list.h includes a number of FreeBSD headers as a workaround for the
LIST_HEAD name collision. To reduce pollution, avoid including list.h
in commonly used headers when it is not explicitly needed.
Reviewed by: hselasky
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D11249
In particular:
- Don't evaluate event conditions with a sleepqueue lock held, since such
code may attempt to acquire arbitrary locks.
- Fix the return value for wait_event_interruptible() in the case that the
wait is interrupted by a signal.
- Implement wait_on_bit_timeout() and wait_on_atomic_t().
- Implement some functions used to test for pending signals.
- Implement a number of wait_event_*() variants and unify the existing
implementations.
- Unify the mechanism used by wait_event_*() and schedule() to put the
calling thread to sleep.
This is required to support updated DRM drivers. Thanks to hselasky for
finding and fixing a number of bugs in the original revision.
Reviewed by: hselasky
MFC after: 2 weeks
Differential Revision: https://reviews.freebsd.org/D10986
ARM and MIPS fail universe builds.
ARM and MIPS are missing the following:
* VM_MEMATTR_WRITE_THROUGH
* VM_MEMATTR_WRITE_COMBINING
Pointy-hat to: jhibbits
arm, mips, and powerpc all implement pmap_mapdev_attr() and pmap_unmapdev(),
so add those archs to the checks. powerpc also includes the atomic_swap_*()
functions, so add that to the supported list as well. Not tested except by
compiling powerpc.
Reviewed by: markj
CPU_FOREACH() is not available until SI_SUB_CPU at SI_ORDER_ANY
when the LinuxKPI is loaded as part of the kernel.
MFC after: 1 week
Sponsored by: Mellanox Technologies
kqueue() does not set non-blocking I/O mode for event driven read of
file descriptors. This means the LinuxKPI internal kqueue read and
write event flags must be updated before the next read and/or write
system call. Else the read and/or write system call may block. This
can happen when there is no more data to read following a previous
read event. Then the application also gets blocked from processing
other events. This situation can also be solved by the applications
setting and using non-blocking I/O mode.
MFC after: 1 week
Sponsored by: Mellanox Technologies
character devices. In Linux the FIONBIO IOCTL is handled by the kernel
and not the drivers. Also need return success for the FIOASYNC ioctl
due to existing logic in kern_fcntl() even though it is not supported
currently.
MFC after: 1 week
Sponsored by: Mellanox Technologies
polling contexts in the LinuxKPI.
After the kqueue() support was added to the LinuxKPI in r319409 the
Linux poll file operation will be used outside the system file polling
callback function, which can cause a NULL-pointer panic inside
selrecord() because curthread->td_sel is set to NULL. This patch moves
the selrecord() call away from poll_wait() and to the system file poll
callback function in the LinuxKPI, which essentially wraps the Linux
one. This is similar to what the cuse(3) module is currently doing.
Refer to sys/fs/cuse/*.[ch] for more details.
MFC after: 1 week
Sponsored by: Mellanox Technologies
ioctl(), read() and write() system call handlers. This error code is
internal to the kernel and should not be seen by user-space programs
according to Linux.
Submitted by: Yanko Yankulov <yanko.yankulov@gmail.com>
MFC after: 1 week
Sponsored by: Mellanox Technologies
devices. The implementation allows read and write filters to be
created and piggybacks on the poll() file operation to determine when
a filter should trigger. The piggyback mechanism is simply to check
for the EWOULDBLOCK or EAGAIN return code from read(), write() or
ioctl() system calls and then update the kqueue() polling state bits.
The implementation is similar to the one found in the cuse(3) module.
Refer to sys/fs/cuse/*.[ch] for more details.
MFC after: 1 week
Sponsored by: Mellanox Technologies
printk_ratelimited() in the LinuxKPI.
While at it fix the inclusion guard of printk.h to be similar to the
rest of the LinuxKPI header files.
MFC after: 1 week
Sponsored by: Mellanox Technologies
task structure to avoid deadlock when tearing down the VM object
during a process exit.
Found by: markj @
MFC after: 1 week
Sponsored by: Mellanox Technologies
it shouldn't be called.
Background:
The Linux VM open operation is called when a new VMA is
created on top of the current VMA. This is done through either mremap
flow or split_vma, usually due to mlock, madvise, munmap and so
on. This is currently not supported by the LinuxKPI.
MFC after: 1 week
Sponsored by: Mellanox Technologies
- Allow "struct linux_file" to be refcounted when its "_file" member
is NULL by using its "f_count" field. The reference counts are
transferred to the file structure when the file descriptor is
installed.
- Add missing vdrop() calls for error cases during open().
- Set the "_file" member of "struct linux_file" during open. This
allows use of refcounting through get_file() and fput() with LinuxKPI
character devices.
MFC after: 1 week
Sponsored by: Mellanox Technologies
kern_yield(0) effectively causes the calling thread to be rescheduled
immediately since it resets the thread's priority to the highest possible
value. This can cause livelocks when the pattern
"while (!trylock()) kern_yield(0);" is used since the thread holding the
lock may linger on the runqueue for the CPU on which the looping thread is
running.
MFC after: 1 week
CPU_FOREACH() is not available until SI_SUB_CPU at SI_ORDER_ANY
when the LinuxKPI is loaded as part of the kernel.
MFC after: 1 week
Sponsored by: Mellanox Technologies
Background:
The same VM object might be shared by multiple processes and the
mm_struct is usually freed when a process exits.
Grab a reference on the mm_struct while the vmap is in the
linux_vma_head list in case the first process which inserted a VM
object has exited.
Tested by: kwm @
MFC after: 1 week
Sponsored by: Mellanox Technologies
linux_page_address() function in the LinuxKPI. This solves an issue
where the return value from linux_page_address() is passed to
kmem_free().
MFC after: 1 week
Sponsored by: Mellanox Technologies
kit, CK, in the LinuxKPI.
When threads are pinned to a CPU core or when there is only one CPU,
it can happen that a higher priority thread can call the CK
synchronize function while a lower priority thread holds the read
lock. Because the CK's synchronize is a simple wait loop this can lead
to a deadlock situation. To solve this problem use the recently
introduced CK's wait callback function.
When detecting a CK blocking condition figure out the lowest priority
among the blockers and update the calling thread's priority and
yield. If another CPU core is holding the read lock, pin the thread to
the blocked CPU core and update the priority. The calling threads
priority and CPU bindings are restored before return.
If a thread holding a CK read lock is detected to be sleeping, pause()
will be used instead of yield().
MFC after: 1 week
Sponsored by: Mellanox Technologies
CPUs when allocating a LinuxKPI workqueue. This also ensures that the
created taskqueue always have a non-zero number of worker threads.
MFC after: 1 week
Sponsored by: Mellanox Technologies
- Move all bitmap related functions from bitops.h to bitmap.h, similar
to what Linux does.
- Apply some minor code cleanup and simplifications to optimize the
generated code when using static inline functions.
- Implement the following list of bitmap functions which are needed by
drm-next and ibcore:
- bitmap_find_next_zero_area_off()
- bitmap_find_next_zero_area()
- bitmap_or()
- bitmap_and()
- bitmap_xor()
- Add missing include directives to the qlnxe driver
(davidcs@ has been notified)
MFC after: 1 week
Sponsored by: Mellanox Technologies
In FreeBSD thread IDs and procedure IDs have distinct number
spaces. When asking for the group leader task ID in the LinuxKPI,
return the procedure ID and let this resolve to the first task in the
procedure having a valid LinuxKPI task structure pointer.
MFC after: 1 week
Sponsored by: Mellanox Technologies
like open, close and fault using the character device pager.
Some notes about the implementation:
1) Linux drivers set the vm_ops and vm_private_data fields during a
mmap() call to indicate that the driver wants to use the LinuxKPI VM
operations. Else these operations are not used.
2) The vm_private_data pointer is associated with a VM area structure
and inserted into an internal LinuxKPI list. If the vm_private_data
pointer already exists, the existing VM area structure is used instead
of the allocated one which gets freed.
3) The LinuxKPI's vm_private_data pointer is used as the callback
handle for the FreeBSD VM object. The VM subsystem in FreeBSD has a
similar list to identify equal handles and will only call the
character device pager's close function once.
4) All LinuxKPI VM operations are serialized through the mmap_sem
sempaphore, which is per procedure, which prevents simultaneous access
to the shared VM area structure when receiving page faults.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
tasks in the belonging procedure already have a valid mm_struct and
reference that instead.
The mm_struct in the LinuxKPI should be shared among all tasks
belonging to the same procedure. This has to do with with the mmap_sem
semaphore which should serialize all VM operations inside a given
procedure. Linux based drivers depend on this behaviour.
MFC after: 1 week
Sponsored by: Mellanox Technologies
Make sure the character device poll callback function does not return
an error code, but a POLLXXX value, in case of failure.
MFC after: 1 week
Sponsored by: Mellanox Technologies
the LinuxKPI for accessing user-space memory in the kernel.
Add functions to hold and wire physical page(s) based on a given range
of user-space virtual addresses.
Add functions to get and put a reference on, wire, hold, mark
accessed, copy and dirty a physical page.
Add new VM related structures and defines as a preparation step for
advancing the memory map capabilities of the LinuxKPI.
Add function to figure out if a virtual address was allocated using
malloc().
Add function to convert a virtual kernel address into its physical
page pointer.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
functions in the LinuxKPI. Add a usage atomic to the task_struct
structure to facilitate refcounting the task structure when returned
from get_pid_task(). The get_task_struct() and put_task_struct()
function is used to manage atomic refcounting. After this change the
task_struct should only be freed through put_task_struct().
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
some associated helper functions in the LinuxKPI. Let the existing
linux_alloc_current() function allocate and initialize the new
structure and let linux_free_current() drop the refcount on the memory
mapping structure. When the mm_struct's refcount reaches zero, the
structure is freed.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
pairwise to support the FreeBSD way of pushing and popping the page
fault flags. Ensure this by requiring every occurrence of pagefault
disable function call to have a corresponding pagefault enable call.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
Support is implemented by mapping Linux's "struct net" into FreeBSD's
"struct vnet". Currently only vnet0 is supported by ibcore.
MFC after: 1 week
Sponsored by: Mellanox Technologies
When locking a mutex and deadlock is detected the first mutex lock
call that sees the deadlock will return -EDEADLK .
MFC after: 1 week
Sponsored by: Mellanox Technologies
Add support for using mutexes during KDB and shutdown. This is also
required for doing mode-switching during panic for drm-next.
Add new mutex functions mutex_init_witness() and mutex_destroy()
allowing LinuxKPI mutexes to be tracked by witness.
Declare mutex_is_locked() and mutex_is_owned() like inline functions
to get cleaner warnings. These functions are used inside WARN_ON()
statements which might look a bit odd if these functions get fully
expanded.
Give mutexes better debug names through the mutex_name() macro when
WITNESS_ALL is defined. The mutex_name() macro can prefix parts of the
filename and line number before the mutex name.
MFC after: 1 week
Sponsored by: Mellanox Technologies
kthread_add() will assert it is called too soon. This fixes a startup
issue when COMPAT_LINUXKPI is in enabled the kernel configuration
file.
Reported by: Michael Butler <imb@protected-networks.net>
MFC after: 1 week
Sponsored by: Mellanox Technologies
Put large functions into linux_slab.c instead of declaring them static
inline.
Add support for more memory allocation wrappers like kmalloc_array()
and __vmalloc().
Make sure either the M_WAITOK or the M_NOWAIT flag is set and mask
away unused memory allocation flags before calling FreeBSD's malloc()
routine.
Move kmalloc_node() definition to slab.h where it belongs.
Implement support for the SLAB_DESTROY_BY_RCU feature when creating a
kmem_cache which basically means kmem_cache memory is freed using
call_rcu().
MFC after: 1 week
Sponsored by: Mellanox Technologies
LinuxKPI. When the type of the argument is constant the temporary
variable cannot be assigned after the barrier. Instead assign the
temporary variable by initialization.
MFC after: 1 week
Sponsored by: Mellanox Technologies
This avoids creating own per-CPU threads and also ensures the tasklet
execution happens on the same CPU core invoking the tasklet.
MFC after: 1 week
Sponsored by: Mellanox Technologies
This change makes the workqueue implementation behave more like in
Linux, both functionality wise and structure wise.
All workqueue code has been moved to linux_work.c
Add an atomic based statemachine to the work_struct to ensure proper
operation. Prior to this change struct_work was directly mapped to a
FreeBSD task. When a taskqueue has multiple threads the same task may
end up being executed on more than one worker thread simultaneously.
This might cause problems with code coming from Linux, which expects
serial behaviour, similar to Linux tasklets.
Move all global workqueue function names into the linux_xxx domain to
avoid symbol name clashes in the future.
Implement a few more workqueue related functions and macros.
Create two multithreaded taskqueues for the LinuxKPI during module
load, one for time-consuming callbacks and one for non-time consuming
callbacks.
MFC after: 1 week
Sponsored by: Mellanox Technologies
WITNESS_ALL is defined. The lock name is based on the filename and
line number where the initialisation happens.
MFC after: 1 week
Sponsored by: Mellanox Technologies
- Optimise the RCU implementation to not allocate and free
ck_epoch_records during runtime. Instead allocate two sets of
ck_epoch_records per CPU for general purpose use. The first set is
only used for reader locks and the second set is only used for
synchronization and barriers and is protected with a regular mutex to
prevent simultaneous issues.
- Move the task structure away from the rcu_head structure and into
the per-CPU structures. This allows the size of the rcu_head structure
to be reduced down to the size of two pointers.
- Fix a bug where the linux_rcu_barrier() function only waited for one
per-CPU epoch record to be completed instead of all.
- Use a critical section or a mutex to protect ck_epoch_begin() and
ck_epoch_end() depending on RCU or SRCU type. All the ck_epoch_xxx()
functions, except ck_epoch_register(), ck_epoch_unregister() and
ck_epoch_recycle() are not re-entrant and needs a critical section or
a mutex to operate in the LinuxKPI, after inspecting the CK
implementation of the above mentioned functions. The simultaneous
issues arise from per-CPU epoch records being shared between multiple
threads depending on the amount of taskswitching and how many threads
are involved with the RCU and SRCU operations.
- Properly free all epoch records by using safe list traversal at
LinuxKPI module unload. It turns out the ck_epoch_recycle() always
have the records on an internal list and use a flag in the epoch
record to track allocated and free entries. This would lead to use
after free during module unload.
- Remove redundant synchronize_rcu() call from the
linux_compat_uninit() function. Let the linux_rcu_runtime_uninit()
function do the final rcu_barrier() instead.
MFC after: 1 week
Sponsored by: Mellanox Technologies
The clang compiler will optimise these functions down to three AMD64
instructions if the bit argument is a constant during compilation.
MFC after: 1 week
Sponsored by: Mellanox Technologies
When allocating unmapped pages, take advantage of the direct map on
AMD64 to get the virtual address corresponding to a page. Else all
pages allocated must be mapped because sometimes the virtual address
of a page is requested.
Move all page allocation and deallocation code into an own C-file.
Add support for GFP_DMA32, GFP_KERNEL, GFP_ATOMIC and __GFP_ZERO
allocation flags.
Make a clear separation between mapped and unmapped allocations.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
The i915kms driver in Linux 4.9 reimplement parts of the scatter list
functions with regards to performance. In other words there is not so
much room for changing structure layouts and functionality if the
i915kms should be built AS-IS. This patch aligns the scatter list
support to what is expected by the i915kms driver. Remove some
comments not needed while at it.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
1) Add better spinlock debug names when WITNESS_ALL is defined.
2) Make sure that the calling thread gets bound to the current CPU
while a spinlock is locked. Some Linux kernel code depends on that the
CPU ID doesn't change while a spinlock is locked.
3) Add support for using LinuxKPI spinlocks during a panic().
MFC after: 1 week
Sponsored by: Mellanox Technologies
Tasklets are implemented using a taskqueue and a small statemachine on
top. The additional statemachine is required to ensure all LinuxKPI
tasklets get serialized. FreeBSD taskqueues do not guarantee
serialisation of its tasks, except when there is only one worker
thread configured.
MFC after: 1 week
Sponsored by: Mellanox Technologies
A set of helper functions have been added to manage the life of the
LinuxKPI task struct. When an external system call or task is invoked,
a check is made to create the task struct by demand. A thread
destructor callback is registered to free the task struct when a
thread exits to avoid memory leaks.
This change lays the ground for emulating the Linux kernel more
closely which is a dependency by the code using the LinuxKPI APIs.
Add new dedicated td_lkpi_task field has been added to struct thread
instead of abusing td_retval[1].
Fix some header file inclusions to make LINT kernel build properly
after this change.
Bump the __FreeBSD_version to force a rebuild of all kernel modules.
MFC after: 1 week
Sponsored by: Mellanox Technologies
The intended use is to annotate frequently used globals which either rarely
change (and thus can be grouped in the same cacheline) or are an atomic counter
(which means it may benefit from being the only variable in the cacheline).
Linker script support is provided only for amd64. Architectures without it risk
having other variables put in, i.e. as if they were not annotated. This is
harmless from correctness point of view.
Reviewed by: bde (previous version)
MFC after: 1 month
the ones obtained through devclass_get_device(). Some minor code
cleanups while at it.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
Linux list_for_each_entry() does not neccessarily end with the iterator
NULL (it may be an offset from NULL if the list member is not the first
element of the member struct).
Reported by: Coverity
CID: 1366940
Reviewed by: hselasky@
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D8780
My plan is to change this function's prototype at some point in the
future to add a new label argument, which can be used to export all of
sysctl as metrics that can be scraped by Prometheus. Switch over this
caller to use the macro wrapper counterpart.
conflict with the opensolaris kernel module.
This patch solves a problem where the kernel linker will incorrectly
resolve opensolaris kmem_xxx() functions as linuxkpi ones, which leads
to a panic when these functions are used.
Submitted by: gallatin @
Sponsored by: Mellanox Technologies
MFC after: 1 week
FreeBSD supports lazy allocation of PCI BAR, that is, when a device
driver's attach method is invoked, even if the device's PCI BAR
address wasn't initialized, the invocation of bus_alloc_resource_any()
(the call chain: pci_alloc_resource() -> pci_alloc_multi_resource() ->
pci_reserve_map() -> pci_write_bar()) would allocate a proper address
for the PCI BAR and write this 'lazy allocated' address into the PCI
BAR.
This model works fine for native FreeBSD device drivers, but _not_ for
device drivers shared with Linux (e.g. dev/mlx5/mlx5_core/mlx5_main.c
and ofed/drivers/net/mlx4/main.c. Both of them use
pci_request_regions(), which doesn't work properly with the PCI BAR
lazy allocation, because pci_resource_type() -> _pci_get_rle() always
returns NULL, so pci_request_regions() doesn't have the opportunity to
invoke bus_alloc_resource_any(). We now use pci_find_bar() in
pci_resource_type(), which is able to locate all available PCI BARs
even if some of them will be lazy allocated.
Submitted by: Dexuan Cui <decui microsoft com>
Reviewed by: hps
MFC after: 1 week
Sponsored by: Microsoft
Differential Revision: https://reviews.freebsd.org/D8071
Linux module parameters have a permissions value. If any write bits
are set we are allowed to modify the module parameter runtime. Reflect
this when creating the static SYSCTL nodes.
Sponsored by: Mellanox Technologies
MFC after: 1 week
Bool module parameters are no longer supported, because there is no
equivalent in FreeBSD.
There are two macros available which control the behaviour of the
LinuxKPI module parameters:
- LINUXKPI_PARAM_PARENT allows the consumer to set the SYSCTL parent
where the modules parameters will be created.
- LINUXKPI_PARAM_PREFIX defines a parameter name prefix, which is
added to all created module parameters.
Sponsored by: Mellanox Technologies
MFC after: 1 week
run after a panic(). This for example allows a LinuxKPI based graphics
stack to receive prints during a panic.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
at it use NULL for some pointer checks.
Bump the FreeBSD version to force recompilation of all kernel modules
due to a structure size change.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
streamline the rest of the xxx_to_jiffies() functions to have a
constant 64-bit argument and use identical range checks for the
result.
Specifically preserve msecs_to_jiffies(0) returning 0. See r282743 for
further details.
MFC after: 1 week
Sponsored by: Mellanox Technologies
error code checks might fail. ERESTART is in the BSD world defined as
-1. While at it add more Linux error codes.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
Linux requires that all IOCTL data resides in userspace. FreeBSD
always moves the main IOCTL structure into a kernel buffer before
invoking the IOCTL handler and then copies it back into userspace,
before returning. Hide this difference in the "linux_copyin()" and
"linux_copyout()" functions by remapping userspace addresses in the
range from 0x10000 to 0x20000, to the kernel IOCTL data buffer.
It is assumed that the userspace code, data and stack segments starts
no lower than memory address 0x400000, which is also stated by "man 1
ld", which means any valid userspace pointer can be passed to regular
LinuxKPI handled IOCTLs.
Bump the FreeBSD version to force recompilation of all kernel modules.
Discussed with: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
"current" inside all LinuxKPI file operation callbacks. The "current"
is frequently used for various debug prints, printing the thread name
and thread ID for example.
Obtained from: kmacy @
MFC after: 1 week
Sponsored by: Mellanox Technologies
Ensure the actual poll result is returned by the "linux_file_poll()"
function instead of zero which means no data is available.
MFC after: 3 days
Sponsored by: Mellanox Technologies
This is kinda critical to the performance when the CPU is slow and
network bandwidth is high, e.g. in the hypervisor.
Reviewed by: rrs, gallatin, Dexuan Cui <decui microsoft com>
Sponsored by: Microsoft OSTC
Differential Revision: https://reviews.freebsd.org/D5765
- Add some new hlist macros.
- Update existing hlist macros removing the need for a temporary
iteration variable.
- Properly define the RCU hlist macros to be SMP safe with regard
to RCU.
- Safe list macro arguments by adding a pair of parentheses.
- Prefix the _list_add() and _list_splice() functions with "linux"
to reflect they are LinuxKPI internal functions.
Obtained from: Linux
MFC after: 1 week
Sponsored by: Mellanox Technologies
- Fix implementation of atomic_add_unless(). The atomic_cmpset_int()
function returns a boolean and not the previous value of the atomic
variable.
- The atomic counters should be signed according to Linux.
- Some minor cosmetics and styling while at it.
Reviewed by: alfred @
MFC after: 1 week
Sponsored by: Mellanox Technologies
idr_alloc_cyclic() in the LinuxKPI. Bump the FreeBSD version to
force recompilation of all KLDs due to IDR structure size change.
MFC after: 2 weeks
Sponsored by: Mellanox Technologies
If a driver's Linux mmap callback passed vm_page_prot through unchanged,
then linux_dev_mmap_single() would try to apply whatever VM_MEMATTR_xxx
value 0 is to the mapping. On x86, VM_MEMATTR_DEFAULT is the PAT value
for write-back (WB) which is 6, while 0 maps to the PAT value for
uncacheable (UC). Thus, any mmap request that did not explicitly set
page_prot was tried to map memory as UC triggering the warning in
sg_pager_getpages().
Tested by: np
Reported by: Krishnamraju Eraparaju @ Chelsio
MFC after: 3 days
Sponsored by: Chelsio Communications
