Age | Commit message (Collapse) | Author |
|
Change the parameters of `comedi_buf_is_mmapped()` to pass a pointer to
the comedi subdevice instead of a pointer to the "async" structure
belonging to the subdevice.
The main aim at the moment is to replace all the `struct comedi_async *`
parameters with `struct comedi_subdevice *` parameters in the comedi
driver API.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Change the parameters of `comedi_buf_reset()` to pass a pointer to the
comedi subdevice instead of a pointer to the "async" structure belonging
to the subdevice.
The main aim at the moment is to replace all the `struct comedi_async *`
parameters with `struct comedi_subdevice *` parameters in the comedi
driver API.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Change the parameters of `comedi_buf_write_n_allocated()` to pass a
pointer to the comedi subdevice instead of a pointer to the "async"
structure belonging to the subdevice.
The main aim at the moment is to replace all the `struct comedi_async *`
parameters with `struct comedi_subdevice *` parameters in the comedi
driver API.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Mmapping a comedi data buffer with lockdep checking enabled produced the
following kernel debug messages:
======================================================
[ INFO: possible circular locking dependency detected ]
3.5.0-rc3-ija1+ #9 Tainted: G C
-------------------------------------------------------
comedi_test/4160 is trying to acquire lock:
(&dev->mutex#2){+.+.+.}, at: [<ffffffffa00313f4>] comedi_mmap+0x57/0x1d9 [comedi]
but task is already holding lock:
(&mm->mmap_sem){++++++}, at: [<ffffffff810c96fe>] vm_mmap_pgoff+0x41/0x76
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (&mm->mmap_sem){++++++}:
[<ffffffff8106d0e8>] lock_acquire+0x97/0x105
[<ffffffff810ce3bc>] might_fault+0x6d/0x90
[<ffffffffa0031ffb>] do_devinfo_ioctl.isra.7+0x11e/0x14c [comedi]
[<ffffffffa003227f>] comedi_unlocked_ioctl+0x256/0xe48 [comedi]
[<ffffffff810f7fcd>] vfs_ioctl+0x18/0x34
[<ffffffff810f87fd>] do_vfs_ioctl+0x382/0x43c
[<ffffffff810f88f9>] sys_ioctl+0x42/0x65
[<ffffffff81415c62>] system_call_fastpath+0x16/0x1b
-> #0 (&dev->mutex#2){+.+.+.}:
[<ffffffff8106c528>] __lock_acquire+0x101d/0x1591
[<ffffffff8106d0e8>] lock_acquire+0x97/0x105
[<ffffffff8140c894>] mutex_lock_nested+0x46/0x2a4
[<ffffffffa00313f4>] comedi_mmap+0x57/0x1d9 [comedi]
[<ffffffff810d5816>] mmap_region+0x281/0x492
[<ffffffff810d5c92>] do_mmap_pgoff+0x26b/0x2a7
[<ffffffff810c971a>] vm_mmap_pgoff+0x5d/0x76
[<ffffffff810d493f>] sys_mmap_pgoff+0xc7/0x10d
[<ffffffff81004d36>] sys_mmap+0x16/0x20
[<ffffffff81415c62>] system_call_fastpath+0x16/0x1b
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(&mm->mmap_sem);
lock(&dev->mutex#2);
lock(&mm->mmap_sem);
lock(&dev->mutex#2);
*** DEADLOCK ***
To avoid the circular dependency, just try to get the lock in
`comedi_mmap()` instead of blocking. Since the comedi device's main mutex
is heavily used, do a down-read of its `attach_lock` rwsemaphore
instead. Trying to down-read `attach_lock` should only fail if
some task has down-write locked it, and that is only done while the
comedi device is being attached to or detached from a low-level hardware
device.
Unfortunately, acquiring the `attach_lock` doesn't prevent another
task replacing the comedi data buffer we are trying to mmap. The
details of the buffer are held in a `struct comedi_buf_map` and pointed
to by `s->async->buf_map` where `s` is the comedi subdevice whose buffer
we are trying to map. The `struct comedi_buf_map` is already reference
counted with a `struct kref`, so we can stop it being freed prematurely.
Modify `comedi_mmap()` to call new function
`comedi_buf_map_from_subdev_get()` to read the subdevice's current
buffer map pointer and increment its reference instead of accessing
`async->buf_map` directly. Call `comedi_buf_map_put()` to decrement the
reference once the buffer map structure has been dealt with. (Note that
`comedi_buf_map_put()` does nothing if passed a NULL pointer.)
`comedi_buf_map_from_subdev_get()` checks the subdevice's buffer map
pointer has been set and the buffer map has been initialized enough for
`comedi_mmap()` to deal with it (specifically, check the `n_pages`
member has been set to a non-zero value). If all is well, the buffer
map's reference is incremented and a pointer to it is returned. The
comedi subdevice's spin-lock is used to protect the checks. Also use
the spin-lock in `__comedi_buf_alloc()` and `__comedi_buf_free()` to
protect changes to the subdevice's buffer map structure pointer and the
buffer map structure's `n_pages` member. (This checking of `n_pages` is
a bit clunky and I [Ian Abbott] plan to deal with it in the future.)
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Cc: <stable@vger.kernel.org> # 3.14.x, 3.15.x
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
If a comedi device is automatically detached by `comedi_auto_unconfig()`
any data buffers associated with subdevices that support asynchronous
commands will be freed. If the buffer is mmapped at the time, bad
things are likely to happen! Prevent this by moving some of the buffer
details from `struct comedi_async` into a new, dynamically allocated,
and kref-counted `struct comedi_buf_map`. This holds a list of pages, a
reference count, and enough information to free the pages. The new
member `buf_map` of `struct comedi_async` points to a `struct
comedi_buf_map` when the buffer size is non-zero.
Provide a new helper function `comedi_buf_is_mapped()` to check whether
an a buffer is mmapped. If it is mmapped, the buffer is not allowed to
be resized and the device is not allowed to be manually detached by the
`COMEDI_DEVCONFIG` ioctl. Provide helper functions
`comedi_buf_map_get()` and `comedi_buf_map_put()` to manipulate the
reference count of the `struct comedi_buf_map`, which will be freed
along with its contents via the 'release' callback of the `kref_put()`
call. The reference count is manipulated by the vma operations and the
mmap file operation.
Now, when the comedi device is automatically detached, the buffer will
be effectively freed by calling `comedi_buf_alloc()` with a new buffer
size of 0. That calls local function `__comedi_buf_free()` which calls
`comedi_buf_map_put()` on the `buf_map` member to free it. It won't
actually be freed until the final 'put'.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
The comedi core module's handling of the `COMEDI_DEVCONFIG` ioctl will
not allow a device to be detached if it is busy. However, comedi
devices can also be auto-detached due to a removal of a hardware device.
One of the things we should do in that case is cancel any asynchronous
commands that are running. Add a new function
`comedi_device_cancel_all()` to do that and call it from
`comedi_device_detach()`.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Low-level comedi drivers registered with the comedi core by
`comedi_driver_register()` are linked together into a simple linked list
headed by the `comedi_drivers` variable and chained by the `next` member
of `struct comedi_driver`. A driver is removed from the list by
`comedi_driver_unregister()`. The driver list is iterated through by
`comedi_device_attach()` when the `COMEDI_DEVCONFIG` ioctl is used to
attach a "legacy" device to a driver, and is also iterated through by
`comedi_read()` in "proc.c" when reading "/proc/comedi".
There is currently no protection against items being added or removed
from the list while it is being iterated. Add a mutex
`comedi_drivers_list_lock` to be locked while adding or removing an item
on the list, or when iterating through the list.
`comedi_driver_unregister()` also checks for and detaches any devices
using the driver. This is currently done before unlinking the driver
from the list, but it makes more sense to unlink the driver from the
list first to prevent `comedi_device_attach()` attempting to use it, so
move the unlinking part to the start of the function. Also, in
`comedi_device_attach()` hold on to the mutex until we've finished
attempting to attach the device to avoid it interfering with the
detachment in `comedi_driver_unregister()`.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Change `comedi_alloc_board_minor()` to return a pointer to the allocated
`struct comedi_device` instead of a minor device number. Return an
`ERR_PTR()` value on error instead of a negative error number. This
saves a call to `comedi_dev_from_minor()` in `comedi_auto_config()`.
Also change it to use a local variable `dev` to hold the pointer to the
`struct comedi_device` instead of using `info->device` all the time.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
`comedi_free_board_minor()` is now only called from the same .c file it
is defined in, so give it static linkage.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Add `comedi_release_hardware_device()` as a replacement for the call
sequence `comedi_find_board_minor()`, `comedi_free_board_minor()`. This
is slightly safer as we can make sure nothing funny happens to the found
`comedi_file_info_table[]` entry in the middle of the sequence. Change
`comedi_auto_unconfig()` to call the new function instead of the old
sequence. Remove `comedi_find_board_minor()` as it has no other
callers.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Reviewed-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
This module parameter is used to enable the auto config mechanism
in the comedi core. Most of the PCI, PCMCIA, and USB drivers have
been converted to use the auto config mechanism and will not attach
if it is disabled.
Since the 'comedi_autoconfig' parameter is defaulted to true, just
remove it so that the comedi drivers that use auto config will
always be able to attach.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Cc: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
`comedi_alloc_subdevice_minor()` doesn't really need a parameter
pointing to a `struct comedi_device` as it can get this information from
the parameter pointing to a `struct comedi_subdevice`. Just pass the
subdevice parameter.
Signed-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Move the declarations of `comedi_alloc_subdevice_minor()` and
`comedi_free_subdevice_minor()` from "comedidev.h" to
"comedi_internal.h" since they are only of interest to the comedi core,
and are not exported to the low-level comedi drivers.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
These functions are only used by the comedi core. Move the
prototypes to comedi_internal.h so they are not exposed to
the comedi drivers.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Cc: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
These functions are only used by the comedi core. Move the
prototypes to comedi_internal.h so they are not exposed to
the comedi drivers.
Tidy up comedi_internal.h a bit so that all the internal stuff
in drivers.c is grouped.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Cc: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
This helper function is only called by the comedi core in comedi_buf.c
and comedi_fops.c. For aesthetic reasons, move it to comedi_buf.c and
remove the inline. Move the prototype from comedidev.h to comedi_internal.h
so it's not exposed outside the comedi core.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
For aesthetic reasons, rename this function to comedi_buf_reset(). This
makes all the asynchronous buffer functions have the same namespace.
Signed-off-by: H Hartley Sweeten <hsweeten@visionengravers.com>
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Move the contents of "comedi_fops.h" into "comedi_internal.h" and delete
"comedi_fops.h". It only contains a couple of external variable
declarations (and #include <linux/types.h>) and one of those isn't even
declared in "comedi_fops.c". The other one is an external declaration
of a variable used to store a module parameter and some of those are
already externally declared in "comedi_internal.h", so they can keep it
company!
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Add a multiple inclusion guard macro to "comedi_internal.h".
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|
|
Use a less generic name for this internal header file included by
various parts of the comedi core.
Signed-off-by: Ian Abbott <abbotti@mev.co.uk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
|