The patch titled
Subject: ocfs2: free up write context when direct IO failed
has been added to the -mm tree. Its filename is
ocfs2-free-up-write-context-when-direct-io-failed.patch
This patch should soon appear at
http://ozlabs.org/~akpm/mmots/broken-out/ocfs2-free-up-write-context-when-d…
and later at
http://ozlabs.org/~akpm/mmotm/broken-out/ocfs2-free-up-write-context-when-d…
Before you just go and hit "reply", please:
a) Consider who else should be cc'ed
b) Prefer to cc a suitable mailing list as well
c) Ideally: find the original patch on the mailing list and do a
reply-to-all to that, adding suitable additional cc's
*** Remember to use Documentation/process/submit-checklist.rst when testing your code ***
The -mm tree is included into linux-next and is updated
there every 3-4 working days
------------------------------------------------------
From: Wengang Wang <wen.gang.wang(a)oracle.com>
Subject: ocfs2: free up write context when direct IO failed
The write context should also be freed even when direct IO failed.
Otherwise a memory leak is introduced and entries remain in
oi->ip_unwritten_list causing the following BUG later in unlink path:
ERROR: bug expression: !list_empty(&oi->ip_unwritten_list)
ERROR: Clear inode of 215043, inode has unwritten extents
...
Call Trace:
? __set_current_blocked+0x42/0x68
ocfs2_evict_inode+0x91/0x6a0 [ocfs2]
? bit_waitqueue+0x40/0x33
evict+0xdb/0x1af
iput+0x1a2/0x1f7
do_unlinkat+0x194/0x28f
SyS_unlinkat+0x1b/0x2f
do_syscall_64+0x79/0x1ae
entry_SYSCALL_64_after_hwframe+0x151/0x0
This patch also logs, with frequency limit, direct IO failures.
Link: http://lkml.kernel.org/r/20181102170632.25921-1-wen.gang.wang@oracle.com
Signed-off-by: Wengang Wang <wen.gang.wang(a)oracle.com>
Reviewed-by: Junxiao Bi <junxiao.bi(a)oracle.com>
Reviewed-by: Changwei Ge <ge.changwei(a)h3c.com>
Reviewed-by: Joseph Qi <jiangqi903(a)gmail.com>
Cc: Mark Fasheh <mark(a)fasheh.com>
Cc: Joel Becker <jlbec(a)evilplan.org>
Cc: <stable(a)vger.kernel.org>
Signed-off-by: Andrew Morton <akpm(a)linux-foundation.org>
---
fs/ocfs2/aops.c | 12 ++++++++++--
fs/ocfs2/cluster/masklog.h | 9 +++++++++
2 files changed, 19 insertions(+), 2 deletions(-)
--- a/fs/ocfs2/aops.c~ocfs2-free-up-write-context-when-direct-io-failed
+++ a/fs/ocfs2/aops.c
@@ -2411,8 +2411,16 @@ static int ocfs2_dio_end_io(struct kiocb
/* this io's submitter should not have unlocked this before we could */
BUG_ON(!ocfs2_iocb_is_rw_locked(iocb));
- if (bytes > 0 && private)
- ret = ocfs2_dio_end_io_write(inode, private, offset, bytes);
+ if (bytes <= 0)
+ mlog_ratelimited(ML_ERROR, "Direct IO failed, bytes = %lld",
+ (long long)bytes);
+ if (private) {
+ if (bytes > 0)
+ ret = ocfs2_dio_end_io_write(inode, private, offset,
+ bytes);
+ else
+ ocfs2_dio_free_write_ctx(inode, private);
+ }
ocfs2_iocb_clear_rw_locked(iocb);
--- a/fs/ocfs2/cluster/masklog.h~ocfs2-free-up-write-context-when-direct-io-failed
+++ a/fs/ocfs2/cluster/masklog.h
@@ -178,6 +178,15 @@ do { \
##__VA_ARGS__); \
} while (0)
+#define mlog_ratelimited(mask, fmt, ...) \
+do { \
+ static DEFINE_RATELIMIT_STATE(_rs, \
+ DEFAULT_RATELIMIT_INTERVAL, \
+ DEFAULT_RATELIMIT_BURST); \
+ if (__ratelimit(&_rs)) \
+ mlog(mask, fmt, ##__VA_ARGS__); \
+} while (0)
+
#define mlog_errno(st) ({ \
int _st = (st); \
if (_st != -ERESTARTSYS && _st != -EINTR && \
_
Patches currently in -mm which might be from wen.gang.wang(a)oracle.com are
ocfs2-free-up-write-context-when-direct-io-failed.patch
Commit '88ba95bedb79 ("backlight: pwm_bl: Compute brightness of LED
linearly to human eye")' allows the possibility to compute a default
brightness table when there isn't the brightness-levels property in the
DT. Unfortunately the changes made broke the pwm backlight for the
non-DT boards.
Usually, the non-DT boards don't pass the brightness levels via platform
data, instead, it sets the max_brightness in their platform data and the
driver calculates the level without a table. The ofending patch assumed
hat when there is no brightness levels table we should create one, but this
is clearly wrong for the non-DT case.
After this patch the code handles the DT and the non-DT case taking in
consideration also if max_brightness is set or not.
Fixes: '88ba95bedb79 ("backlight: pwm_bl: Compute brightness of LED linearly to human eye")'
Cc: <stable(a)vger.kernel.org>
Reported-by: Robert Jarzmik <robert.jarzmik(a)free.fr>
Signed-off-by: Enric Balletbo i Serra <enric.balletbo(a)collabora.com>
Tested-by: Robert Jarzmik <robert.jarzmik(a)free.fr>
Acked-by: Daniel Thompson <daniel.thompson(a)linaro.org>
---
Changes in v2:
- Rebase on top of mainline
- Add Tested-by and Acked-by tags.
drivers/video/backlight/pwm_bl.c | 41 +++++++++++++++++++++++++++-----
1 file changed, 35 insertions(+), 6 deletions(-)
diff --git a/drivers/video/backlight/pwm_bl.c b/drivers/video/backlight/pwm_bl.c
index 678b27063198..f9ef0673a083 100644
--- a/drivers/video/backlight/pwm_bl.c
+++ b/drivers/video/backlight/pwm_bl.c
@@ -562,7 +562,30 @@ static int pwm_backlight_probe(struct platform_device *pdev)
goto err_alloc;
}
- if (!data->levels) {
+ if (data->levels) {
+ /*
+ * For the DT case, only when brightness levels is defined
+ * data->levels is filled. For the non-DT case, data->levels
+ * can come from platform data, however is not usual.
+ */
+ for (i = 0; i <= data->max_brightness; i++) {
+ if (data->levels[i] > pb->scale)
+ pb->scale = data->levels[i];
+
+ pb->levels = data->levels;
+ }
+ } else if (!data->max_brightness) {
+ /*
+ * If no brightness levels are provided and max_brightness is
+ * not set, use the default brightness table. For the DT case,
+ * max_brightness is set to 0 when brightness levels is not
+ * specified. For the non-DT case, max_brightness is usually
+ * set to some value.
+ */
+
+ /* Get the PWM period (in nanoseconds) */
+ pwm_get_state(pb->pwm, &state);
+
ret = pwm_backlight_brightness_default(&pdev->dev, data,
state.period);
if (ret < 0) {
@@ -570,13 +593,19 @@ static int pwm_backlight_probe(struct platform_device *pdev)
"failed to setup default brightness table\n");
goto err_alloc;
}
- }
- for (i = 0; i <= data->max_brightness; i++) {
- if (data->levels[i] > pb->scale)
- pb->scale = data->levels[i];
+ for (i = 0; i <= data->max_brightness; i++) {
+ if (data->levels[i] > pb->scale)
+ pb->scale = data->levels[i];
- pb->levels = data->levels;
+ pb->levels = data->levels;
+ }
+ } else {
+ /*
+ * That only happens for the non-DT case, where platform data
+ * sets the max_brightness value.
+ */
+ pb->scale = data->max_brightness;
}
pb->lth_brightness = data->lth_brightness * (state.period / pb->scale);
--
2.19.1
Hello,
on an mxs machine running a v4.9-rt kernel I saw the following
backtrace:
[ 25.309794] BUG: sleeping function called from invalid context at kernel/locking/rtmutex.c:987
[ 25.309810] in_atomic(): 1, irqs_disabled(): 128, pid: 221, name: mrp62439d
[ 25.309834] CPU: 0 PID: 221 Comm: mrp62439d Not tainted 4.9.109-20180709-1-rt76-00043-g80a57dee2c6b-dirty #43
[ 25.309840] Hardware name: Freescale MXS (Device Tree)
[ 25.309913] [<c000fad8>] (unwind_backtrace) from [<c000df34>] (show_stack+0x18/0x1c)
[ 25.309957] [<c000df34>] (show_stack) from [<c0042924>] (___might_sleep+0xf8/0x164)
[ 25.309996] [<c0042924>] (___might_sleep) from [<c04ec920>] (rt_spin_lock+0x20/0x6c)
[ 25.310039] [<c04ec920>] (rt_spin_lock) from [<c02e5a20>] (gpio_to_desc+0x1c/0xbc)
[ 25.310076] [<c02e5a20>] (gpio_to_desc) from [<c02ec358>] (mxs_gpio_set_irq_type+0x114/0x194)
[ 25.310107] [<c02ec358>] (mxs_gpio_set_irq_type) from [<c00558d8>] (__irq_set_trigger+0x64/0x158)
[ 25.310132] [<c00558d8>] (__irq_set_trigger) from [<c0055f74>] (__setup_irq+0x5a8/0x694)
[ 25.310156] [<c0055f74>] (__setup_irq) from [<c0056408>] (request_threaded_irq+0xe4/0x18c)
[ 25.310187] [<c0056408>] (request_threaded_irq) from [<c02e8650>] (gpio_ioctl+0x304/0x680)
[ 25.310217] [<c02e8650>] (gpio_ioctl) from [<c01476a4>] (do_vfs_ioctl+0x98/0x970)
[ 25.310241] [<c01476a4>] (do_vfs_ioctl) from [<c0147fb8>] (SyS_ioctl+0x3c/0x64)
[ 25.310264] [<c0147fb8>] (SyS_ioctl) from [<c000a4d8>] (__sys_trace_return+0x0/0x10)
The problem is that __irq_set_trigger is called in atomic context (even on rt)
and mxs_gpio_set_irq_type calls gpio_get_value which takes the
(sleeping on rt) gpio_lock.
I didn't test 833eacc7b5913da9896bacd30db7d490aa777868 yet (I used
val = readl(port->base + PINCTRL_DIN(port)) & pin_mask;
instead) but this commit should have the same effect.
Can you please backport this? (There are maybe some more similar
problems that were introduced in v4.5-rc1 / 0f4630f3720e7e, I didn't
check these though.)
Best regards
Uwe
--
Pengutronix e.K. | Uwe Kleine-König |
Industrial Linux Solutions | http://www.pengutronix.de/ |
From: Huacai Chen <chenhc(a)lemote.com>
[ Upstream commit c61c7def1fa0a722610d89790e0255b74f3c07dd ]
Commit ea7e0480a4b6 ("MIPS: VDSO: Always map near top of user memory")
set VDSO_RANDOMIZE_SIZE to 256MB for 64bit kernel. But take a look at
arch/mips/mm/mmap.c we can see that MIN_GAP is 128MB, which means the
mmap_base may be at (user_address_top - 128MB). This make the stack be
surrounded by mmaped areas, then stack expanding fails and causes a
segmentation fault. Therefore, VDSO_RANDOMIZE_SIZE should be less than
MIN_GAP and this patch reduce it to 64MB.
Signed-off-by: Huacai Chen <chenhc(a)lemote.com>
Signed-off-by: Paul Burton <paul.burton(a)mips.com>
Fixes: ea7e0480a4b6 ("MIPS: VDSO: Always map near top of user memory")
Patchwork: https://patchwork.linux-mips.org/patch/20910/
Cc: Ralf Baechle <ralf(a)linux-mips.org>
Cc: James Hogan <jhogan(a)kernel.org>
Cc: linux-mips(a)linux-mips.org
Cc: Fuxin Zhang <zhangfx(a)lemote.com>
Cc: Zhangjin Wu <wuzhangjin(a)gmail.com>
Cc: Huacai Chen <chenhuacai(a)gmail.com>
Cc: stable(a)vger.kernel.org # 4.19
---
arch/mips/include/asm/processor.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/arch/mips/include/asm/processor.h b/arch/mips/include/asm/processor.h
index 49d6046ca1d0..c373eb605040 100644
--- a/arch/mips/include/asm/processor.h
+++ b/arch/mips/include/asm/processor.h
@@ -81,7 +81,7 @@ extern unsigned int vced_count, vcei_count;
#endif
-#define VDSO_RANDOMIZE_SIZE (TASK_IS_32BIT_ADDR ? SZ_1M : SZ_256M)
+#define VDSO_RANDOMIZE_SIZE (TASK_IS_32BIT_ADDR ? SZ_1M : SZ_64M)
extern unsigned long mips_stack_top(void);
#define STACK_TOP mips_stack_top()
--
2.19.1
On Tue, Nov 6, 2018 at 3:06 PM Petr Vorel <pvorel(a)suse.cz> wrote:
>
> Hi Amir,
>
> > There must be some confusion.
> > FAN_MARK_MOUNT was NOT added in v4.19-rc2.
> > It has been there from the start.
> > FAN_MARK_INODE was NOT added either
> > the define FAN_MARK_INODE 0 is just a convenience readability define
> > it does not change the API.
> I'm sorry, you're right.
>
> > You may be confusing with FAN_MARK_FILESYSTEM
> > just was just added in kernel v4.20-rc1.
> > The extension of tests to cover FAN_MARK_FILESYSTEM
> > is waiting in my queue:
> > https://github.com/amir73il/ltp/commits/fanotify_sb
>
> > And it already includes runtime checks for FAN_MARK_FILESYSTEM
> > support.
>
> > Did I miss anything?
> Testing your branch on older kernel, fanotify10 fails earlier than new TCONF
Because fanotify10 checks for a bug that existed since the beginning
and fixed by 9bdda4e9cf2d fsnotify: fix ignore mask logic in fsnotify().
So test SHOULD fail until the backported patch is applied to the old kernel.
The patch does not apply cleanly to kernels <= v4.17.
Tested backport patch for v4.14.y attached.
Thanks,
Amir.
When the platform BIOS is unable to report all the media error records
it requires the OS to restart the scrub at a prescribed location. The
driver detects the overflow condition, but then fails to report it to
the ARS state machine after reaping the records. Propagate -ENOSPC
correctly to continue the ARS operation.
Cc: <stable(a)vger.kernel.org>
Fixes: 1cf03c00e7c1 ("nfit: scrub and register regions in a workqueue")
Reported-by: Jacek Zloch <jacek.zloch(a)intel.com>
Signed-off-by: Dan Williams <dan.j.williams(a)intel.com>
---
drivers/acpi/nfit/core.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)
diff --git a/drivers/acpi/nfit/core.c b/drivers/acpi/nfit/core.c
index f8c638f3c946..5970b8f5f768 100644
--- a/drivers/acpi/nfit/core.c
+++ b/drivers/acpi/nfit/core.c
@@ -2928,9 +2928,9 @@ static int acpi_nfit_query_poison(struct acpi_nfit_desc *acpi_desc)
return rc;
if (ars_status_process_records(acpi_desc))
- return -ENOMEM;
+ dev_err(acpi_desc->dev, "Failed to process ARS records\n");
- return 0;
+ return rc;
}
static int ars_register(struct acpi_nfit_desc *acpi_desc,
From: Masahisa Kojima <masahisa.kojima(a)linaro.org>
[ Upstream commit 8d5b0bf611ec5b7618d5b772dddc93b8afa78cb8 ]
We observed that packets and bytes count are not reset
when user performs interface down. Eventually, tx queue is
exhausted and packets will not be sent out.
To avoid this problem, resets tx queue in ndo_stop.
Fixes: 533dd11a12f6 ("net: socionext: Add Synquacer NetSec driver")
Signed-off-by: Masahisa Kojima <masahisa.kojima(a)linaro.org>
Signed-off-by: Yoshitoyo Osaki <osaki.yoshitoyo(a)socionext.com>
Signed-off-by: David S. Miller <davem(a)davemloft.net>
Signed-off-by: Sasha Levin <sashal(a)kernel.org>
---
drivers/net/ethernet/socionext/netsec.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/drivers/net/ethernet/socionext/netsec.c b/drivers/net/ethernet/socionext/netsec.c
index 4289ccb26e4e..d2caeb9edc04 100644
--- a/drivers/net/ethernet/socionext/netsec.c
+++ b/drivers/net/ethernet/socionext/netsec.c
@@ -940,6 +940,9 @@ static void netsec_uninit_pkt_dring(struct netsec_priv *priv, int id)
dring->head = 0;
dring->tail = 0;
dring->pkt_cnt = 0;
+
+ if (id == NETSEC_RING_TX)
+ netdev_reset_queue(priv->ndev);
}
static void netsec_free_dring(struct netsec_priv *priv, int id)
--
2.17.1