System crashes, in initiator mode operation, with
qla2xxx_copy_atioqueues() in stack trace when firmware dump is
attempted.
Check for atio_q_length alone does not indicate if atio_ring is
allocated, make explicit check of atio_ring to avoid the crash.
Applicable to ISP24xx, ISP25xx, ISP81xx & ISP83xx line of HBAs.
Signed-off-by: Arun Easi <arun.easi@qlogic.com>
Signed-off-by: Saurav Kashyap <saurav.kashyap@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
The original patch was not covering all the adapters and firmwares.
This commit reverts 3a11711ad0.
Signed-off-by: Joe Carnuccio <joe.carnuccio@qlogic.com>
Signed-off-by: Saurav Kashyap <saurav.kashyap@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Reinitialize resource queue prior to freeing resource entries to ensure they
are not referenced. This fixes an issue with target_destoy accessing memory
after it was freed.
Signed-off-by: Wen Xiong <wenxiong@linux.vnet.ibm.com>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
The abort command issued by ipr_cancel_op() is being added to the wrong
HRRQ free queue after the command returns. Fix it by using the HRRQ
pointer in the ipr command struct itself.
Signed-off-by: Kleber Sacilotto de Souza <klebers@linux.vnet.ibm.com>
Signed-off-by: Wen Xiong <wenxiong@linux.vnet.ibm.com>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
This patch fixes a reference count bug in the SCSI tape driver which can be
reproduced with the following:
* Boot with slub_debug=FZPU, tape drive attached
* echo 1 > /sys/devices/... tape device pci path .../remove
* Wait for device removal
* echo 1 > /sys/kernel/slab/blkdev_queue/validate
* Slub debug complains about corrupted poison pattern
In commit 523e1d39 (block: make gendisk hold a reference to its queue)
add_disk() and disk_release() were modified to get/put an additional
reference on a disk queue to fix a reference counting discrepency
between bdev release and SCSI device removal. The ST driver never
calls add_disk(), so this commit introduced an extra kref put when the
ST driver frees its struct gendisk.
Attempts were made to fix this bug at the block level [1] but later
abandoned due to floppy driver issues [2].
[1] https://lkml.org/lkml/2012/8/27/354
[2] https://lkml.org/lkml/2012/9/22/113
Signed-off-by: Joe Lawrence <joe.lawrence@stratus.com>
Tested-by: Ewan D. Milne <emilne@redhat.com>
Acked-by: Kai Mäkisara <Kai.Makisara@kolumbus.fi>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
In fact the disc_resp buffer will be overwrite by smp response, so we never
found this typo, correct it by using the right one.
Signed-off-by: John Gong <john_gong@usish.com>
Signed-off-by: Jack Wang <jack_wang@usish.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
Loading ipr modules failed(-22) with msi-x enabled adapter. In ipr_test_msi(),
We need to pass the first vector of msix vectors instead of using pdev->irq
to request_irq() when adapter enables msix feature.
Signed-off-by: Wen Xiong <wenxiong@linux.vnet.ibm.com>
Acked-by: Brian King <brking@linux.vnet.ibm.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>
We need to verify that the given sockets actually are l2cap sockets. If
they aren't, we are not supposed to access bt_sk(sock) and we shouldn't
start the session if the offsets turn out to be valid local BT addresses.
That is, if someone passes a TCP socket to HIDCONNADD, then we access some
random offset in the TCP socket (which isn't even guaranteed to be valid).
Fix this by checking that the socket is an l2cap socket.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
We print this error twice in the first error-path so remove it. One error
message is enough.
Signed-off-by: David Herrmann <dh.herrmann@gmail.com>
Acked-by: Marcel Holtmann <marcel@holtmann.org>
Signed-off-by: Gustavo Padovan <gustavo.padovan@collabora.co.uk>
cache target when the device being cached is not itself wrapped with
device-mapper.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)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=LUeu
-----END PGP SIGNATURE-----
Merge tag 'dm-3.9-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-dm
Pull device-mapper fixes from Alasdair Kergon:
"A pair of patches to fix the writethrough mode of the device-mapper
cache target when the device being cached is not itself wrapped with
device-mapper."
* tag 'dm-3.9-fixes-2' of git://git.kernel.org/pub/scm/linux/kernel/git/agk/linux-dm:
dm cache: reduce bio front_pad size in writeback mode
dm cache: fix writes to cache device in writethrough mode
ASPM
Revert "PCI/ACPI: Request _OSC control before scanning PCI root bus"
kexec
PCI: Don't try to disable Bus Master on disconnected PCI devices
Platform ROM images
PCI: Add PCI ROM helper for platform-provided ROM images
nouveau: Attempt to use platform-provided ROM image
radeon: Attempt to use platform-provided ROM image
Hotplug
PCI/ACPI: Always resume devices on ACPI wakeup notifications
PCI/PM: Disable runtime PM of PCIe ports
EISA
EISA/PCI: Fix bus res reference
EISA/PCI: Init EISA early, before PNP
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
iQIcBAABAgAGBQJRXzmPAAoJEFmIoMA60/r8svoQAIb1Z11DAdPfTaXThwEIURZb
KC4wzfLMAZCRvfY6j6qZw6Qi3NT6ZqgrZxSXgV0myB6Luwcv80+D2BHmDnRoxRYD
ywAm2m8vwoUq5s7MsaGHRLViBafL5TDs1KQLVEPkJ+B1gCzo61UWXPVDu+z9u+Fx
t9fNV5Xc/uaxciXZYVnPGTCFjCVTqa95x16bF6NgmG1j1DHYuX7fBVSsJc2CS77m
P2oyrfLmsKUA7jKHD4ntbIslYcqUC5NF8wcRad324Pzi93IwdktUSn8tqqNFCX9T
EE8hxCp24CPm6ngyCFIIkqYM/MMDM8YVqBwuZjngm3+tjaS3hBIrr1jBDenUgncY
2WRomgRzBwaa8nTr5xhin3kYGRkQAEf3M8RWG70YBppTidvvoUt0bfxSfSlN9R/t
g/7Eu4u35+lJ9F+Gg4unc0JrydkSJG1tfE+0LaEOhqXgCwloPR3kk08qcHPA2NHQ
dlEMJ3FSUHvINoRE7pm6+E2lcbGvZjCa9Cvd9IbdJ6r4wVYI7pHl8Zlp/sf3HZdV
sKRAw7fSazoFCX73lHR0f1zx7l1OdTmDSxEHv/pQA+xVt7K0DcOq+Mj0B84pVYcl
WSRUcr6jAi91syuKX+8TRpbfACb25PkPrEFEsjsGcBq+I/a15e8a+sIW31Z66PNu
ICejTqh/2/aOlgFg9OBs
=gIGr
-----END PGP SIGNATURE-----
Merge tag 'pci-v3.9-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci
Pull PCI fixes from Bjorn Helgaas:
"PCI updates for v3.9:
ASPM
Revert "PCI/ACPI: Request _OSC control before scanning PCI root bus"
kexec
PCI: Don't try to disable Bus Master on disconnected PCI devices
Platform ROM images
PCI: Add PCI ROM helper for platform-provided ROM images
nouveau: Attempt to use platform-provided ROM image
radeon: Attempt to use platform-provided ROM image
Hotplug
PCI/ACPI: Always resume devices on ACPI wakeup notifications
PCI/PM: Disable runtime PM of PCIe ports
EISA
EISA/PCI: Fix bus res reference
EISA/PCI: Init EISA early, before PNP"
* tag 'pci-v3.9-fixes-1' of git://git.kernel.org/pub/scm/linux/kernel/git/helgaas/pci:
PCI/PM: Disable runtime PM of PCIe ports
PCI/ACPI: Always resume devices on ACPI wakeup notifications
PCI: Don't try to disable Bus Master on disconnected PCI devices
Revert "PCI/ACPI: Request _OSC control before scanning PCI root bus"
radeon: Attempt to use platform-provided ROM image
nouveau: Attempt to use platform-provided ROM image
EISA/PCI: Init EISA early, before PNP
EISA/PCI: Fix bus res reference
PCI: Add PCI ROM helper for platform-provided ROM images
Use the wrapper functions for getting and setting the driver data using
spi_device instead of using dev_{get|set}_drvdata with &spi->dev, so we
can directly pass a struct spi_device.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
debugfs exposes device state and list of me clients and their
properties
Signed-off-by: Tomas Winkler <tomas.winkler@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use the wrapper functions for getting and setting the driver data using
spi_device instead of using dev_{get|set}_drvdata with &spi->dev, so we
can directly pass a struct spi_device.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use the wrapper functions for getting and setting the driver data using
spi_device instead of using dev_{get|set}_drvdata with &spi->dev, so we
can directly pass a struct spi_device.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
tty_port_tty_get might return a tty which is NULL. But it is
dereferenced unconditionally in edge_send. Stop dereferencing that by
sending usb_serial_port pointer around.
Signed-off-by: Jiri Slaby <jslaby@suse.cz>
Cc: Johan Hovold <jhovold@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use the wrapper functions for getting and setting the driver data using
spi_device instead of using dev_{get|set}_drvdata with &spi->dev, so we
can directly pass a struct spi_device.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
It might be a kernel disaster if one sysfs entry is freed but
still referenced by sysfs tree.
Recently Dave and Sasha reported one use-after-free problem on
sysfs entry, and the problem has been troubleshooted with help
of debug message added in this patch.
Given sysfs_get_dirent/sysfs_put are exported APIs, even inside
sysfs they are called in many contexts(kobject/attribe add/delete,
inode init/drop, dentry lookup/release, readdir, ...), it is healthful
to check the removed flag before freeing one entry and dump message
if it is freeing without being removed first.
Cc: Dave Jones <davej@redhat.com>
Cc: Sasha Levin <levinsasha928@gmail.com>
Signed-off-by: Ming Lei <ming.lei@canonical.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Here is the big Gadget & PHY pull request. Many of us have
been really busy lately getting multiple drivers to a better
position.
Since this pull request is so large, I will divide it in sections
so it's easier to grasp what's included.
- cleanups:
. UDC drivers no longer touch gadget->dev, that's now udc-core
responsibility
. Many more UDC drivers converted to usb_gadget_map/unmap_request()
. UDC drivers no longer initialize DMA-related fields from gadget's
device structure
. UDC drivers don't touch gadget.dev.driver directly
. UDC drivers don't assign gadget.dev.release directly
. Removal of some unused DMA_ADDR_INVALID
. Introduction of CONFIG_USB_PHY
. All phy drivers have been moved to drivers/usb/phy and renamed to
a common naming scheme
. Fix PHY layer so it never returns a NULL pointer, also fix all
callers to avoid using IS_ERR_OR_NULL()
. Sparse fixes all over the place
. drivers/usb/otg/ has been deleted
. Marvel drivers (mv_udc, ehci-mv, mv_otg and mv_u3d) improved clock
usage
- new features:
. UDC core now provides a generic way for tracking and reporting
UDC's state (not attached, resuming, suspended, addressed,
default, etc)
. twl4030-usb learned that it shouldn't be enabled during init
. Full DT support for DWC3 has been implemented
. ab8500-usb learned about pinctrl framework
. nop PHY learned about DeviceTree and regulators
. DWC3 learned about suspend/resume
. DWC3 can now be compiled in host-only and gadget-only (as well as
DRD) configurations
. UVC now enables streaming endpoint based on negotiated speed
. isp1301 now implements the PHY API properly
. configfs-based interface for gadget drivers which will lead to
the removal of all code which just combines functions together
to build functional gadget drivers.
. f_serial and f_obex were converted to new configfs interface while
maintaining old interface around.
- non-critical fixes:
. UVC gadget driver got fixes for Endpoint usage and stream calculation
. ab8500-usb fixed unbalanced clock and regulator API usage
. twl4030-usb got a fix for when OMAP3 is booted with cable connected
. fusb300_udc got a fix for DMA usage
. UVC got fixes for two assertions of the USB Video Class Compliance
specification revision 1.1
. build warning issues caused by recent addition of __must_check to
regulator API
These are all changes which deserve a mention, all other changes are related
to these one or minor spelling fixes and other similar tasks.
Signed-of-by: Felipe Balbi <balbi@ti.com>
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)
iQIcBAABAgAGBQJRXG8GAAoJEIaOsuA1yqREJzYQAKMLW3J/TKTWS1DDuf7qMtMz
Ug6qChKZXgT1/QrNjsq2tx4jYIkNdSMtRKiUx0BnIptlUx6gM22gcsN8mXX/UJjC
FYAiWl+tYe85e9uayqqt+qVCZjTZCc7St4wQalugDHefvA7yCbiZpSaJRGlJMK+x
mePJ7MfrulDsYBXr0u+m2LOJ0qxMDi40k3/UN3aUu5yzrmBiRpVq1mySruvLwGFp
Pr6vBnprEc6bW5sRdUR4SICKLvLk5sHwHpvpkzDLYBIb/jXQwbfQri/HKeh4VMk8
trbsvHZyB7H8uuFsCHiBc6VtjcbZ4mxPUK+1PCq8hG077avdkm3ox0BERk9aRTeC
jg4mdpyWjgovwi882woPEQXNZoaAXpVDyI8tBRx92a+rGJjXSHhLQI+4Ffi4ZvzV
d+q1ZzrHxuzwa/BwcPETY76umXQqXWXg+ap1bHDY0RZFoPLdXMpl583NXGSn3gOD
dUlD0UlgYwb75333tRIPNQn3qOx0HVd6MlYPMNzl9t9c9qqfX78AYRny6ZucupRg
t9VZ6FO3D2yre9W7u3U3q2c9H7uSAKr/8xaNfvdsIWPncgvvYVIyE8MnR7AiHoPv
ZYESs/Gs6w9vUsHa9K4J16Ape7D3AMcGpXUoPUxTBHrwBexzt4j27VWtcL4ns/9r
0kcltUJ4Zq+PIjc7xgxe
=aF4r
-----END PGP SIGNATURE-----
Merge tag 'usb-for-v3.10' of git://git.kernel.org/pub/scm/linux/kernel/git/balbi/usb into usb-next
Felipe writes:
usb: patches for v3.10 merge window
Here is the big Gadget & PHY pull request. Many of us have
been really busy lately getting multiple drivers to a better
position.
Since this pull request is so large, I will divide it in sections
so it's easier to grasp what's included.
- cleanups:
. UDC drivers no longer touch gadget->dev, that's now udc-core
responsibility
. Many more UDC drivers converted to usb_gadget_map/unmap_request()
. UDC drivers no longer initialize DMA-related fields from gadget's
device structure
. UDC drivers don't touch gadget.dev.driver directly
. UDC drivers don't assign gadget.dev.release directly
. Removal of some unused DMA_ADDR_INVALID
. Introduction of CONFIG_USB_PHY
. All phy drivers have been moved to drivers/usb/phy and renamed to
a common naming scheme
. Fix PHY layer so it never returns a NULL pointer, also fix all
callers to avoid using IS_ERR_OR_NULL()
. Sparse fixes all over the place
. drivers/usb/otg/ has been deleted
. Marvel drivers (mv_udc, ehci-mv, mv_otg and mv_u3d) improved clock
usage
- new features:
. UDC core now provides a generic way for tracking and reporting
UDC's state (not attached, resuming, suspended, addressed,
default, etc)
. twl4030-usb learned that it shouldn't be enabled during init
. Full DT support for DWC3 has been implemented
. ab8500-usb learned about pinctrl framework
. nop PHY learned about DeviceTree and regulators
. DWC3 learned about suspend/resume
. DWC3 can now be compiled in host-only and gadget-only (as well as
DRD) configurations
. UVC now enables streaming endpoint based on negotiated speed
. isp1301 now implements the PHY API properly
. configfs-based interface for gadget drivers which will lead to
the removal of all code which just combines functions together
to build functional gadget drivers.
. f_serial and f_obex were converted to new configfs interface while
maintaining old interface around.
- non-critical fixes:
. UVC gadget driver got fixes for Endpoint usage and stream calculation
. ab8500-usb fixed unbalanced clock and regulator API usage
. twl4030-usb got a fix for when OMAP3 is booted with cable connected
. fusb300_udc got a fix for DMA usage
. UVC got fixes for two assertions of the USB Video Class Compliance
specification revision 1.1
. build warning issues caused by recent addition of __must_check to
regulator API
These are all changes which deserve a mention, all other changes are related
to these one or minor spelling fixes and other similar tasks.
Signed-of-by: Felipe Balbi <balbi@ti.com>
Add local function `comedi_clear_board_dev()` as a safer alternative to
`comedi_clear_board_minor()` when we already have a pointer to a `struct
comedi_device`. It uses the board minor device number stored in the
`struct comedi_device` (which must have already been initialized) and
only clears the entry in `comedi_board_minor_table[]` if it points to
the specified `struct comedi_device`. Rather than returning the old
table entry, it returns `true` if the entry matched (and so has just
been cleared) and returns `false` otherwise.
Call `comedi_clear_board_dev()` instead of `comedi_clear_board_minor()`
in `comedi_unlocked_ioctl()` (in the code that frees a dynamically
allocated comedi device detached by the `COMEDI_DEVCONFIG` ioctl). That
ought to return `true` but check it just in case before freeing the
device. There is still a race condition here which needs to be dealt
with once we've implemented reference counting for `struct
comedi_device`s.
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>
`struct comedi_file_info` is no longer used so remove it.
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_alloc_board_minor()` allocates and initializes a `struct
comedi_file_info` and a `struct comedi_device`, and assigns a board
minor device number (if there are any available), storing a pointer to
the allocated `struct comedi_file_info` in
`comedi_board_minor_table[minor]` where `minor` is the board minor
device number.
There is no longer anything useful in the `struct comedi_file_info`
apart from the pointer to the `struct comedi_device` that was allocated,
so the `struct comedi_file_info` is superfluous.
Change `comedi_board_minor_table[]` to hold pointers to the actual
`struct comedi_device`'s. `comedi_alloc_board_minor()` no longer needs
to allocate a `struct comedi_file_info`. Replace
`comedi_free_board_file_info()` with `comedi_free_board_dev()` with its
parameter pointing to the `struct comedi_device` to be freed (there is
no longer a `struct comedi_file_info` to be freed).
There are consequential changes to `comedi_dev_from_board_minor()`,
`comedi_clear_board_minor()` (which now returns a `struct comedi_device
*`), `comedi_free_board_minor()`, `comedi_release_hardware_device()` and
`comedi_unlocked_ioctl()` (when dealing with detachment of a dynamically
allocated comedi device by the `COMEDI_DEVCONFIG` ioctl).
`comedi_dev_from_file_info()` is no longer used as a result of the above
changes so remove it.
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_alloc_subdevice_minor()` allocates and initializes a `struct
comedi_file_info` and assigns a subdevice minor device number (if there
are any available), storing a pointer to the allocated `struct
comedi_file_info` in `comedi_subdevice_minor_table[i]` where `i` is the
array index corresponding to the subdevice minor device number (indexed
by subdevice minor device number minus `COMEDI_NUM_BOARD_MINORS`).
The information stored in the `struct comedi_file_info` can be derived
from the subdevice structure (`struct comedi_subdevice`) itself, so the
`struct comedi_file_info` is superfluous.
Change `comedi_subdevice_minor_table[]` to hold pointers to the actual
`struct comedi_subdevice`'s. `comedi_alloc_subdevice_minor()` no longer
needs to allocate a `struct comedi_file_info` and
`comedi_free_subdevice_info()` no longer has a `struct comedi_file_info`
to free.
Replace `comedi_file_info_from_minor()` with
`comedi_subdevice_from_minor()`, returning a (possibly NULL) pointer to
a `struct comedi_subdevice` from the table. This has knock-on effects
for `comedi_dev_from_subdevice_minor()`, `comedi_read_subdevice()` and
`comedi_write_subdevice()`. In particular, `comedi_read_subdevice()`
and `comedi_write_subdevice()` now need to check the subdevice flags to
see if the determine whether to override the comedi device's default
read/write subdevice.
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>
Merge the code from `comedi_file_info_from_board_minor()` into
`comedi_dev_from_board_minor()` and remove it since nothing else calls
it.
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>
Merge the code from `comedi_clear_subdevice_minor()` into
`comedi_free_subdevice_minor()` and remove it since nothing else calls
it.
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>
Refactor `comedi_dev_from_minor()` to call one of two new functions
`comedi_dev_from_board_minor()` (for minor device numbers less than
`COMEDI_NUM_BOARD_MINORS`) or `comedi_dev_from_subdevice_minor()` (for
minor device numbers greater than or equal to `COMEDI_NUM_BOARD_MINORS`,
which are subdevice minor device numbers).
Remove `comedi_file_info_from_minor()` as it is no longer used.
This is a step towards removing `struct comedi_file_info`.
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>
Since the parameters of `comedi_read_subdevice()` and
`comedi_write_subdevice()` have changed, there is no longer a need to
keep a local variable `struct comedi_file_info *info` hanging around in
the file operation functions and device attribute access functions.
Remove the variable from those function and replace the call sequence
`info = comedi_file_info_from_minor(minor)`, `dev =
comedi_dev_from_file_info(info)` with the simpler `dev =
comedi_dev_from_minor(minor)`.
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_read_subdevice()` and `comedi_write_subdevice()` currently take
a single parameter pointing to a `struct comedi_file_info`. I'm trying
to get rid of `struct comedi_file_info` so as part of that plan,
`comedi_read_subdevice()` and `comedi_write_subdevice()` need to change.
Change them to take two parameters: a pointer to a `struct
comedi_device` and a minor device number. If the minor device number is
a "board" minor device number (`minor < COMEDI_NUM_BOARD_MINORS`) we'll
always return the default read or write subdevice. If the minor device
number if a "subdevice" minor device number a different read or write
subdevice may be returned. In that case, use the subdevice minor device
number to look up the information about whether the default read or
write subdevice needs to be overridden.
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>
The comedi device attribute functions such as
`show_max_read_buffer_kb()` call `dev_get_drvdata()` to get a pointer to
a `struct comedi_file_info` from the private driver data field of class
device. Change them to use the minor device number to look up this
pointer value so they behave more like the file operation functions.
Check the pointer is non-NULL as the entry in the minor device table
could have been set to NULL. Note that there is still a race condition
in the use of this pointer value after acquiring the mutex which needs
to be dealt with once reference counting has been implemented for comedi
devices.
The calls to `dev_set_drvdata()` from `comedi_alloc_board_minor()` and
`comedi_alloc_subdevice_minor()` are no longer needed so remove them.
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>
The comedi device attribute functions such as
`show_read_buffer_kb()` or `show_write_buffer_kb()` call
`comedi_read_subdevice()` or `comedi_write_subdevice()` without
acquiring the comedi device's mutex first, although the functions do
acquire the mutex afterwards. Change them to acquire the mutex first.
This is consistent with most of the comedi device file operation
functions (apart from `comedi_read()` and `comedi_write()` which
probably need looking at).
Despite the use of the mutex, there are still race conditions as the
`struct comedi_file_info *info` variable value set before acquiring the
mutex could be stale after acquiring the mutex. This problem will be
dealt with once reference counting has been implemented for the comedi
devices.
This patch also adds local variable `struct comedi_device *dev` to the
functions to reduce the use of the `info` variable a little bit.
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>
Most of the comedi core uses the identifier `dev` to point to a `struct
comedi_device`. The device sysfs attribute functions such as
`show_max_read_buffer_kb()` use the parameter id `dev` to point to a
`struct device`. Rename the parameter to `csdev` for "class device" for
consistency with the functions that call `device_create()` to create
these class devices (`comedi_alloc_board_minor()` and
`comedi_alloc_subdevice_minor()`).
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>
The `hardware_device` member of `struct comedi_device_info` is only set
to point to a hardware device by the auto-configuration code (specifically
`comedi_alloc_board_minor() with a non-NULL `hardware_device` parameter)
so that it can be found again by the auto-unconfiguration code
(specifically `comedi_release_hardware_device()`). However, as
`info->device->hw_dev` (where `info` is a pointer to the `struct
comedi_device_info`) is set to the same value as `info->hardware_device`
(by calling `comedi_set_hw_dev()` with the same hardware device pointer)
we can look for that instead, so there is no need for the
`hardware_device` member any more. Get rid of it.
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_clear_minor()` calls either `comedi_clear_board_minor()` or
`comedi_clear_subdevice_minor()` depending on its minor device number
parameter. The one it calls only depends on where it is called from, so
bypass the calls to `comedi_clear_minor()` and remove it.
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>
The comedi core reserves minor device numbers from 0 to
`COMEDI_NUM_BOARD_MINORS - 1` (0 to 0x30 - 1) for the main comedi
"board" devices and reserves minor device numbers from
`COMEDI_NUM_BOARD_MINORS` to `COMEDI_NUM_MINORS - 1` (0x30 to 0x100 - 1)
for comedi subdevices (or at least those that claim to support
asynchronous comedi commands). There is an array
`comedi_file_info_table[COMEDI_NUM_MINORS]` used to hold pointers to
information for each board minor device number and subdevice minor
device number that has been allocated (with NULL pointers for those not
allocated), along with a protective lock `comedi_file_info_table_lock`.
Since the ranges of board minor device numbers and subdevice minor
device numbers do not overlap, we can use separate tables and separate
locks for the different types of minor device numbers. This will allow
us to use different pointer types for the elements of each table in the
future without just using a generic `void *`. (At the moment, the table
elements point to a `struct comedi_file_info` allocated dynamically for
each allocated board minor device or subdevice minor device, but I plan
to get rid of that data structure.)
Replace `comedi_file_info_table[COMEDI_NUM_MINORS]` with two new arrays
of the same type, `comedi_board_minor_table[COMEDI_NUM_BOARD_MINORS]`
for board minors, and
`comedi_subdevice_minor_table[COMEDI_NUM_SUBDEVICE_MINORS]` for
subdevice minors (where `COMEDI_NUM_SUBDEVICE_MINORS` is
`COMEDI_NUM_MINORS - COMEDI_NUM_BOARD_MINORS`).
`comedi_subdevice_minor_table[]` is indexed by the subdevice minor
number minus `COMEDI_NUM_BOARD_MINORS` since `COMEDI_NUM_BOARD_MINORS`
is the first valid subdevice minor number.
Replace `comedi_file_info_table_lock` with
`comedi_board_minor_table_lock` for board minors and
`comedi_subdevice_minor_table_lock` for subdevice minors.
Refactor `comedi_clear_minor()` to call one of two new functions
`comedi_clear_board_minor()` and `comedi_clear_subdevice_minor()`
depending on the minor device number passed as a parameter. Similarly,
refactor `comedi_file_info_from_minor()` to call one of two new
functions `comedi_file_info_from_board_minor()` and
`comedi_file_info_from_subdevice_minor()` depending on the minor device
number parameter.
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>
The macro `COMEDI_FIRST_SUBDEVICE_MINOR` just expands to another macro
`COMEDI_NUM_BOARD_MINORS`. Replace uses of
`COMEDI_FIRST_SUBDEVICE_MINOR` with `COMEDI_NUM_BOARD_MINORS` and get
rid of the former.
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>
The macro definitions `COMEDI_NUM_MINORS` and
`COMEDI_FIRST_SUBDEVICE_MINOR` are only used in "comedi_fops.c" so move
them to there from "comedidev.h".
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>
The spin-lock `comedi_file_info_table_lock` is used to protect against
simultaneous modification and access of `comedi_file_info_table[]`.
Change it from a spin-lock to a mutex as it is only used in contexts
where sleeping is allowed.
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>
Call `comedi_set_hw_dev()` to set the `hw_dev` member of `struct
comedi_device` in `comedi_alloc_board_minor()` instead of in
`comedi_auto_config()`. Don't bother to check for an error returned by
`comedi_set_hw_dev()` here; it only fails when changing a non-NULL
pointer to a different non-NULL pointer and since the `struct
comedi_device` has just been allocated and initialized, its `hw_dev`
will be NULL already. Calling `comedi_set_hw_dev()` with a non-NULL
hardware device pointer increments the kref counter for the hardware
device.
If `comedi_alloc_board_minor()` fails further down the function, we rely
on its call to `comedi_device_cleanup()` to call `comedi_clear_hw_dev()`
(via `comedi_device_detach()` and `cleanup_device()`) to clear `hw_dev`
and decrement its kref counter. (That's the "beneficial side-effect"
mentioned in the patch that replaced `__comedi_device_detach()`.)
Remove the call to `comedi_set_hw_dev()` from `comedi_auto_config()` as
the call to `comedi_alloc_board_minor()` does it for us.
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_device_postconfig()` calls `comedi_device_detach()` on failure.
Remove that call and make the callers of `comedi_device_postconfig()`
call `comedi_device_detach()` themselves if it returns an error. This
seems more logical as the callers of `comedi_device_postconfig()` called
`comedi_device_detach()` anyway if they didn't call
`comedi_device_postconfig()`.
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_device_detach()` does nothing if the `struct comedi_device`'s
`attached` member is false, otherwise it calls
`__comedi_device_detach()` to do the real work.
`__comedi_device_detach()` is called from various other functions in
"drivers.c" (`comedi_device_postconfig()`, `comedi_device_attach()`, and
`comedi_auto_config()`) to bypass the check for the `attached` member
being false.
If we make `__comedi_device_detach()` safe to call when the `attached`
member is already false, we can remove the check in
`comedi_device_detach()`, subsume `__comedi_device_detach()` within
`comedi_device_detach()`, and replace all the calls to
`__comedi_device_detach()` with calls to `comedi_device_detach()`.
In fact, it is already safe to call `__comedi_device_detach()` when the
`attached` member is false. We just need to remove the warning message
it outputs when the `driver` member is NULL. Then the function becomes
idempotent without outputting spurious warnings. (It is idempotent
because `dev->driver->detach()` will only be called once at most and the
call to `cleanup_device()` is idempotent itself.)
Combine `comedi_device_detach()` with `__comedi_device_detach()`,
removing the check for the `attached` member being false and removing
the warning about the `driver` member being NULL, and replace all calls
to `__comedi_device_detach()` with calls to the combined
`comedi_device_detach()`.
A beneficial side-effect of the above change is that a call to
`comedi_device_detach()` will always result in a call to
`cleanup_device()` and so always result in a call to
`comedi_clear_hw_dev()`. We will make use of this beneficial
side-effect in a later patch.
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>
On failure of the call to the low-level comedi device driver's
`->attach()` handler from `__comedi_device_attach()`, reverse the
current ordering of the calls to `module_put()` and
`comedi_device_detach()` because `__comedi_device_detach()` will call
code in the module being put.
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>
For a legacy device attachment with the `COMEDI_DEVCONFIG` ioctl,
`do_devconfig_ioctl()` calls `comedi_device_attach()` to find a matching
device driver and attach the device. It then tries to increment the
device driver's module count and if that fails it detaches the device.
So on successful attachment of a device by the `COMEDI_DEVCONFIG` ioctl,
the device driver's module count will have been incremented.
`comedi_device_attach()` is called from nowhere else. It already
increments the device driver's module count temporarily and decrements
it again; if it gets as far as calling `comedi_device_postconfig()` the
module count is decremented within that function.
Simplify the above by removing the decrement of the device driver module
count from `comedi_device_postconfig()`. If the call to
`comedi_device_postconfig()` succeeds, `comedi_device_attach()` will
return with the module count still incremented, otherwise decrement the
module count before returning the error. Don't try and increment the
module count in `do_devconfig_ioctl()` after a successful return from
`comedi_device_attach()` as the module count has now already been
incremented.
`comedi_device_postconfig()` is also called by `comedi_auto_config()`
which currently has to increment the device driver's module count
temporarily so that `comedi_device_postconfig()` can decrement it, but
always returns with no overall change to the module count. Remove all
the module count manipulations from `comedi_device_postconfig()`. There
is no other reason for `comedi_auto_config()` to increment the device
driver's module count temporarily, since it is only called (indirectly)
from the device driver itself (usually via one of the wrappers
`comedi_pci_auto_config()` or `comedi_usb_auto_config()`).
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>
Move a few functions and variables to avoid a couple of forward
declarations.
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>
If a dynamically allocated (non-legacy, and automatically configured)
comedi device has been successfully unconfigured by use of the
`COMEDI_DEVCONFIG` ioctl, then remove the device afterwards.
(Dynamically identified comedi devices are identified by their minor
device number being `comedi_num_legacy_minors` or greater.) This is
done in `comedi_unlocked_ioctl()` on return from `do_devconfig_ioctl()`.
Note that there is an unlikely race condition with some other thread
that has just called `comedi_file_info_from_minor()` or
`comedi_dev_from_minor()` and is about to use the device, but that race
condition also exists for automatically removed devices and will be
dealt with properly once reference counting of comedi devices has been
implemented. We do avoid a race condition between automatic removal and
removal by the `COMEDI_DEVCONFIG` ioctl though.
Also add an extra precaution in `do_devconfig_ioctl()` to avoid
configuring a dynamically allocated device since there is a tight
window avoiding the race condition where this could happen and the
device is about to be removed anyway.
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>
When `comedi_unlocked_ioctl()` handled the `COMEDI_DEVCONFIG` ioctl via
a call to `do_devconfig()`, this used to be allowed to attach a legacy
hardware device to any board minor device, and the auto-unconfiguration
code for non-legacy used to search all board minor devices to find the
hardware device being removed. After calling `do_devconfig()`
successfully, `comedi_unlocked_ioctl()` sets the `hardware_device`
member of the `struct comedi_file_info` to NULL to stop the
auto-unconfiguration code finding it by accident and unconfiguring the
wrong comedi device as a result.
The above is no longer necessary because `do_devconfig()` now only
allows legacy hardware devices to be attached to comedi board minor
device numbers below `comedi_num_legacy_minors` and the
auto-unconfiguration code (particularly
`comedi_release_hardware_device()`) only searches comedi board minor
device numbers greater than or equal to `comedi_num_legacy_minors`, so
there is no overlap.
Remove the now unnecessary setting of `info->hardware_device = NULL` in
`comedi_unlocked_ioctl()`.
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_alloc_board_minor()` is called for both reserved "legacy"
devices created during module initialization, and for dynamically
created devices (via `comedi_auto_config()`). All the minor device
numbers from 0 to `comedi_num_legacy_minors - 1` are for legacy devices
and all those from `comedi_num_legacy_minors` to
`COMEDI_NUM_BOARD_MINORS - 1` are for dynamically created devices.
`comedi_release_hardware_device()` is called (via
`comedi_auto_unconfig()`) when a dynamically created device is being
removed. It needs to search the table of minor device numbers to see
which one is associated with the hardware device. It currently starts
the search at minor device number 0. Change it to start from
`comedi_num_legacy_minors` to skip over those belonging to legacy
devices. Also change `comedi_alloc_board_minor()` to skip the legacy
devices when searching for a free minor device number for the
dynamically created device. (The validity of the `hardware_device`
parameter is used to distinguish the legacy devices from the dynamically
created ones.)
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>
