This is an automatic generated email to let you know that the following patch were queued:
Subject: media: rc: sir_ir: detect presence of port
Author: Sean Young <sean(a)mess.org>
Date: Wed Nov 8 16:19:45 2017 -0500
Without this test, sir_ir clumsy claims resources for a device which
does not exist.
The 0-day kernel test robot reports the following errors (in a loop):
sir_ir sir_ir.0: Trapped in interrupt
genirq: Flags mismatch irq 4. 00000000 (ttyS0) vs. 00000000 (sir_ir)
When sir_ir is loaded with the default io and irq, the following happens:
- sir_ir claims irq 4
- user space opens /dev/ttyS0
- in serial8250_do_startup(), some setup is done for ttyS0, which causes
irq 4 to fire (in THRE test)
- sir_ir does not realise it was not for it, and spins until the "trapped
in interrupt"
- now serial driver calls setup_irq() and fails and we get the
"Flags mismatch" error.
There is no port present at 0x3e8 so simply check for the presence of a
port, as suggested by Linus.
Reported-by: kbuild test robot <fengguang.wu(a)intel.com>
Tested-by: Fengguang Wu <fengguang.wu(a)intel.com>
Signed-off-by: Sean Young <sean(a)mess.org>
Cc: <stable(a)vger.kernel.org> # 4.12+
Signed-off-by: Mauro Carvalho Chehab <mchehab(a)s-opensource.com>
drivers/media/rc/sir_ir.c | 40 ++++++++++++++++++++++++++++++++++++----
1 file changed, 36 insertions(+), 4 deletions(-)
---
diff --git a/drivers/media/rc/sir_ir.c b/drivers/media/rc/sir_ir.c
index 76120664b700..9ee2c9196b4d 100644
--- a/drivers/media/rc/sir_ir.c
+++ b/drivers/media/rc/sir_ir.c
@@ -57,7 +57,7 @@ static void add_read_queue(int flag, unsigned long val);
static irqreturn_t sir_interrupt(int irq, void *dev_id);
static void send_space(unsigned long len);
static void send_pulse(unsigned long len);
-static void init_hardware(void);
+static int init_hardware(void);
static void drop_hardware(void);
/* Initialisation */
@@ -263,11 +263,36 @@ static void send_pulse(unsigned long len)
}
}
-static void init_hardware(void)
+static int init_hardware(void)
{
+ u8 scratch, scratch2, scratch3;
unsigned long flags;
spin_lock_irqsave(&hardware_lock, flags);
+
+ /*
+ * This is a simple port existence test, borrowed from the autoconfig
+ * function in drivers/tty/serial/8250/8250_port.c
+ */
+ scratch = sinp(UART_IER);
+ soutp(UART_IER, 0);
+#ifdef __i386__
+ outb(0xff, 0x080);
+#endif
+ scratch2 = sinp(UART_IER) & 0x0f;
+ soutp(UART_IER, 0x0f);
+#ifdef __i386__
+ outb(0x00, 0x080);
+#endif
+ scratch3 = sinp(UART_IER) & 0x0f;
+ soutp(UART_IER, scratch);
+ if (scratch2 != 0 || scratch3 != 0x0f) {
+ /* we fail, there's nothing here */
+ spin_unlock_irqrestore(&hardware_lock, flags);
+ pr_err("port existence test failed, cannot continue\n");
+ return -ENODEV;
+ }
+
/* reset UART */
outb(0, io + UART_MCR);
outb(0, io + UART_IER);
@@ -285,6 +310,8 @@ static void init_hardware(void)
/* turn on UART */
outb(UART_MCR_DTR | UART_MCR_RTS | UART_MCR_OUT2, io + UART_MCR);
spin_unlock_irqrestore(&hardware_lock, flags);
+
+ return 0;
}
static void drop_hardware(void)
@@ -334,14 +361,19 @@ static int sir_ir_probe(struct platform_device *dev)
pr_err("IRQ %d already in use.\n", irq);
return retval;
}
+
+ retval = init_hardware();
+ if (retval) {
+ del_timer_sync(&timerlist);
+ return retval;
+ }
+
pr_info("I/O port 0x%.4x, IRQ %d.\n", io, irq);
retval = devm_rc_register_device(&sir_ir_dev->dev, rcdev);
if (retval < 0)
return retval;
- init_hardware();
-
return 0;
}
Hi Mark,
On 15-11-2017 10:54, Mark Brown wrote:
> On Wed, Nov 15, 2017 at 02:45:45AM +0000, alexander.levin(a)verizon.com wrote:
>
>> We can no longer rely on the return value of
>> devm_snd_dmaengine_pcm_register(...) to check if the DMA
>> handle is declared in the DT.
>> Previously this check activated PIO mode but currently
>> dma_request_chan returns either a valid channel or -EPROBE_DEFER.
> So when did the corresponding change to the dmaengine API go in?
Looks like this was introduced in a8135d0d79e9 ("dmaengine: core:
Introduce new, universal API to request a channel") but looking
at soc-generic-dmaengine-pcm.c I see that this would never work
anyway because dmaengine_pcm_request_chan_of() also returns
either zero or -EPROBE_DEFER.
>
>> In order to activate PIO mode check instead if the interrupt
>> line is declared. This reflects better what is documented in
>> the DT bindings (see Documentation/devicetree/bindings/sound/
>> designware-i2s.txt).
>> Also, initialize use_pio variable which was never being set
>> causing PIO mode to never work.
> Though if PIO mode never worked presumably this isn't that urgent...
Yeah, and we also have this. Its my fault, I was using a
different tree for testing at the time. But if we could backport
this it would be nice because we and our clients use this PIO
driver in ARC AXS101 Development Platform. (A side note is that
the DT bindings were only recently introduced but I think its
more easier to add the bindings instead of fixing this driver ...).
Best Regards,
Jose Miguel Abreu
The snd_usb_copy_string_desc() retrieves the usb string corresponding to
the index number thought the usb_string(). And for NULL-terminated, insert
'0' by using the return value of usb_string() as the index of buffer to
hold the string.
The problem is that the usb_string() also returns the length of the string
read(>= 0), but it can also return a negative value, the error or status
value of usb_control_msg(). If iClockSource is '0' as shown below,
usb_string() will return -EINVAL. This will result in '0' being inserted
into buf[-22], and the following KASAN out-of-bound error message will be
output.
AudioControl Interface Descriptor:
bLength 8
bDescriptorType 36
bDescriptorSubtype 10 (CLOCK_SOURCE)
bClockID 1
bmAttributes 0x07 Internal programmable Clock (synced to SOF)
bmControls 0x07
Clock Frequency Control (read/write)
Clock Validity Control (read-only)
bAssocTerminal 0
iClockSource 0
To fix out-of-bound error, insert 0 only if the return value of
usb_string() is greater than 0.
==================================================================
BUG: KASAN: stack-out-of-bounds in parse_audio_unit+0x1327/0x1960 [snd_usb_audio]
Write of size 1 at addr ffff88007e66735a by task systemd-udevd/18376
CPU: 0 PID: 18376 Comm: systemd-udevd Not tainted 4.13.0+ #3
Hardware name: LG Electronics 15N540-RFLGL/White Tip Mountain, BIOS 15N5
Call Trace:
dump_stack+0x63/0x8d
print_address_description+0x70/0x290
? parse_audio_unit+0x1327/0x1960 [snd_usb_audio]
kasan_report+0x265/0x350
__asan_store1+0x4a/0x50
parse_audio_unit+0x1327/0x1960 [snd_usb_audio]
? save_stack+0xb5/0xd0
? save_stack_trace+0x1b/0x20
? save_stack+0x46/0xd0
? kasan_kmalloc+0xad/0xe0
? kmem_cache_alloc_trace+0xff/0x230
? snd_usb_create_mixer+0xb0/0x4b0 [snd_usb_audio]
? usb_audio_probe+0x4de/0xf40 [snd_usb_audio]
? usb_probe_interface+0x1f5/0x440
? driver_probe_device+0x3ed/0x660
? build_feature_ctl+0xb10/0xb10 [snd_usb_audio]
? save_stack_trace+0x1b/0x20
? init_object+0x69/0xa0
? snd_usb_find_csint_desc+0xa8/0xf0 [snd_usb_audio]
snd_usb_mixer_controls+0x1dc/0x370 [snd_usb_audio]
? build_audio_procunit+0x890/0x890 [snd_usb_audio]
? snd_usb_create_mixer+0xb0/0x4b0 [snd_usb_audio]
? kmem_cache_alloc_trace+0xff/0x230
? usb_ifnum_to_if+0xbd/0xf0
snd_usb_create_mixer+0x25b/0x4b0 [snd_usb_audio]
? snd_usb_create_stream+0x255/0x2c0 [snd_usb_audio]
usb_audio_probe+0x4de/0xf40 [snd_usb_audio]
? snd_usb_autosuspend.part.7+0x30/0x30 [snd_usb_audio]
? __pm_runtime_idle+0x90/0x90
? kernfs_activate+0xa6/0xc0
? usb_match_one_id_intf+0xdc/0x130
? __pm_runtime_set_status+0x2d4/0x450
usb_probe_interface+0x1f5/0x440
Cc: <stable(a)vger.kernel.org>
Signed-off-by: Jaejoong Kim <climbbb.kim(a)gmail.com>
---
The AudioControl Interface Descriptor in commit message is from
lsusb output with real usb audio DAC.
The usb audio product causing the OOB are as follows:
http://www.lg.com/uk/lg-friends/lg-AFD-1200
It only prints OOB error and usb audio works well. :)
sound/usb/mixer.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/sound/usb/mixer.c b/sound/usb/mixer.c
index e630813..5a83c2c 100644
--- a/sound/usb/mixer.c
+++ b/sound/usb/mixer.c
@@ -204,7 +204,8 @@ static int snd_usb_copy_string_desc(struct mixer_build *state,
int index, char *buf, int maxlen)
{
int len = usb_string(state->chip->dev, index, buf, maxlen - 1);
- buf[len] = 0;
+ if (len > 0)
+ buf[len] = 0;
return len;
}
--
2.7.4
From: Bob Moore <robert.moore(a)intel.com>
[ Upstream commit 57707a9a7780fab426b8ae9b4c7b65b912a748b3 ]
ACPICA commit 9f76de2d249b18804e35fb55d14b1c2604d627a1
ACPICA commit b2e89d72ef1e9deefd63c3fd1dee90f893575b3a
ACPICA commit 23b5bbe6d78afd3c5abf3adb91a1b098a3000b2e
The declared buffer length must be the same as the length of the
byte initializer list, otherwise not a valid resource descriptor.
Link: https://github.com/acpica/acpica/commit/9f76de2d
Link: https://github.com/acpica/acpica/commit/b2e89d72
Link: https://github.com/acpica/acpica/commit/23b5bbe6
Signed-off-by: Bob Moore <robert.moore(a)intel.com>
Signed-off-by: Lv Zheng <lv.zheng(a)intel.com>
Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki(a)intel.com>
Signed-off-by: Sasha Levin <alexander.levin(a)verizon.com>
---
drivers/acpi/acpica/utresrc.c | 17 ++++++++++++-----
1 file changed, 12 insertions(+), 5 deletions(-)
diff --git a/drivers/acpi/acpica/utresrc.c b/drivers/acpi/acpica/utresrc.c
index 5cd017c7ac0e..94b3ee013761 100644
--- a/drivers/acpi/acpica/utresrc.c
+++ b/drivers/acpi/acpica/utresrc.c
@@ -421,8 +421,10 @@ acpi_ut_walk_aml_resources(struct acpi_walk_state *walk_state,
ACPI_FUNCTION_TRACE(ut_walk_aml_resources);
- /* The absolute minimum resource template is one end_tag descriptor */
-
+ /*
+ * The absolute minimum resource template is one end_tag descriptor.
+ * However, we will treat a lone end_tag as just a simple buffer.
+ */
if (aml_length < sizeof(struct aml_resource_end_tag)) {
return_ACPI_STATUS(AE_AML_NO_RESOURCE_END_TAG);
}
@@ -454,9 +456,8 @@ acpi_ut_walk_aml_resources(struct acpi_walk_state *walk_state,
/* Invoke the user function */
if (user_function) {
- status =
- user_function(aml, length, offset, resource_index,
- context);
+ status = user_function(aml, length, offset,
+ resource_index, context);
if (ACPI_FAILURE(status)) {
return_ACPI_STATUS(status);
}
@@ -480,6 +481,12 @@ acpi_ut_walk_aml_resources(struct acpi_walk_state *walk_state,
*context = aml;
}
+ /* Check if buffer is defined to be longer than the resource length */
+
+ if (aml_length > (offset + length)) {
+ return_ACPI_STATUS(AE_AML_NO_RESOURCE_END_TAG);
+ }
+
/* Normal exit */
return_ACPI_STATUS(AE_OK);
--
2.11.0