- 08 Feb, 2016 12 commits
-
-
Vitaly Kuznetsov authored
We have to call vmbus_initiate_unload() on crash to make kdump work but the crash can also be happening in interrupt (e.g. Sysrq + c results in such) where we can't schedule or the following will happen: [ 314.905786] bad: scheduling from the idle thread! Just skipping the wait (and even adding some random wait here) won't help: to make host-side magic working we're supposed to receive CHANNELMSG_UNLOAD (and actually confirm the fact that we received it) but we can't use interrupt-base path (vmbus_isr()-> vmbus_on_msg_dpc()). Implement a simple busy wait ignoring all the other messages and use it if we're in an interrupt context. Signed-off-by:
Vitaly Kuznetsov <vkuznets@redhat.com> Signed-off-by:
K. Y. Srinivasan <kys@microsoft.com> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Vitaly Kuznetsov authored
When we pick a CPU to use for a new subchannel we try find a non-used one on the appropriate NUMA node, we keep track of them with the primary->alloced_cpus_in_node mask. Under normal circumstances we don't run out of available CPUs but it is possible when we we don't initialize some cpus in Linux, e.g. when we boot with 'nr_cpus=' limitation. Avoid the infinite loop in init_vp_index() by checking that we still have non-used CPUs in the alloced_cpus_in_node mask and resetting it in case we don't. Signed-off-by:
-
Kamal Mostafa authored
Use the local uapi headers to keep in sync with "recently" added #define's (e.g. VSS_OP_REGISTER1). Fixes: 3eb2094c ("Adding makefile for tools/hv") Cc: <stable@vger.kernel.org> Signed-off-by:
Kamal Mostafa <kamal@canonical.com> Signed-off-by:
-
K. Y. Srinivasan authored
Add vendor and device attributes to VMBUS devices. These will be used by Hyper-V tools as well user-level RDMA libraries that will use the vendor/device tuple to discover the RDMA device. Signed-off-by:
-
K. Y. Srinivasan authored
Cleanup vmbus_set_event() by inlining the hypercall to post the event and since the return value of vmbus_set_event() is not checked, make it void. As part of this cleanup, get rid of the function hv_signal_event() as it is only callled from vmbus_set_event(). Signed-off-by:
-
Dan Carpenter authored
We recently messed up the error handling here. We can return with the pipe->lock held or sometimes we unlock twice by mistake. Fixes: 2f3be882 ('goldfish_pipe: Pin pages to memory while copying and other cleanups') Signed-off-by:
Dan Carpenter <dan.carpenter@oracle.com> Reviewed-by:
Christoffer Dall <christoffer.dall@linaro.org> Signed-off-by:
-
Eric Biggers authored
The error code passed to ERR_PTR() always should be negated. Also, the return value of scif_add_mmu_notifier() was never checked. Signed-off-by:
Eric Biggers <ebiggers3@gmail.com> Reviewed-by:
Sudeep Dutt <sudeep.dutt@intel.com> Signed-off-by:
-
Geliang Tang authored
list_next_entry has been defined in list.h, so I replace list_entry_next with it. Signed-off-by:
Geliang Tang <geliangtang@163.com> Reviewed-by:
-
Dan Carpenter authored
Signed integer overflow is undefined. Also I added a check for "(offset < 0)" in scif_unregister() because that makes it match the other conditions and because I didn't want to subtract a negative. Fixes: ba612aa8 ('misc: mic: SCIF memory registration and unregistration') Signed-off-by:
-
Christian Colic authored
Rewrite comparisons to NULL "data->reg_base == NULL" as "!data->reg_base" to conform to checkpatch. Signed-off-by:
Christian Colic <colic.christian@gmail.com> Signed-off-by:
-
Christian Colic authored
Added a space around the "+" at: "reg_base+addr2" to clear up a checkpatch check. Signed-off-by:
-
Arnd Bergmann authored
The goldfish nand driver divides a signed 64-bit number (loff_t) in multiple places using the do_div() function. This has always been unreliable but now produces a compiler warning (since 4.5-rc1): goldfish/goldfish_nand.c: In function 'goldfish_nand_erase': goldfish/goldfish_nand.c:107:91: error: comparison of distinct pointer types lacks a cast [-Werror] goldfish/goldfish_nand.c: In function 'goldfish_nand_read_oob': goldfish/goldfish_nand.c:145:91: error: comparison of distinct pointer types lacks a cast [-Werror] This changes the code to the equivalent div_s64{,_rem} that works correctly for negative numbers (which we should never get here). The warning has shown up on ARM allmodconfig builds after the goldfish bus driver has become visible on ARM. Signed-off-by:Arnd Bergmann <arnd@arndb.de> Fixes: bd2f348d ("goldfish: refactor goldfish platform configs") Acked-by:
Alan Cox <alan@linux.intel.com> Signed-off-by:
-
- 07 Feb, 2016 28 commits
-
-
Alexander Usyskin authored
The amthif FW client can appear after the end of client enumeration. Amthif host client initialization is done now at FW client discovery time. Signed-off-by:
Alexander Usyskin <alexander.usyskin@intel.com> Signed-off-by:
Tomas Winkler <tomas.winkler@intel.com> Signed-off-by:
-
Alexander Usyskin authored
Signal the FW that it can send an HBM enumeration answer immediately, without waiting for FW initialization completion, meaning before all the FW clients are ready and registered. Organize enumeration response options to enum as a byproduct. Signed-off-by:
-
Alexander Usyskin authored
Since clients can be now added and removed during runtime we need to run bus rescan whenever me_clients list is modified. Signed-off-by:
-
Alexander Usyskin authored
The reserved host clients can be obsoleted now, a portion of the platforms is shipped without iAMT enabled, where the reservation is not relevant and for platforms with iAMT dynamic allocation is sufficient. Dropping reserved ids makes enumeration more flexible and generic Signed-off-by:
-
Alexander Usyskin authored
The FW can initiate client disconnection only because an error condition, hence it make sense to bump the debug message to the warning level to have an entery in the log. Signed-off-by:
-
Alexander Usyskin authored
Enable by default connection to fixed address clients from user-space for skylake and newer platform. Signed-off-by:
-
Alexander Usyskin authored
The read callback created from a flow control request for a fixed address client have NULL in the file pointer. Fill the file pointer using a data from a write callback. This allows us to drop workaround introduced in: commit eeabfcf5 ("mei: connection to fixed address clients from user-space") Signed-off-by:
-
Alexander Usyskin authored
A fixed address client in the FW doesn't have a notion of connection and can send message after the file associated with it was already closed. Silently discard such messages. Add inline helpers to detect whether a message is hbm or intended for a fixed address client Signed-off-by:
-
Alexander Usyskin authored
Clean write and write_waiting queues in disconnect. Requests in those queues are stale and processing will lead to fat warnings. In multi thread operations on disconnect and in FW disconnect case - write/read/event waiters should end wait and return error. Wake all waiters for disconnecting client to achieve that. Drop wake all and write queue clean on reset, as now we waking all waiters and cleaning write queues on disconnect. No need to do it twice. Signed-off-by:
-
Alexander Usyskin authored
In case of link reset all blocked writes should be interrupted. Note, that currently blocking write is used only through bus layer. Signed-off-by:
-
Alexander Usyskin authored
Signed-off-by:
-
Alexander Usyskin authored
The bus data transfer interface was missing the check if the device is in enabled state, this may lead to stack corruption during link reset. Cc: <stable@vger.kernel.org> #4.0 Signed-off-by:
-
Tomas Winkler authored
Call wake_up cl->ev_wait only in case there is no bus client registered to the event notification. Second, since we don't have exclusive waiter wake_up_interruptible_all is not used correctly here. Signed-off-by:
-
Tomas Winkler authored
In this particular case this more correct and safer to check if the RX event is set in the event mask rather than query waitqueue_active Since the check is already performed in the mei_cl_bus_rx_event function, it is just required to check for its return value. Second, since we don't have exclusive waiter wake_up_interruptible_all is not used correctly here. Signed-off-by:
-
Alexander Usyskin authored
In case of link reset all waiting readers should be interrupted. Signed-off-by:
-
Alexander Usyskin authored
Switch using cl->rx_wait wait queue also for amthif, there is nothing special about amthif in that matter in Rx flow. The cl->wait is reserved for hbm flows and asynchronous events Signed-off-by:
-
Tomas Winkler authored
Remove duplicated parameter validation from mei_amthif_write functions, The parameter check is already performed by the caller function mei_write Signed-off-by:
-
Tomas Winkler authored
Now when we have per client rd_completed list we can remove the amthif specific amthif_rd_complete_list. In addition in the function mei_amthif_read do not loop over the rd_completed list like the original code as the code path is unlocked. Reviewed-by:
-
Alexander Usyskin authored
A next amthif write can be executed only after the previous one has completed. Signed-off-by:
-
Tomas Winkler authored
The driver uses three names file, fp, and file_object for struct file type. To improve code clarity and adjust to my taste rename file_object to more common and shorter fp. Reviewed-by:
-
Tomas Winkler authored
The struct file file pointer is used as an opaque handle to for a connected client, for this part the pointer should be immutable and should be set to count. Reviewed-by:
-
Alexander Usyskin authored
Since the driver now uses a list for storing read packets instead of single variable a pending read is no longer blocking other connections. A pending read will be discarded up the file closure. Signed-off-by:
-
Alexander Usyskin authored
If empty message come from FW (buf_idx == 0) then the current code will still try to copy data from not filled buffer to the user-space, instead the code should behave the same as when end of a message has been reached, clean resources and return 0 Signed-off-by:
-
Alexander Usyskin authored
If registering of character device failed stop the device properly. Signed-off-by:
-
Tomas Winkler authored
There is a possible integer overflow following by a buffer overflow when accumulating messages coming from the FW to compose a full payload. Occurrence of wrap around has to be prevented for next message size calculation. For unsigned integer the addition overflow has occurred when the result is smaller than one of the arguments. To simplify the fix, the types of buf.size and buf_idx are set to the same width, namely size_t also to be aligned with the type of length parameter in file read/write ops. Signed-off-by:
-
Alexander Usyskin authored
HBM features list is ready while sending enumerate request and enumerating clients, output it to debugfs in these states too. Signed-off-by:
-
Alexander Usyskin authored
In case of many active host clients clients (41 and more) 1K buffer is not enough for full information print. Calculate buffer size according to real clients number. Signed-off-by:
-
Tomas Winkler authored
Use tracing events also for reading and writing pci configuration space <debugfs>/tracing/events/mei/mei_pci_reg_{read,write} Signed-off-by:
-
