This finishes the port of the irq configuration and handling from
"uio_pdrv_genirq" to "uio_dmem_genirq". It changes the atomic
bit-manipulation routines to their non-atomic counterparts as we are
already guarding the code by spinlock.
Split out from commit 34cb275283 ("UIO: Fix concurrency issue").
Signed-off-by: Rafael Mendonca <rafaelmendsr@gmail.com>
Link: https://lore.kernel.org/r/20220930224100.816175-4-rafaelmendsr@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This fixes a concurrency issue addressed in commit 34cb275283 ("UIO: Fix
concurrency issue"):
"In a SMP case there was a race condition issue between
Uio_pdrv_genirq_irqcontrol() running on one CPU and irq handler on
another CPU. Fix it by spin_locking shared resources access inside irq
handler."
The implementation of "uio_dmem_genirq" was based on "uio_pdrv_genirq" and
it is used in a similar manner to the "uio_pdrv_genirq" driver with respect
to interrupt configuration and handling. At the time "uio_dmem_genirq" was
merged, both had the same implementation of the 'uio_info' handlers
irqcontrol() and handler(), thus, both had the same concurrency issue
mentioned by the above commit. However, the above patch was only applied to
the "uio_pdrv_genirq" driver.
Split out from commit 34cb275283 ("UIO: Fix concurrency issue").
Fixes: 0a0c3b5a24 ("Add new uio device for dynamic memory allocation")
Signed-off-by: Rafael Mendonca <rafaelmendsr@gmail.com>
Link: https://lore.kernel.org/r/20220930224100.816175-3-rafaelmendsr@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Commit b74351287d ("uio: fix a sleep-in-atomic-context bug in
uio_dmem_genirq_irqcontrol()") started calling disable_irq() without
holding the spinlock because it can sleep. However, that fix introduced
another bug: if interrupt is already disabled and a new disable request
comes in, then the spinlock is not unlocked:
root@localhost:~# printf '\x00\x00\x00\x00' > /dev/uio0
root@localhost:~# printf '\x00\x00\x00\x00' > /dev/uio0
root@localhost:~# [ 14.851538] BUG: scheduling while atomic: bash/223/0x00000002
[ 14.851991] Modules linked in: uio_dmem_genirq uio myfpga(OE) bochs drm_vram_helper drm_ttm_helper ttm drm_kms_helper drm snd_pcm ppdev joydev psmouse snd_timer snd e1000fb_sys_fops syscopyarea parport sysfillrect soundcore sysimgblt input_leds pcspkr i2c_piix4 serio_raw floppy evbug qemu_fw_cfg mac_hid pata_acpi ip_tables x_tables autofs4 [last unloaded: parport_pc]
[ 14.854206] CPU: 0 PID: 223 Comm: bash Tainted: G OE 6.0.0-rc7 #21
[ 14.854786] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.16.0-0-gd239552ce722-prebuilt.qemu.org 04/01/2014
[ 14.855664] Call Trace:
[ 14.855861] <TASK>
[ 14.856025] dump_stack_lvl+0x4d/0x67
[ 14.856325] dump_stack+0x14/0x1a
[ 14.856583] __schedule_bug.cold+0x4b/0x5c
[ 14.856915] __schedule+0xe81/0x13d0
[ 14.857199] ? idr_find+0x13/0x20
[ 14.857456] ? get_work_pool+0x2d/0x50
[ 14.857756] ? __flush_work+0x233/0x280
[ 14.858068] ? __schedule+0xa95/0x13d0
[ 14.858307] ? idr_find+0x13/0x20
[ 14.858519] ? get_work_pool+0x2d/0x50
[ 14.858798] schedule+0x6c/0x100
[ 14.859009] schedule_hrtimeout_range_clock+0xff/0x110
[ 14.859335] ? tty_write_room+0x1f/0x30
[ 14.859598] ? n_tty_poll+0x1ec/0x220
[ 14.859830] ? tty_ldisc_deref+0x1a/0x20
[ 14.860090] schedule_hrtimeout_range+0x17/0x20
[ 14.860373] do_select+0x596/0x840
[ 14.860627] ? __kernel_text_address+0x16/0x50
[ 14.860954] ? poll_freewait+0xb0/0xb0
[ 14.861235] ? poll_freewait+0xb0/0xb0
[ 14.861517] ? rpm_resume+0x49d/0x780
[ 14.861798] ? common_interrupt+0x59/0xa0
[ 14.862127] ? asm_common_interrupt+0x2b/0x40
[ 14.862511] ? __uart_start.isra.0+0x61/0x70
[ 14.862902] ? __check_object_size+0x61/0x280
[ 14.863255] core_sys_select+0x1c6/0x400
[ 14.863575] ? vfs_write+0x1c9/0x3d0
[ 14.863853] ? vfs_write+0x1c9/0x3d0
[ 14.864121] ? _copy_from_user+0x45/0x70
[ 14.864526] do_pselect.constprop.0+0xb3/0xf0
[ 14.864893] ? do_syscall_64+0x6d/0x90
[ 14.865228] ? do_syscall_64+0x6d/0x90
[ 14.865556] __x64_sys_pselect6+0x76/0xa0
[ 14.865906] do_syscall_64+0x60/0x90
[ 14.866214] ? syscall_exit_to_user_mode+0x2a/0x50
[ 14.866640] ? do_syscall_64+0x6d/0x90
[ 14.866972] ? do_syscall_64+0x6d/0x90
[ 14.867286] ? do_syscall_64+0x6d/0x90
[ 14.867626] entry_SYSCALL_64_after_hwframe+0x63/0xcd
[...] stripped
[ 14.872959] </TASK>
('myfpga' is a simple 'uio_dmem_genirq' driver I wrote to test this)
The implementation of "uio_dmem_genirq" was based on "uio_pdrv_genirq" and
it is used in a similar manner to the "uio_pdrv_genirq" driver with respect
to interrupt configuration and handling. At the time "uio_dmem_genirq" was
introduced, both had the same implementation of the 'uio_info' handlers
irqcontrol() and handler(). Then commit 34cb275283 ("UIO: Fix concurrency
issue"), which was only applied to "uio_pdrv_genirq", ended up making them
a little different. That commit, among other things, changed disable_irq()
to disable_irq_nosync() in the implementation of irqcontrol(). The
motivation there was to avoid a deadlock between irqcontrol() and
handler(), since it added a spinlock in the irq handler, and disable_irq()
waits for the completion of the irq handler.
By changing disable_irq() to disable_irq_nosync() in irqcontrol(), we also
avoid the sleeping-while-atomic bug that commit b74351287d ("uio: fix a
sleep-in-atomic-context bug in uio_dmem_genirq_irqcontrol()") was trying to
fix. Thus, this fixes the missing unlock in irqcontrol() by importing the
implementation of irqcontrol() handler from the "uio_pdrv_genirq" driver.
In the end, it reverts commit b74351287d ("uio: fix a
sleep-in-atomic-context bug in uio_dmem_genirq_irqcontrol()") and change
disable_irq() to disable_irq_nosync().
It is worth noting that this still does not address the concurrency issue
fixed by commit 34cb275283 ("UIO: Fix concurrency issue"). It will be
addressed separately in the next commits.
Split out from commit 34cb275283 ("UIO: Fix concurrency issue").
Fixes: b74351287d ("uio: fix a sleep-in-atomic-context bug in uio_dmem_genirq_irqcontrol()")
Signed-off-by: Rafael Mendonca <rafaelmendsr@gmail.com>
Link: https://lore.kernel.org/r/20220930224100.816175-2-rafaelmendsr@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The return value of dma_set_coherent_mask() is not always 0.
To catch the exception in case that dma is not support the mask.
Fixes: 0a0c3b5a24 ("Add new uio device for dynamic memory allocation")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Link: https://lore.kernel.org/r/20211204000326.1592687-1-jiasheng@iscas.ac.cn
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This moves move pm_runtime_disable on a devm_add_action_or_reset() handler.
And with the use of the devm_uio_register_device() function, the remove
hook is no longer required.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Link: https://lore.kernel.org/r/20201120075625.12272-2-alexandru.ardelean@analog.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This change converts the simple allocations in the driver to used
device-managed allocation functions.
This removes the error path entirely in the probe function, and reduces
some code in the remove function.
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Link: https://lore.kernel.org/r/20201120075625.12272-1-alexandru.ardelean@analog.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
uio_pdrv_genirq and uio_dmem_genirq interrupts are handled in
userspace. So the condition for the interrupt hasn't normally not been
cleared when top half returns. disable_irq_nosync is called in top half,
but since that normally is lazy the irq isn't actually disabled.
For level triggered interrupts this will always result in a spurious
additional fire since the level in to the interrupt controller still is
active. The actual interrupt handler isn't run though since this
spurious irq is just recorded, and later on discared (for level).
This commit disables lazy masking for level triggered interrupts. It
leaves edge triggered interrupts as before, because they work with the
lazy scheme.
All other UIO drivers already seem to clear the interrupt cause at
driver levels.
Example of double fire. First goes all the way up to
uio_pdrv_genirq_handler, second is terminated in handle_fasteoi_irq and
marked as pending.
<idle>-0 [000] d... 8.245870: gic_handle_irq: irq 29
<idle>-0 [000] d.h. 8.245873: uio_pdrv_genirq_handler: disable irq 29
<idle>-0 [000] d... 8.245878: gic_handle_irq: irq 29
<idle>-0 [000] d.h. 8.245880: handle_fasteoi_irq: irq 29 PENDING
HInt-34 [001] d... 8.245897: uio_pdrv_genirq_irqcontrol: enable irq 29
Tested on 5.7rc2 using uio_pdrv_genirq and a custom Xilinx MPSoC board.
Signed-off-by: Thommy Jakobsson <thommyj@gmail.com>
Link: https://lore.kernel.org/r/20200628141229.16121-1-thommyj@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Fix the following coccicheck warning:
drivers/uio/uio_dmem_genirq.c:47:5-8: Unneeded variable: "ret". Return
"0" on line 71
Signed-off-by: Jason Yan <yanaijie@huawei.com>
Link: https://lore.kernel.org/r/20200428063155.42349-1-yanaijie@huawei.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The driver may sleep while holding a spinlock.
The function call path (from bottom to top) in Linux 4.19 is:
kernel/irq/manage.c, 523:
synchronize_irq in disable_irq
drivers/uio/uio_dmem_genirq.c, 140:
disable_irq in uio_dmem_genirq_irqcontrol
drivers/uio/uio_dmem_genirq.c, 134:
_raw_spin_lock_irqsave in uio_dmem_genirq_irqcontrol
synchronize_irq() can sleep at runtime.
To fix this bug, disable_irq() is called without holding the spinlock.
This bug is found by a static analysis tool STCheck written by myself.
Signed-off-by: Jia-Ju Bai <baijiaju1990@gmail.com>
Link: https://lore.kernel.org/r/20191218094405.6009-1-baijiaju1990@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This change also does a bit of a unification for the IRQ init code.
But the actual problem is that UIO_IRQ_NONE == 0, so for the DT case where
UIO_IRQ_NONE gets assigned to `uioinfo->irq`, a 2nd initialization will get
triggered (for the IRQ) and this one will exit via `goto bad1`.
As far as things seem to go, the only case where UIO_IRQ_NONE seems valid,
is when using a device-tree. The driver has some legacy support for old
platform_data structures. It looks like, for platform_data a non-existent
IRQ is an invalid case (or was considered an invalid case).
Which is why -ENXIO is treated only when a DT is used.
Signed-off-by: Dragos Bogdan <dragos.bogdan@analog.com>
Signed-off-by: Alexandru Ardelean <alexandru.ardelean@analog.com>
Acked-by: Damian Hobson-Garcia <dhobsong@igel.co.jp>
Link: https://lore.kernel.org/r/20191105073212.16719-1-alexandru.ardelean@analog.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
We don't need dev_err() messages when platform_get_irq() fails now that
platform_get_irq() prints an error message itself when something goes
wrong. Let's remove these prints with a simple semantic patch.
// <smpl>
@@
expression ret;
struct platform_device *E;
@@
ret =
(
platform_get_irq(E, ...)
|
platform_get_irq_byname(E, ...)
);
if ( \( ret < 0 \| ret <= 0 \) )
{
(
-if (ret != -EPROBE_DEFER)
-{ ...
-dev_err(...);
-... }
|
...
-dev_err(...);
)
...
}
// </smpl>
While we're here, remove braces on if statements that only have one
statement (manually).
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Stephen Boyd <swboyd@chromium.org>
Link: https://lore.kernel.org/r/20190730181557.90391-46-swboyd@chromium.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Based on 2 normalized pattern(s):
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license version 2 as
published by the free software foundation
this program is free software you can redistribute it and or modify
it under the terms of the gnu general public license version 2 as
published by the free software foundation #
extracted by the scancode license scanner the SPDX license identifier
GPL-2.0-only
has been chosen to replace the boilerplate/reference in 4122 file(s).
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Reviewed-by: Enrico Weigelt <info@metux.net>
Reviewed-by: Kate Stewart <kstewart@linuxfoundation.org>
Reviewed-by: Allison Randal <allison@lohutok.net>
Cc: linux-spdx@vger.kernel.org
Link: https://lkml.kernel.org/r/20190604081206.933168790@linutronix.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
In preparation to remove the node name pointer from struct device_node,
convert printf users to use the %pOFn format specifier.
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Rob Herring <robh@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The variable i contains a total number of resources (including
IORESOURCE_IRQ). However, we want the dmem_region_start to point
after the last resource of type IORESOURCE_MEM. The original behaviour
leads (very likely) to skipping several UIO mapping regions and makes
them useless. Fix this by computing dmem_region_start from the uiomem
which points to the last used UIO mapping.
Fixes: 0a0c3b5a24 ("Add new uio device for dynamic memory allocation")
Signed-off-by: Jan Viktorin <viktorin@rehivetech.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
When platform_get_irq() is failed after "priv" allocated,
it need to free "priv". But the label of bad0 doesn't try
to free about "priv". So this patch changes that lable to "bad1".
But "bad1" has pm_runtime_disable() call, this function should
be called when uio_register_device() is failed. So it is moved
into handling error for uio_register_device().
Signed-off-by: Daeseok Youn <daeseok.youn@gmail.com>
Signed-off-by: Hans J. Koch <hjk@hansjkoch.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Use the wrapper function for retrieving the platform data instead of
accessing dev->platform_data directly. This is a cosmetic change
to make the code simpler and enhance the readability.
Signed-off-by: Jingoo Han <jg1.han@samsung.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This eliminates having an #ifdef returning NULL for the case
when OF is disabled.
Signed-off-by: Sachin Kamat <sachin.kamat@linaro.org>
Acked-by: Damian Hobson-Garcia <dhobsong@igel.co.jp>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The same condition should be used both when allocating and freeing the
driver private data. When dev.of_node is non NULL, allocate a new
private data structure, otherwise use the values from the platform data.
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Damian Hobson-Garcia <dhobsong@igel.co.jp>
Cc: "Hans J. Koch" <hjk@hansjkoch.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The uio device should not fail on open just because one memory allocation
fails. The device might export several regions, the failure of some of
which may or may not be a problem for the user space driver. Failing
regions will remain unmapped, and successful regions will be mapped and
exported to user space. Also deals with the case where failing to map
a region after successfully allocating others would not unmap the
successfully allocated regions before dying.
Signed-off-by: Damian Hobson-Garcia <dhobsong@igel.co.jp>
Cc: "Hans J. Koch" <hjk@hansjkoch.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
DMA_ERROR_CODE is not defined on all architectures and is architecture
specific. Instead, use the constant, ~0 to indicate unmapped regions.
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Reported-by: Geert Uytterhoeven <geert@linux-m68k.org>
Signed-off-by: Damian Hobson-Garcia <dhobsong@igel.co.jp>
Cc: "Hans J. Koch" <hjk@hansjkoch.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Assigning the virtual address returned from dma_alloc_coherent to the the
internal_addr element of uioinfo produces the following sparse errors since
internal_addr is a void __iomem * and dma_alloc_coherent returns void *.
+ drivers/uio/uio_dmem_genirq.c:65:39: sparse: incorrect type in assignment (different address spaces)
drivers/uio/uio_dmem_genirq.c:65:39: expected void [noderef] <asn:2>*internal_addr
drivers/uio/uio_dmem_genirq.c:65:39: got void *[assigned] addr
+ drivers/uio/uio_dmem_genirq.c:93:17: sparse: incorrect type in argument 3 (different address spaces)
drivers/uio/uio_dmem_genirq.c:93:17: expected void *vaddr
drivers/uio/uio_dmem_genirq.c:93:17: got void [noderef] <asn:2>*internal_addr
Store the void * in the driver's private data instead.
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Damian Hobson-Garcia <dhobsong@igel.co.jp>
Cc: "Hans J. Koch" <hjk@hansjkoch.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This device extends the uio_pdrv_genirq driver to provide limited
dynamic memory allocation for UIO devices. This allows UIO devices
to use CMA and IOMMU allocated memory regions. This driver is based
on the uio_pdrv_genirq driver and provides the same generic interrupt
handling capabilities. Like uio_prdv_genirq,
a fixed number of memory regions, defined in the platform device's
.resources field are exported to userpace. This driver adds the ability
to export additional regions whose number and size are known at boot time,
but whose memory is not allocated until the uio device file is opened for
the first time. When the device file is closed, the allocated memory block
is freed. Physical (DMA) addresses for the dynamic regions are provided to
the userspace via /sys/class/uio/uioX/maps/mapY/addr in the same way as
static addresses are when the uio device file is open, when no processes
are holding the device file open, the address returned to userspace is
DMA_ERROR_CODE.
Signed-off-by: Damian Hobson-Garcia <dhobsong@igel.co.jp>
Signed-off-by: "Hans J. Koch" <hjk@hansjkoch.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Rafael Mendonca