a wakeup GPIO will setup the GPIO and the
corresponding PDC interrupt as its parent.
Change-Id: I13b206773e5f04452eed87ad1490da50b034ac4c
Co-developed-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Signed-off-by: Lina Iyer <ilina@codeaurora.org>
Change the if condition to be inclusive to allow remove_offload
ioctl to send the rmv_offload_request through qmi on the last
connection.
Change-Id: I6eec7799fd1a76ab168bc588b3e9c74dbfd215fd
Signed-off-by: Michael Adisumarta <madisuma@codeaurora.org>
Remove code check-in not belonging to IPA driver.
Change-Id: I6f79d72d054c7c3b9d5448ae83150d5afaedacff
Signed-off-by: Bojun Pan <bojunp@codeaurora.org>
Remove the cpu affinity of ion pool refill kthreads which were
binded to min capability cpus, since there was no observable
power impact with when not affined.
Note: It was found that the code to create cpumask for min
capability cpus was broken and kthreads were free to run on
any cpus. No observable power impact was seen, hence better
off with removing this code entirely.
Change-Id: If6a32d9e4d3c576e63e3b0ec405be1fd4639a5e0
Signed-off-by: Sudarshan Rajagopalan <sudaraja@codeaurora.org>
Signed-off-by: Isaac J. Manjarres <isaacm@codeaurora.org>
