diff options
Diffstat (limited to 'Documentation')
| -rw-r--r-- | Documentation/hid/uhid.txt | 179 | ||||
| -rw-r--r-- | Documentation/this_cpu_ops.txt | 213 | ||||
| -rw-r--r-- | Documentation/x86/tlb.txt | 2 |
3 files changed, 264 insertions, 130 deletions
diff --git a/Documentation/hid/uhid.txt b/Documentation/hid/uhid.txt index 54c8f9706a95..c8656dd029a9 100644 --- a/Documentation/hid/uhid.txt +++ b/Documentation/hid/uhid.txt @@ -1,28 +1,13 @@ UHID - User-space I/O driver support for HID subsystem ======================================================== -The HID subsystem needs two kinds of drivers. In this document we call them: +UHID allows user-space to implement HID transport drivers. Please see +hid-transport.txt for an introduction into HID transport drivers. This document +relies heavily on the definitions declared there. - 1. The "HID I/O Driver" is the driver that performs raw data I/O to the - low-level device. Internally, they register an hid_ll_driver structure with - the HID core. They perform device setup, read raw data from the device and - push it into the HID subsystem and they provide a callback so the HID - subsystem can send data to the device. - - 2. The "HID Device Driver" is the driver that parses HID reports and reacts on - them. There are generic drivers like "generic-usb" and "generic-bluetooth" - which adhere to the HID specification and provide the standardizes features. - But there may be special drivers and quirks for each non-standard device out - there. Internally, they use the hid_driver structure. - -Historically, the USB stack was the first subsystem to provide an HID I/O -Driver. However, other standards like Bluetooth have adopted the HID specs and -may provide HID I/O Drivers, too. The UHID driver allows to implement HID I/O -Drivers in user-space and feed the data into the kernel HID-subsystem. - -This allows user-space to operate on the same level as USB-HID, Bluetooth-HID -and similar. It does not provide a way to write HID Device Drivers, though. Use -hidraw for this purpose. +With UHID, a user-space transport driver can create kernel hid-devices for each +device connected to the user-space controlled bus. The UHID API defines the I/O +events provided from the kernel to user-space and vice versa. There is an example user-space application in ./samples/uhid/uhid-example.c @@ -42,8 +27,9 @@ by setting O_NONBLOCK. struct uhid_event { __u32 type; union { - struct uhid_create_req create; - struct uhid_data_req data; + struct uhid_create2_req create2; + struct uhid_output_req output; + struct uhid_input2_req input2; ... } u; }; @@ -54,8 +40,11 @@ multiple write()'s. A single event must always be sent as a whole. Furthermore, only a single event can be sent per read() or write(). Pending data is ignored. If you want to handle multiple events in a single syscall, then use vectored I/O with readv()/writev(). +The "type" field defines the payload. For each type, there is a +payload-structure available in the union "u" (except for empty payloads). This +payload contains management and/or device data. -The first thing you should do is sending an UHID_CREATE event. This will +The first thing you should do is sending an UHID_CREATE2 event. This will register the device. UHID will respond with an UHID_START event. You can now start sending data to and reading data from UHID. However, unless UHID sends the UHID_OPEN event, the internally attached HID Device Driver has no user attached. @@ -69,12 +58,20 @@ ref-counting for you. You may decide to ignore UHID_OPEN/UHID_CLOSE, though. I/O is allowed even though the device may have no users. -If you want to send data to the HID subsystem, you send an HID_INPUT event with -your raw data payload. If the kernel wants to send data to the device, you will -read an UHID_OUTPUT or UHID_OUTPUT_EV event. +If you want to send data on the interrupt channel to the HID subsystem, you send +an HID_INPUT2 event with your raw data payload. If the kernel wants to send data +on the interrupt channel to the device, you will read an UHID_OUTPUT event. +Data requests on the control channel are currently limited to GET_REPORT and +SET_REPORT (no other data reports on the control channel are defined so far). +Those requests are always synchronous. That means, the kernel sends +UHID_GET_REPORT and UHID_SET_REPORT events and requires you to forward them to +the device on the control channel. Once the device responds, you must forward +the response via UHID_GET_REPORT_REPLY and UHID_SET_REPORT_REPLY to the kernel. +The kernel blocks internal driver-execution during such round-trips (times out +after a hard-coded period). If your device disconnects, you should send an UHID_DESTROY event. This will -unregister the device. You can now send UHID_CREATE again to register a new +unregister the device. You can now send UHID_CREATE2 again to register a new device. If you close() the fd, the device is automatically unregistered and destroyed internally. @@ -82,73 +79,79 @@ internally. write() ------- write() allows you to modify the state of the device and feed input data into -the kernel. The following types are supported: UHID_CREATE, UHID_DESTROY and -UHID_INPUT. The kernel will parse the event immediately and if the event ID is +the kernel. The kernel will parse the event immediately and if the event ID is not supported, it will return -EOPNOTSUPP. If the payload is invalid, then -EINVAL is returned, otherwise, the amount of data that was read is returned and -the request was handled successfully. +the request was handled successfully. O_NONBLOCK does not affect write() as +writes are always handled immediately in a non-blocking fashion. Future requests +might make use of O_NONBLOCK, though. - UHID_CREATE: + UHID_CREATE2: This creates the internal HID device. No I/O is possible until you send this - event to the kernel. The payload is of type struct uhid_create_req and + event to the kernel. The payload is of type struct uhid_create2_req and contains information about your device. You can start I/O now. - UHID_CREATE2: - Same as UHID_CREATE, but the HID report descriptor data (rd_data) is an array - inside struct uhid_create2_req, instead of a pointer to a separate array. - Enables use from languages that don't support pointers, e.g. Python. - UHID_DESTROY: This destroys the internal HID device. No further I/O will be accepted. There may still be pending messages that you can receive with read() but no further UHID_INPUT events can be sent to the kernel. - You can create a new device by sending UHID_CREATE again. There is no need to + You can create a new device by sending UHID_CREATE2 again. There is no need to reopen the character device. - UHID_INPUT: - You must send UHID_CREATE before sending input to the kernel! This event - contains a data-payload. This is the raw data that you read from your device. - The kernel will parse the HID reports and react on it. - UHID_INPUT2: - Same as UHID_INPUT, but the data array is the last field of uhid_input2_req. - Enables userspace to write only the required bytes to kernel (ev.type + - ev.u.input2.size + the part of the data array that matters), instead of - the entire struct uhid_input2_req. - - UHID_FEATURE_ANSWER: - If you receive a UHID_FEATURE request you must answer with this request. You - must copy the "id" field from the request into the answer. Set the "err" field - to 0 if no error occurred or to EIO if an I/O error occurred. + You must send UHID_CREATE2 before sending input to the kernel! This event + contains a data-payload. This is the raw data that you read from your device + on the interrupt channel. The kernel will parse the HID reports. + + UHID_GET_REPORT_REPLY: + If you receive a UHID_GET_REPORT request you must answer with this request. + You must copy the "id" field from the request into the answer. Set the "err" + field to 0 if no error occurred or to EIO if an I/O error occurred. If "err" is 0 then you should fill the buffer of the answer with the results - of the feature request and set "size" correspondingly. + of the GET_REPORT request and set "size" correspondingly. + + UHID_SET_REPORT_REPLY: + This is the SET_REPORT equivalent of UHID_GET_REPORT_REPLY. Unlike GET_REPORT, + SET_REPORT never returns a data buffer, therefore, it's sufficient to set the + "id" and "err" fields correctly. read() ------ -read() will return a queued output report. These output reports can be of type -UHID_START, UHID_STOP, UHID_OPEN, UHID_CLOSE, UHID_OUTPUT or UHID_OUTPUT_EV. No -reaction is required to any of them but you should handle them according to your -needs. Only UHID_OUTPUT and UHID_OUTPUT_EV have payloads. +read() will return a queued output report. No reaction is required to any of +them but you should handle them according to your needs. UHID_START: This is sent when the HID device is started. Consider this as an answer to - UHID_CREATE. This is always the first event that is sent. + UHID_CREATE2. This is always the first event that is sent. Note that this + event might not be available immediately after write(UHID_CREATE2) returns. + Device drivers might required delayed setups. + This event contains a payload of type uhid_start_req. The "dev_flags" field + describes special behaviors of a device. The following flags are defined: + UHID_DEV_NUMBERED_FEATURE_REPORTS: + UHID_DEV_NUMBERED_OUTPUT_REPORTS: + UHID_DEV_NUMBERED_INPUT_REPORTS: + Each of these flags defines whether a given report-type uses numbered + reports. If numbered reports are used for a type, all messages from + the kernel already have the report-number as prefix. Otherwise, no + prefix is added by the kernel. + For messages sent by user-space to the kernel, you must adjust the + prefixes according to these flags. UHID_STOP: This is sent when the HID device is stopped. Consider this as an answer to UHID_DESTROY. - If the kernel HID device driver closes the device manually (that is, you - didn't send UHID_DESTROY) then you should consider this device closed and send - an UHID_DESTROY event. You may want to reregister your device, though. This is - always the last message that is sent to you unless you reopen the device with - UHID_CREATE. + If you didn't destroy your device via UHID_DESTROY, but the kernel sends an + UHID_STOP event, this should usually be ignored. It means that the kernel + reloaded/changed the device driver loaded on your HID device (or some other + maintenance actions happened). + You can usually ignored any UHID_STOP events safely. UHID_OPEN: This is sent when the HID device is opened. That is, the data that the HID device provides is read by some other process. You may ignore this event but it is useful for power-management. As long as you haven't received this event there is actually no other process that reads your data so there is no need to - send UHID_INPUT events to the kernel. + send UHID_INPUT2 events to the kernel. UHID_CLOSE: This is sent when there are no more processes which read the HID data. It is @@ -156,27 +159,29 @@ needs. Only UHID_OUTPUT and UHID_OUTPUT_EV have payloads. UHID_OUTPUT: This is sent if the HID device driver wants to send raw data to the I/O - device. You should read the payload and forward it to the device. The payload - is of type "struct uhid_data_req". + device on the interrupt channel. You should read the payload and forward it to + the device. The payload is of type "struct uhid_data_req". This may be received even though you haven't received UHID_OPEN, yet. - UHID_OUTPUT_EV (obsolete): - Same as UHID_OUTPUT but this contains a "struct input_event" as payload. This - is called for force-feedback, LED or similar events which are received through - an input device by the HID subsystem. You should convert this into raw reports - and send them to your device similar to events of type UHID_OUTPUT. - This is no longer sent by newer kernels. Instead, HID core converts it into a - raw output report and sends it via UHID_OUTPUT. - - UHID_FEATURE: - This event is sent if the kernel driver wants to perform a feature request as - described in the HID specs. The report-type and report-number are available in - the payload. - The kernel serializes feature requests so there will never be two in parallel. - However, if you fail to respond with a UHID_FEATURE_ANSWER in a time-span of 5 - seconds, then the requests will be dropped and a new one might be sent. - Therefore, the payload also contains an "id" field that identifies every - request. - -Document by: - David Herrmann <dh.herrmann@googlemail.com> + UHID_GET_REPORT: + This event is sent if the kernel driver wants to perform a GET_REPORT request + on the control channeld as described in the HID specs. The report-type and + report-number are available in the payload. + The kernel serializes GET_REPORT requests so there will never be two in + parallel. However, if you fail to respond with a UHID_GET_REPORT_REPLY, the + request might silently time out. + Once you read a GET_REPORT request, you shall forward it to the hid device and + remember the "id" field in the payload. Once your hid device responds to the + GET_REPORT (or if it fails), you must send a UHID_GET_REPORT_REPLY to the + kernel with the exact same "id" as in the request. If the request already + timed out, the kernel will ignore the response silently. The "id" field is + never re-used, so conflicts cannot happen. + + UHID_SET_REPORT: + This is the SET_REPORT equivalent of UHID_GET_REPORT. On receipt, you shall + send a SET_REPORT request to your hid device. Once it replies, you must tell + the kernel about it via UHID_SET_REPORT_REPLY. + The same restrictions as for UHID_GET_REPORT apply. + +---------------------------------------------------- +Written 2012, David Herrmann <dh.herrmann@gmail.com> diff --git a/Documentation/this_cpu_ops.txt b/Documentation/this_cpu_ops.txt index 1a4ce7e3e05f..0ec995712176 100644 --- a/Documentation/this_cpu_ops.txt +++ b/Documentation/this_cpu_ops.txt @@ -2,26 +2,26 @@ this_cpu operations ------------------- this_cpu operations are a way of optimizing access to per cpu -variables associated with the *currently* executing processor through -the use of segment registers (or a dedicated register where the cpu -permanently stored the beginning of the per cpu area for a specific -processor). +variables associated with the *currently* executing processor. This is +done through the use of segment registers (or a dedicated register where +the cpu permanently stored the beginning of the per cpu area for a +specific processor). -The this_cpu operations add a per cpu variable offset to the processor -specific percpu base and encode that operation in the instruction +this_cpu operations add a per cpu variable offset to the processor +specific per cpu base and encode that operation in the instruction operating on the per cpu variable. -This means there are no atomicity issues between the calculation of +This means that there are no atomicity issues between the calculation of the offset and the operation on the data. Therefore it is not -necessary to disable preempt or interrupts to ensure that the +necessary to disable preemption or interrupts to ensure that the processor is not changed between the calculation of the address and the operation on the data. Read-modify-write operations are of particular interest. Frequently processors have special lower latency instructions that can operate -without the typical synchronization overhead but still provide some -sort of relaxed atomicity guarantee. The x86 for example can execute -RMV (Read Modify Write) instructions like inc/dec/cmpxchg without the +without the typical synchronization overhead, but still provide some +sort of relaxed atomicity guarantees. The x86, for example, can execute +RMW (Read Modify Write) instructions like inc/dec/cmpxchg without the lock prefix and the associated latency penalty. Access to the variable without the lock prefix is not synchronized but @@ -30,6 +30,38 @@ data specific to the currently executing processor. Only the current processor should be accessing that variable and therefore there are no concurrency issues with other processors in the system. +Please note that accesses by remote processors to a per cpu area are +exceptional situations and may impact performance and/or correctness +(remote write operations) of local RMW operations via this_cpu_*. + +The main use of the this_cpu operations has been to optimize counter +operations. + +The following this_cpu() operations with implied preemption protection +are defined. These operations can be used without worrying about +preemption and interrupts. + + this_cpu_add() + this_cpu_read(pcp) + this_cpu_write(pcp, val) + this_cpu_add(pcp, val) + this_cpu_and(pcp, val) + this_cpu_or(pcp, val) + this_cpu_add_return(pcp, val) + this_cpu_xchg(pcp, nval) + this_cpu_cmpxchg(pcp, oval, nval) + this_cpu_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2) + this_cpu_sub(pcp, val) + this_cpu_inc(pcp) + this_cpu_dec(pcp) + this_cpu_sub_return(pcp, val) + this_cpu_inc_return(pcp) + this_cpu_dec_return(pcp) + + +Inner working of this_cpu operations +------------------------------------ + On x86 the fs: or the gs: segment registers contain the base of the per cpu area. It is then possible to simply use the segment override to relocate a per cpu relative address to the proper per cpu area for @@ -48,22 +80,21 @@ results in a single instruction mov ax, gs:[x] instead of a sequence of calculation of the address and then a fetch -from that address which occurs with the percpu operations. Before +from that address which occurs with the per cpu operations. Before this_cpu_ops such sequence also required preempt disable/enable to prevent the kernel from moving the thread to a different processor while the calculation is performed. -The main use of the this_cpu operations has been to optimize counter -operations. +Consider the following this_cpu operation: this_cpu_inc(x) -results in the following single instruction (no lock prefix!) +The above results in the following single instruction (no lock prefix!) inc gs:[x] instead of the following operations required if there is no segment -register. +register: int *y; int cpu; @@ -73,10 +104,10 @@ register. (*y)++; put_cpu(); -Note that these operations can only be used on percpu data that is +Note that these operations can only be used on per cpu data that is reserved for a specific processor. Without disabling preemption in the surrounding code this_cpu_inc() will only guarantee that one of the -percpu counters is correctly incremented. However, there is no +per cpu counters is correctly incremented. However, there is no guarantee that the OS will not move the process directly before or after the this_cpu instruction is executed. In general this means that the value of the individual counters for each processor are @@ -86,9 +117,9 @@ that is of interest. Per cpu variables are used for performance reasons. Bouncing cache lines can be avoided if multiple processors concurrently go through the same code paths. Since each processor has its own per cpu -variables no concurrent cacheline updates take place. The price that +variables no concurrent cache line updates take place. The price that has to be paid for this optimization is the need to add up the per cpu -counters when the value of the counter is needed. +counters when the value of a counter is needed. Special operations: @@ -100,33 +131,39 @@ Takes the offset of a per cpu variable (&x !) and returns the address of the per cpu variable that belongs to the currently executing processor. this_cpu_ptr avoids multiple steps that the common get_cpu/put_cpu sequence requires. No processor number is -available. Instead the offset of the local per cpu area is simply -added to the percpu offset. +available. Instead, the offset of the local per cpu area is simply +added to the per cpu offset. +Note that this operation is usually used in a code segment when +preemption has been disabled. The pointer is then used to +access local per cpu data in a critical section. When preemption +is re-enabled this pointer is usually no longer useful since it may +no longer point to per cpu data of the current processor. Per cpu variables and offsets ----------------------------- -Per cpu variables have *offsets* to the beginning of the percpu +Per cpu variables have *offsets* to the beginning of the per cpu area. They do not have addresses although they look like that in the code. Offsets cannot be directly dereferenced. The offset must be -added to a base pointer of a percpu area of a processor in order to +added to a base pointer of a per cpu area of a processor in order to form a valid address. Therefore the use of x or &x outside of the context of per cpu operations is invalid and will generally be treated like a NULL pointer dereference. -In the context of per cpu operations + DEFINE_PER_CPU(int, x); - x is a per cpu variable. Most this_cpu operations take a cpu - variable. +In the context of per cpu operations the above implies that x is a per +cpu variable. Most this_cpu operations take a cpu variable. - &x is the *offset* a per cpu variable. this_cpu_ptr() takes - the offset of a per cpu variable which makes this look a bit - strange. + int __percpu *p = &x; +&x and hence p is the *offset* of a per cpu variable. this_cpu_ptr() +takes the offset of a per cpu variable which makes this look a bit +strange. Operations on a field of a per cpu structure @@ -152,7 +189,7 @@ If we have an offset to struct s: struct s __percpu *ps = &p; - z = this_cpu_dec(ps->m); + this_cpu_dec(ps->m); z = this_cpu_inc_return(ps->n); @@ -172,29 +209,52 @@ if we do not make use of this_cpu ops later to manipulate fields: Variants of this_cpu ops ------------------------- -this_cpu ops are interrupt safe. Some architecture do not support +this_cpu ops are interrupt safe. Some architectures do not support these per cpu local operations. In that case the operation must be replaced by code that disables interrupts, then does the operations -that are guaranteed to be atomic and then reenable interrupts. Doing +that are guaranteed to be atomic and then re-enable interrupts. Doing so is expensive. If there are other reasons why the scheduler cannot change the processor we are executing on then there is no reason to -disable interrupts. For that purpose the __this_cpu operations are -provided. For example. - - __this_cpu_inc(x); - -Will increment x and will not fallback to code that disables +disable interrupts. For that purpose the following __this_cpu operations +are provided. + +These operations have no guarantee against concurrent interrupts or +preemption. If a per cpu variable is not used in an interrupt context +and the scheduler cannot preempt, then they are safe. If any interrupts +still occur while an operation is in progress and if the interrupt too +modifies the variable, then RMW actions can not be guaranteed to be +safe. + + __this_cpu_add() + __this_cpu_read(pcp) + __this_cpu_write(pcp, val) + __this_cpu_add(pcp, val) + __this_cpu_and(pcp, val) + __this_cpu_or(pcp, val) + __this_cpu_add_return(pcp, val) + __this_cpu_xchg(pcp, nval) + __this_cpu_cmpxchg(pcp, oval, nval) + __this_cpu_cmpxchg_double(pcp1, pcp2, oval1, oval2, nval1, nval2) + __this_cpu_sub(pcp, val) + __this_cpu_inc(pcp) + __this_cpu_dec(pcp) + __this_cpu_sub_return(pcp, val) + __this_cpu_inc_return(pcp) + __this_cpu_dec_return(pcp) + + +Will increment x and will not fall-back to code that disables interrupts on platforms that cannot accomplish atomicity through address relocation and a Read-Modify-Write operation in the same instruction. - &this_cpu_ptr(pp)->n vs this_cpu_ptr(&pp->n) -------------------------------------------- The first operation takes the offset and forms an address and then -adds the offset of the n field. +adds the offset of the n field. This may result in two add +instructions emitted by the compiler. The second one first adds the two offsets and then does the relocation. IMHO the second form looks cleaner and has an easier time @@ -202,4 +262,73 @@ with (). The second form also is consistent with the way this_cpu_read() and friends are used. -Christoph Lameter, April 3rd, 2013 +Remote access to per cpu data +------------------------------ + +Per cpu data structures are designed to be used by one cpu exclusively. +If you use the variables as intended, this_cpu_ops() are guaranteed to +be "atomic" as no other CPU has access to these data structures. + +There are special cases where you might need to access per cpu data +structures remotely. It is usually safe to do a remote read access +and that is frequently done to summarize counters. Remote write access +something which could be problematic because this_cpu ops do not +have lock semantics. A remote write may interfere with a this_cpu +RMW operation. + +Remote write accesses to percpu data structures are highly discouraged +unless absolutely necessary. Please consider using an IPI to wake up +the remote CPU and perform the update to its per cpu area. + +To access per-cpu data structure remotely, typically the per_cpu_ptr() +function is used: + + + DEFINE_PER_CPU(struct data, datap); + + struct data *p = per_cpu_ptr(&datap, cpu); + +This makes it explicit that we are getting ready to access a percpu +area remotely. + +You can also do the following to convert the datap offset to an address + + struct data *p = this_cpu_ptr(&datap); + +but, passing of pointers calculated via this_cpu_ptr to other cpus is +unusual and should be avoided. + +Remote access are typically only for reading the status of another cpus +per cpu data. Write accesses can cause unique problems due to the +relaxed synchronization requirements for this_cpu operations. + +One example that illustrates some concerns with write operations is +the following scenario that occurs because two per cpu variables +share a cache-line but the relaxed synchronization is applied to +only one process updating the cache-line. + +Consider the following example + + + struct test { + atomic_t a; + int b; + }; + + DEFINE_PER_CPU(struct test, onecacheline); + +There is some concern about what would happen if the field 'a' is updated +remotely from one processor and the local processor would use this_cpu ops +to update field b. Care should be taken that such simultaneous accesses to +data within the same cache line are avoided. Also costly synchronization +may be necessary. IPIs are generally recommended in such scenarios instead +of a remote write to the per cpu area of another processor. + +Even in cases where the remote writes are rare, please bear in +mind that a remote write will evict the cache line from the processor +that most likely will access it. If the processor wakes up and finds a +missing local cache line of a per cpu area, its performance and hence +the wake up times will be affected. + +Christoph Lameter, August 4th, 2014 +Pranith Kumar, Aug 2nd, 2014 diff --git a/Documentation/x86/tlb.txt b/Documentation/x86/tlb.txt index 2b3a82e69151..39d172326703 100644 --- a/Documentation/x86/tlb.txt +++ b/Documentation/x86/tlb.txt @@ -35,7 +35,7 @@ invlpg instruction (or instructions _near_ it) show up high in profiles. If you believe that individual invalidations being called too often, you can lower the tunable: - /sys/debug/kernel/x86/tlb_single_page_flush_ceiling + /sys/kernel/debug/x86/tlb_single_page_flush_ceiling This will cause us to do the global flush for more cases. Lowering it to 0 will disable the use of the individual flushes. |