The patch below does not apply to the 4.19-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable(a)vger.kernel.org>.
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
>From dcb10920179ab74caf88a6f2afadecfc2743b910 Mon Sep 17 00:00:00 2001
From: Fabrice Gasnier <fabrice.gasnier(a)st.com>
Date: Tue, 17 Sep 2019 14:38:16 +0200
Subject: [PATCH] iio: adc: stm32-adc: fix a race when using several adcs with
dma and irq
End of conversion may be handled by using IRQ or DMA. There may be a
race when two conversions complete at the same time on several ADCs.
EOC can be read as 'set' for several ADCs, with:
- an ADC configured to use IRQs. EOCIE bit is set. The handler is normally
called in this case.
- an ADC configured to use DMA. EOCIE bit isn't set. EOC triggers the DMA
request instead. It's then automatically cleared by DMA read. But the
handler gets called due to status bit is temporarily set (IRQ triggered
by the other ADC).
So both EOC status bit in CSR and EOCIE control bit must be checked
before invoking the interrupt handler (e.g. call ISR only for
IRQ-enabled ADCs).
Fixes: 2763ea0585c9 ("iio: adc: stm32: add optional dma support")
Signed-off-by: Fabrice Gasnier <fabrice.gasnier(a)st.com>
Cc: <Stable(a)vger.kernel.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron(a)huawei.com>
diff --git a/drivers/iio/adc/stm32-adc-core.c b/drivers/iio/adc/stm32-adc-core.c
index 84ac326bb714..93a096a91f8c 100644
--- a/drivers/iio/adc/stm32-adc-core.c
+++ b/drivers/iio/adc/stm32-adc-core.c
@@ -44,6 +44,8 @@
* @eoc1: adc1 end of conversion flag in @csr
* @eoc2: adc2 end of conversion flag in @csr
* @eoc3: adc3 end of conversion flag in @csr
+ * @ier: interrupt enable register offset for each adc
+ * @eocie_msk: end of conversion interrupt enable mask in @ier
*/
struct stm32_adc_common_regs {
u32 csr;
@@ -51,6 +53,8 @@ struct stm32_adc_common_regs {
u32 eoc1_msk;
u32 eoc2_msk;
u32 eoc3_msk;
+ u32 ier;
+ u32 eocie_msk;
};
struct stm32_adc_priv;
@@ -276,6 +280,8 @@ static const struct stm32_adc_common_regs stm32f4_adc_common_regs = {
.eoc1_msk = STM32F4_EOC1,
.eoc2_msk = STM32F4_EOC2,
.eoc3_msk = STM32F4_EOC3,
+ .ier = STM32F4_ADC_CR1,
+ .eocie_msk = STM32F4_EOCIE,
};
/* STM32H7 common registers definitions */
@@ -284,8 +290,24 @@ static const struct stm32_adc_common_regs stm32h7_adc_common_regs = {
.ccr = STM32H7_ADC_CCR,
.eoc1_msk = STM32H7_EOC_MST,
.eoc2_msk = STM32H7_EOC_SLV,
+ .ier = STM32H7_ADC_IER,
+ .eocie_msk = STM32H7_EOCIE,
};
+static const unsigned int stm32_adc_offset[STM32_ADC_MAX_ADCS] = {
+ 0, STM32_ADC_OFFSET, STM32_ADC_OFFSET * 2,
+};
+
+static unsigned int stm32_adc_eoc_enabled(struct stm32_adc_priv *priv,
+ unsigned int adc)
+{
+ u32 ier, offset = stm32_adc_offset[adc];
+
+ ier = readl_relaxed(priv->common.base + offset + priv->cfg->regs->ier);
+
+ return ier & priv->cfg->regs->eocie_msk;
+}
+
/* ADC common interrupt for all instances */
static void stm32_adc_irq_handler(struct irq_desc *desc)
{
@@ -296,13 +318,28 @@ static void stm32_adc_irq_handler(struct irq_desc *desc)
chained_irq_enter(chip, desc);
status = readl_relaxed(priv->common.base + priv->cfg->regs->csr);
- if (status & priv->cfg->regs->eoc1_msk)
+ /*
+ * End of conversion may be handled by using IRQ or DMA. There may be a
+ * race here when two conversions complete at the same time on several
+ * ADCs. EOC may be read 'set' for several ADCs, with:
+ * - an ADC configured to use DMA (EOC triggers the DMA request, and
+ * is then automatically cleared by DR read in hardware)
+ * - an ADC configured to use IRQs (EOCIE bit is set. The handler must
+ * be called in this case)
+ * So both EOC status bit in CSR and EOCIE control bit must be checked
+ * before invoking the interrupt handler (e.g. call ISR only for
+ * IRQ-enabled ADCs).
+ */
+ if (status & priv->cfg->regs->eoc1_msk &&
+ stm32_adc_eoc_enabled(priv, 0))
generic_handle_irq(irq_find_mapping(priv->domain, 0));
- if (status & priv->cfg->regs->eoc2_msk)
+ if (status & priv->cfg->regs->eoc2_msk &&
+ stm32_adc_eoc_enabled(priv, 1))
generic_handle_irq(irq_find_mapping(priv->domain, 1));
- if (status & priv->cfg->regs->eoc3_msk)
+ if (status & priv->cfg->regs->eoc3_msk &&
+ stm32_adc_eoc_enabled(priv, 2))
generic_handle_irq(irq_find_mapping(priv->domain, 2));
chained_irq_exit(chip, desc);
diff --git a/drivers/iio/adc/stm32-adc-core.h b/drivers/iio/adc/stm32-adc-core.h
index 94aa2d2577dc..2579d514c2a3 100644
--- a/drivers/iio/adc/stm32-adc-core.h
+++ b/drivers/iio/adc/stm32-adc-core.h
@@ -25,6 +25,7 @@
* --------------------------------------------------------
*/
#define STM32_ADC_MAX_ADCS 3
+#define STM32_ADC_OFFSET 0x100
#define STM32_ADCX_COMN_OFFSET 0x300
/* STM32F4 - Registers for each ADC instance */
The patch below does not apply to the 4.19-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable(a)vger.kernel.org>.
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
>From 11a19a90870ea5496a8ded69b86f5b476b6d3355 Mon Sep 17 00:00:00 2001
From: Josef Bacik <josef(a)toxicpanda.com>
Date: Mon, 9 Sep 2019 10:12:04 -0400
Subject: [PATCH] btrfs: allocate new inode in NOFS context
A user reported a lockdep splat
======================================================
WARNING: possible circular locking dependency detected
5.2.11-gentoo #2 Not tainted
------------------------------------------------------
kswapd0/711 is trying to acquire lock:
000000007777a663 (sb_internal){.+.+}, at: start_transaction+0x3a8/0x500
but task is already holding lock:
000000000ba86300 (fs_reclaim){+.+.}, at: __fs_reclaim_acquire+0x0/0x30
which lock already depends on the new lock.
the existing dependency chain (in reverse order) is:
-> #1 (fs_reclaim){+.+.}:
kmem_cache_alloc+0x1f/0x1c0
btrfs_alloc_inode+0x1f/0x260
alloc_inode+0x16/0xa0
new_inode+0xe/0xb0
btrfs_new_inode+0x70/0x610
btrfs_symlink+0xd0/0x420
vfs_symlink+0x9c/0x100
do_symlinkat+0x66/0xe0
do_syscall_64+0x55/0x1c0
entry_SYSCALL_64_after_hwframe+0x49/0xbe
-> #0 (sb_internal){.+.+}:
__sb_start_write+0xf6/0x150
start_transaction+0x3a8/0x500
btrfs_commit_inode_delayed_inode+0x59/0x110
btrfs_evict_inode+0x19e/0x4c0
evict+0xbc/0x1f0
inode_lru_isolate+0x113/0x190
__list_lru_walk_one.isra.4+0x5c/0x100
list_lru_walk_one+0x32/0x50
prune_icache_sb+0x36/0x80
super_cache_scan+0x14a/0x1d0
do_shrink_slab+0x131/0x320
shrink_node+0xf7/0x380
balance_pgdat+0x2d5/0x640
kswapd+0x2ba/0x5e0
kthread+0x147/0x160
ret_from_fork+0x24/0x30
other info that might help us debug this:
Possible unsafe locking scenario:
CPU0 CPU1
---- ----
lock(fs_reclaim);
lock(sb_internal);
lock(fs_reclaim);
lock(sb_internal);
*** DEADLOCK ***
3 locks held by kswapd0/711:
#0: 000000000ba86300 (fs_reclaim){+.+.}, at: __fs_reclaim_acquire+0x0/0x30
#1: 000000004a5100f8 (shrinker_rwsem){++++}, at: shrink_node+0x9a/0x380
#2: 00000000f956fa46 (&type->s_umount_key#30){++++}, at: super_cache_scan+0x35/0x1d0
stack backtrace:
CPU: 7 PID: 711 Comm: kswapd0 Not tainted 5.2.11-gentoo #2
Hardware name: Dell Inc. Precision Tower 3620/0MWYPT, BIOS 2.4.2 09/29/2017
Call Trace:
dump_stack+0x85/0xc7
print_circular_bug.cold.40+0x1d9/0x235
__lock_acquire+0x18b1/0x1f00
lock_acquire+0xa6/0x170
? start_transaction+0x3a8/0x500
__sb_start_write+0xf6/0x150
? start_transaction+0x3a8/0x500
start_transaction+0x3a8/0x500
btrfs_commit_inode_delayed_inode+0x59/0x110
btrfs_evict_inode+0x19e/0x4c0
? var_wake_function+0x20/0x20
evict+0xbc/0x1f0
inode_lru_isolate+0x113/0x190
? discard_new_inode+0xc0/0xc0
__list_lru_walk_one.isra.4+0x5c/0x100
? discard_new_inode+0xc0/0xc0
list_lru_walk_one+0x32/0x50
prune_icache_sb+0x36/0x80
super_cache_scan+0x14a/0x1d0
do_shrink_slab+0x131/0x320
shrink_node+0xf7/0x380
balance_pgdat+0x2d5/0x640
kswapd+0x2ba/0x5e0
? __wake_up_common_lock+0x90/0x90
kthread+0x147/0x160
? balance_pgdat+0x640/0x640
? __kthread_create_on_node+0x160/0x160
ret_from_fork+0x24/0x30
This is because btrfs_new_inode() calls new_inode() under the
transaction. We could probably move the new_inode() outside of this but
for now just wrap it in memalloc_nofs_save().
Reported-by: Zdenek Sojka <zsojka(a)seznam.cz>
Fixes: 712e36c5f2a7 ("btrfs: use GFP_KERNEL in btrfs_alloc_inode")
CC: stable(a)vger.kernel.org # 4.16+
Reviewed-by: Filipe Manana <fdmanana(a)suse.com>
Signed-off-by: Josef Bacik <josef(a)toxicpanda.com>
Reviewed-by: David Sterba <dsterba(a)suse.com>
Signed-off-by: David Sterba <dsterba(a)suse.com>
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index a0546401bc0a..0f2754eaa05b 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -6305,13 +6305,16 @@ static struct inode *btrfs_new_inode(struct btrfs_trans_handle *trans,
u32 sizes[2];
int nitems = name ? 2 : 1;
unsigned long ptr;
+ unsigned int nofs_flag;
int ret;
path = btrfs_alloc_path();
if (!path)
return ERR_PTR(-ENOMEM);
+ nofs_flag = memalloc_nofs_save();
inode = new_inode(fs_info->sb);
+ memalloc_nofs_restore(nofs_flag);
if (!inode) {
btrfs_free_path(path);
return ERR_PTR(-ENOMEM);
The patch below does not apply to the 4.4-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable(a)vger.kernel.org>.
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
>From 4203e968947071586a98b5314fd7ffdea3b4f971 Mon Sep 17 00:00:00 2001
From: Josef Bacik <josef(a)toxicpanda.com>
Date: Mon, 30 Sep 2019 16:27:25 -0400
Subject: [PATCH] btrfs: fix incorrect updating of log root tree
We've historically had reports of being unable to mount file systems
because the tree log root couldn't be read. Usually this is the "parent
transid failure", but could be any of the related errors, including
"fsid mismatch" or "bad tree block", depending on which block got
allocated.
The modification of the individual log root items are serialized on the
per-log root root_mutex. This means that any modification to the
per-subvol log root_item is completely protected.
However we update the root item in the log root tree outside of the log
root tree log_mutex. We do this in order to allow multiple subvolumes
to be updated in each log transaction.
This is problematic however because when we are writing the log root
tree out we update the super block with the _current_ log root node
information. Since these two operations happen independently of each
other, you can end up updating the log root tree in between writing out
the dirty blocks and setting the super block to point at the current
root.
This means we'll point at the new root node that hasn't been written
out, instead of the one we should be pointing at. Thus whatever garbage
or old block we end up pointing at complains when we mount the file
system later and try to replay the log.
Fix this by copying the log's root item into a local root item copy.
Then once we're safely under the log_root_tree->log_mutex we update the
root item in the log_root_tree. This way we do not modify the
log_root_tree while we're committing it, fixing the problem.
CC: stable(a)vger.kernel.org # 4.4+
Reviewed-by: Chris Mason <clm(a)fb.com>
Reviewed-by: Filipe Manana <fdmanana(a)suse.com>
Signed-off-by: Josef Bacik <josef(a)toxicpanda.com>
Signed-off-by: David Sterba <dsterba(a)suse.com>
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 77b6797fcac3..2488eb4b70fc 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -2932,7 +2932,8 @@ static int walk_log_tree(struct btrfs_trans_handle *trans,
* in the tree of log roots
*/
static int update_log_root(struct btrfs_trans_handle *trans,
- struct btrfs_root *log)
+ struct btrfs_root *log,
+ struct btrfs_root_item *root_item)
{
struct btrfs_fs_info *fs_info = log->fs_info;
int ret;
@@ -2940,10 +2941,10 @@ static int update_log_root(struct btrfs_trans_handle *trans,
if (log->log_transid == 1) {
/* insert root item on the first sync */
ret = btrfs_insert_root(trans, fs_info->log_root_tree,
- &log->root_key, &log->root_item);
+ &log->root_key, root_item);
} else {
ret = btrfs_update_root(trans, fs_info->log_root_tree,
- &log->root_key, &log->root_item);
+ &log->root_key, root_item);
}
return ret;
}
@@ -3041,6 +3042,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info = root->fs_info;
struct btrfs_root *log = root->log_root;
struct btrfs_root *log_root_tree = fs_info->log_root_tree;
+ struct btrfs_root_item new_root_item;
int log_transid = 0;
struct btrfs_log_ctx root_log_ctx;
struct blk_plug plug;
@@ -3104,17 +3106,25 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
goto out;
}
+ /*
+ * We _must_ update under the root->log_mutex in order to make sure we
+ * have a consistent view of the log root we are trying to commit at
+ * this moment.
+ *
+ * We _must_ copy this into a local copy, because we are not holding the
+ * log_root_tree->log_mutex yet. This is important because when we
+ * commit the log_root_tree we must have a consistent view of the
+ * log_root_tree when we update the super block to point at the
+ * log_root_tree bytenr. If we update the log_root_tree here we'll race
+ * with the commit and possibly point at the new block which we may not
+ * have written out.
+ */
btrfs_set_root_node(&log->root_item, log->node);
+ memcpy(&new_root_item, &log->root_item, sizeof(new_root_item));
root->log_transid++;
log->log_transid = root->log_transid;
root->log_start_pid = 0;
- /*
- * Update or create log root item under the root's log_mutex to prevent
- * races with concurrent log syncs that can lead to failure to update
- * log root item because it was not created yet.
- */
- ret = update_log_root(trans, log);
/*
* IO has been started, blocks of the log tree have WRITTEN flag set
* in their headers. new modifications of the log will be written to
@@ -3135,6 +3145,14 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
mutex_unlock(&log_root_tree->log_mutex);
mutex_lock(&log_root_tree->log_mutex);
+
+ /*
+ * Now we are safe to update the log_root_tree because we're under the
+ * log_mutex, and we're a current writer so we're holding the commit
+ * open until we drop the log_mutex.
+ */
+ ret = update_log_root(trans, log, &new_root_item);
+
if (atomic_dec_and_test(&log_root_tree->log_writers)) {
/* atomic_dec_and_test implies a barrier */
cond_wake_up_nomb(&log_root_tree->log_writer_wait);
The patch below does not apply to the 4.9-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable(a)vger.kernel.org>.
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
>From 4203e968947071586a98b5314fd7ffdea3b4f971 Mon Sep 17 00:00:00 2001
From: Josef Bacik <josef(a)toxicpanda.com>
Date: Mon, 30 Sep 2019 16:27:25 -0400
Subject: [PATCH] btrfs: fix incorrect updating of log root tree
We've historically had reports of being unable to mount file systems
because the tree log root couldn't be read. Usually this is the "parent
transid failure", but could be any of the related errors, including
"fsid mismatch" or "bad tree block", depending on which block got
allocated.
The modification of the individual log root items are serialized on the
per-log root root_mutex. This means that any modification to the
per-subvol log root_item is completely protected.
However we update the root item in the log root tree outside of the log
root tree log_mutex. We do this in order to allow multiple subvolumes
to be updated in each log transaction.
This is problematic however because when we are writing the log root
tree out we update the super block with the _current_ log root node
information. Since these two operations happen independently of each
other, you can end up updating the log root tree in between writing out
the dirty blocks and setting the super block to point at the current
root.
This means we'll point at the new root node that hasn't been written
out, instead of the one we should be pointing at. Thus whatever garbage
or old block we end up pointing at complains when we mount the file
system later and try to replay the log.
Fix this by copying the log's root item into a local root item copy.
Then once we're safely under the log_root_tree->log_mutex we update the
root item in the log_root_tree. This way we do not modify the
log_root_tree while we're committing it, fixing the problem.
CC: stable(a)vger.kernel.org # 4.4+
Reviewed-by: Chris Mason <clm(a)fb.com>
Reviewed-by: Filipe Manana <fdmanana(a)suse.com>
Signed-off-by: Josef Bacik <josef(a)toxicpanda.com>
Signed-off-by: David Sterba <dsterba(a)suse.com>
diff --git a/fs/btrfs/tree-log.c b/fs/btrfs/tree-log.c
index 77b6797fcac3..2488eb4b70fc 100644
--- a/fs/btrfs/tree-log.c
+++ b/fs/btrfs/tree-log.c
@@ -2932,7 +2932,8 @@ static int walk_log_tree(struct btrfs_trans_handle *trans,
* in the tree of log roots
*/
static int update_log_root(struct btrfs_trans_handle *trans,
- struct btrfs_root *log)
+ struct btrfs_root *log,
+ struct btrfs_root_item *root_item)
{
struct btrfs_fs_info *fs_info = log->fs_info;
int ret;
@@ -2940,10 +2941,10 @@ static int update_log_root(struct btrfs_trans_handle *trans,
if (log->log_transid == 1) {
/* insert root item on the first sync */
ret = btrfs_insert_root(trans, fs_info->log_root_tree,
- &log->root_key, &log->root_item);
+ &log->root_key, root_item);
} else {
ret = btrfs_update_root(trans, fs_info->log_root_tree,
- &log->root_key, &log->root_item);
+ &log->root_key, root_item);
}
return ret;
}
@@ -3041,6 +3042,7 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
struct btrfs_fs_info *fs_info = root->fs_info;
struct btrfs_root *log = root->log_root;
struct btrfs_root *log_root_tree = fs_info->log_root_tree;
+ struct btrfs_root_item new_root_item;
int log_transid = 0;
struct btrfs_log_ctx root_log_ctx;
struct blk_plug plug;
@@ -3104,17 +3106,25 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
goto out;
}
+ /*
+ * We _must_ update under the root->log_mutex in order to make sure we
+ * have a consistent view of the log root we are trying to commit at
+ * this moment.
+ *
+ * We _must_ copy this into a local copy, because we are not holding the
+ * log_root_tree->log_mutex yet. This is important because when we
+ * commit the log_root_tree we must have a consistent view of the
+ * log_root_tree when we update the super block to point at the
+ * log_root_tree bytenr. If we update the log_root_tree here we'll race
+ * with the commit and possibly point at the new block which we may not
+ * have written out.
+ */
btrfs_set_root_node(&log->root_item, log->node);
+ memcpy(&new_root_item, &log->root_item, sizeof(new_root_item));
root->log_transid++;
log->log_transid = root->log_transid;
root->log_start_pid = 0;
- /*
- * Update or create log root item under the root's log_mutex to prevent
- * races with concurrent log syncs that can lead to failure to update
- * log root item because it was not created yet.
- */
- ret = update_log_root(trans, log);
/*
* IO has been started, blocks of the log tree have WRITTEN flag set
* in their headers. new modifications of the log will be written to
@@ -3135,6 +3145,14 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
mutex_unlock(&log_root_tree->log_mutex);
mutex_lock(&log_root_tree->log_mutex);
+
+ /*
+ * Now we are safe to update the log_root_tree because we're under the
+ * log_mutex, and we're a current writer so we're holding the commit
+ * open until we drop the log_mutex.
+ */
+ ret = update_log_root(trans, log, &new_root_item);
+
if (atomic_dec_and_test(&log_root_tree->log_writers)) {
/* atomic_dec_and_test implies a barrier */
cond_wake_up_nomb(&log_root_tree->log_writer_wait);
The patch below does not apply to the 4.19-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable(a)vger.kernel.org>.
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
>From c67d970f0ea8dcc423e112137d34334fa0abb8ec Mon Sep 17 00:00:00 2001
From: Filipe Manana <fdmanana(a)suse.com>
Date: Mon, 30 Sep 2019 10:20:25 +0100
Subject: [PATCH] Btrfs: fix memory leak due to concurrent append writes with
fiemap
When we have a buffered write that starts at an offset greater than or
equals to the file's size happening concurrently with a full ranged
fiemap, we can end up leaking an extent state structure.
Suppose we have a file with a size of 1Mb, and before the buffered write
and fiemap are performed, it has a single extent state in its io tree
representing the range from 0 to 1Mb, with the EXTENT_DELALLOC bit set.
The following sequence diagram shows how the memory leak happens if a
fiemap a buffered write, starting at offset 1Mb and with a length of
4Kb, are performed concurrently.
CPU 1 CPU 2
extent_fiemap()
--> it's a full ranged fiemap
range from 0 to LLONG_MAX - 1
(9223372036854775807)
--> locks range in the inode's
io tree
--> after this we have 2 extent
states in the io tree:
--> 1 for range [0, 1Mb[ with
the bits EXTENT_LOCKED and
EXTENT_DELALLOC_BITS set
--> 1 for the range
[1Mb, LLONG_MAX[ with
the EXTENT_LOCKED bit set
--> start buffered write at offset
1Mb with a length of 4Kb
btrfs_file_write_iter()
btrfs_buffered_write()
--> cached_state is NULL
lock_and_cleanup_extent_if_need()
--> returns 0 and does not lock
range because it starts
at current i_size / eof
--> cached_state remains NULL
btrfs_dirty_pages()
btrfs_set_extent_delalloc()
(...)
__set_extent_bit()
--> splits extent state for range
[1Mb, LLONG_MAX[ and now we
have 2 extent states:
--> one for the range
[1Mb, 1Mb + 4Kb[ with
EXTENT_LOCKED set
--> another one for the range
[1Mb + 4Kb, LLONG_MAX[ with
EXTENT_LOCKED set as well
--> sets EXTENT_DELALLOC on the
extent state for the range
[1Mb, 1Mb + 4Kb[
--> caches extent state
[1Mb, 1Mb + 4Kb[ into
@cached_state because it has
the bit EXTENT_LOCKED set
--> btrfs_buffered_write() ends up
with a non-NULL cached_state and
never calls anything to release its
reference on it, resulting in a
memory leak
Fix this by calling free_extent_state() on cached_state if the range was
not locked by lock_and_cleanup_extent_if_need().
The same issue can happen if anything else other than fiemap locks a range
that covers eof and beyond.
This could be triggered, sporadically, by test case generic/561 from the
fstests suite, which makes duperemove run concurrently with fsstress, and
duperemove does plenty of calls to fiemap. When CONFIG_BTRFS_DEBUG is set
the leak is reported in dmesg/syslog when removing the btrfs module with
a message like the following:
[77100.039461] BTRFS: state leak: start 6574080 end 6582271 state 16402 in tree 0 refs 1
Otherwise (CONFIG_BTRFS_DEBUG not set) detectable with kmemleak.
CC: stable(a)vger.kernel.org # 4.16+
Reviewed-by: Josef Bacik <josef(a)toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana(a)suse.com>
Reviewed-by: David Sterba <dsterba(a)suse.com>
Signed-off-by: David Sterba <dsterba(a)suse.com>
diff --git a/fs/btrfs/file.c b/fs/btrfs/file.c
index 8fe4eb7e5045..27e5b269e729 100644
--- a/fs/btrfs/file.c
+++ b/fs/btrfs/file.c
@@ -1591,7 +1591,6 @@ static noinline ssize_t btrfs_buffered_write(struct kiocb *iocb,
struct btrfs_fs_info *fs_info = btrfs_sb(inode->i_sb);
struct btrfs_root *root = BTRFS_I(inode)->root;
struct page **pages = NULL;
- struct extent_state *cached_state = NULL;
struct extent_changeset *data_reserved = NULL;
u64 release_bytes = 0;
u64 lockstart;
@@ -1611,6 +1610,7 @@ static noinline ssize_t btrfs_buffered_write(struct kiocb *iocb,
return -ENOMEM;
while (iov_iter_count(i) > 0) {
+ struct extent_state *cached_state = NULL;
size_t offset = offset_in_page(pos);
size_t sector_offset;
size_t write_bytes = min(iov_iter_count(i),
@@ -1758,9 +1758,20 @@ static noinline ssize_t btrfs_buffered_write(struct kiocb *iocb,
if (copied > 0)
ret = btrfs_dirty_pages(inode, pages, dirty_pages,
pos, copied, &cached_state);
+
+ /*
+ * If we have not locked the extent range, because the range's
+ * start offset is >= i_size, we might still have a non-NULL
+ * cached extent state, acquired while marking the extent range
+ * as delalloc through btrfs_dirty_pages(). Therefore free any
+ * possible cached extent state to avoid a memory leak.
+ */
if (extents_locked)
unlock_extent_cached(&BTRFS_I(inode)->io_tree,
lockstart, lockend, &cached_state);
+ else
+ free_extent_state(cached_state);
+
btrfs_delalloc_release_extents(BTRFS_I(inode), reserve_bytes,
true);
if (ret) {
From: Marco Felsch <m.felsch(a)pengutronix.de>
Since commit ec757001c818 ("gpio: Enable nonexclusive gpiods from DT
nodes") we are able to get GPIOD_FLAGS_BIT_NONEXCLUSIVE marked gpios.
Currently the gpiolib uses the wrong flags variable for the check. We
need to check the gpiod_flags instead of the of_gpio_flags else we
return -EBUSY for GPIOD_FLAGS_BIT_NONEXCLUSIVE marked and requested
gpiod's.
Fixes: ec757001c818 gpio: Enable nonexclusive gpiods from DT nodes
Cc: stable(a)vger.kernel.org
Signed-off-by: Marco Felsch <m.felsch(a)pengutronix.de>
[Bartosz: the function was moved to gpiolib-of.c so updated the patch]
Signed-off-by: Bartosz Golaszewski <bgolaszewski(a)baylibre.com>
[Bartosz: backported to v5.3.y]
Signed-off-by: Bartosz Golaszewski <bgolaszewski(a)baylibre.com>
---
drivers/gpio/gpiolib.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/drivers/gpio/gpiolib.c b/drivers/gpio/gpiolib.c
index d9074191edef..e4203c1eb869 100644
--- a/drivers/gpio/gpiolib.c
+++ b/drivers/gpio/gpiolib.c
@@ -4303,7 +4303,7 @@ struct gpio_desc *gpiod_get_from_of_node(struct device_node *node,
transitory = flags & OF_GPIO_TRANSITORY;
ret = gpiod_request(desc, label);
- if (ret == -EBUSY && (flags & GPIOD_FLAGS_BIT_NONEXCLUSIVE))
+ if (ret == -EBUSY && (dflags & GPIOD_FLAGS_BIT_NONEXCLUSIVE))
return desc;
if (ret)
return ERR_PTR(ret);
--
2.23.0
The patch below does not apply to the 5.3-stable tree.
If someone wants it applied there, or to any other stable or longterm
tree, then please email the backport, including the original git commit
id to <stable(a)vger.kernel.org>.
thanks,
greg k-h
------------------ original commit in Linus's tree ------------------
>From 1436a78c63495dd94c8d4f84a76d78d5317d481b Mon Sep 17 00:00:00 2001
From: Marco Felsch <m.felsch(a)pengutronix.de>
Date: Tue, 17 Sep 2019 16:56:36 +0200
Subject: [PATCH] iio: light: fix vcnl4000 devicetree hooks
Since commit ebd457d55911 ("iio: light: vcnl4000 add devicetree hooks")
the of_match_table is supported but the data shouldn't be a string.
Instead it shall be one of 'enum vcnl4000_device_ids'. Also the matching
logic for the vcnl4020 was wrong. Since the data retrieve mechanism is
still based on the i2c_device_id no failures did appeared till now.
Fixes: ebd457d55911 ("iio: light: vcnl4000 add devicetree hooks")
Signed-off-by: Marco Felsch <m.felsch(a)pengutronix.de>
Reviewed-by: Angus Ainslie (Purism) angus(a)akkea.ca
Cc: <Stable(a)vger.kernel.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron(a)huawei.com>
diff --git a/drivers/iio/light/vcnl4000.c b/drivers/iio/light/vcnl4000.c
index 51421ac32517..f522cb863e8c 100644
--- a/drivers/iio/light/vcnl4000.c
+++ b/drivers/iio/light/vcnl4000.c
@@ -398,19 +398,19 @@ static int vcnl4000_probe(struct i2c_client *client,
static const struct of_device_id vcnl_4000_of_match[] = {
{
.compatible = "vishay,vcnl4000",
- .data = "VCNL4000",
+ .data = (void *)VCNL4000,
},
{
.compatible = "vishay,vcnl4010",
- .data = "VCNL4010",
+ .data = (void *)VCNL4010,
},
{
- .compatible = "vishay,vcnl4010",
- .data = "VCNL4020",
+ .compatible = "vishay,vcnl4020",
+ .data = (void *)VCNL4010,
},
{
.compatible = "vishay,vcnl4200",
- .data = "VCNL4200",
+ .data = (void *)VCNL4200,
},
{},
};