summaryrefslogtreecommitdiff
path: root/fs/btrfs/extent_io.c
diff options
context:
space:
mode:
authorGreg Kroah-Hartman <gregkh@google.com>2020-07-22 09:39:11 +0200
committerGreg Kroah-Hartman <gregkh@google.com>2020-07-22 09:39:11 +0200
commit18971409ef0ad1c104bb96fd2ce74d1fd907082b (patch)
tree4e9151404808b38a4a8c5f2572f5b23dbcf095c0 /fs/btrfs/extent_io.c
parent94db81b5f333b2fb868347998b92d72c8665d348 (diff)
parent554bbfc0d87fcbc842a18997c2a11a772dc3f003 (diff)
Merge 4.4.231 into android-4.4-p
Changes in 4.4.231 KVM: s390: reduce number of IO pins to 1 spi: spidev: fix a race between spidev_release and spidev_remove spi: spidev: fix a potential use-after-free in spidev_release() scsi: mptscsih: Fix read sense data size net: cxgb4: fix return error value in t4_prep_fw smsc95xx: check return value of smsc95xx_reset smsc95xx: avoid memory leak in smsc95xx_bind ALSA: compress: fix partial_drain completion state arm64: kgdb: Fix single-step exception handling oops ALSA: opl3: fix infoleak in opl3 ALSA: hda - let hs_mic be picked ahead of hp_mic ALSA: usb-audio: add quirk for MacroSilicon MS2109 KVM: x86: bit 8 of non-leaf PDPEs is not reserved Revert "ath9k: Fix general protection fault in ath9k_hif_usb_rx_cb" btrfs: fix fatal extent_buffer readahead vs releasepage race drm/radeon: fix double free ARC: entry: fix potential EFA clobber when TIF_SYSCALL_TRACE ARC: elf: use right ELF_ARCH bnxt_en: fix NULL dereference in case SR-IOV configuration fails ipv4: fill fl4_icmp_{type,code} in ping_v4_sendmsg l2tp: remove skb_dst_set() from l2tp_xmit_skb() llc: make sure applications use ARPHRD_ETHER net: Added pointer check for dst->ops->neigh_lookup in dst_neigh_lookup_skb genetlink: remove genl_bind tcp: make sure listeners don't initialize congestion-control state tcp: md5: add missing memory barriers in tcp_md5_do_add()/tcp_md5_hash_key() tcp: md5: refine tcp_md5_do_add()/tcp_md5_hash_key() barriers tcp: md5: allow changing MD5 keys in all socket states i2c: eg20t: Load module automatically if ID matches Revert "usb/ehci-platform: Set PM runtime as active on resume" Revert "usb/xhci-plat: Set PM runtime as active on resume" Revert "usb/ohci-platform: Fix a warning when hibernating" usb: gadget: udc: atmel: fix uninitialized read in debug printk staging: comedi: verify array index is correct before using it perf stat: Zero all the 'ena' and 'run' array slot stats for interval mode mtd: rawnand: brcmnand: fix CS0 layout HID: magicmouse: do not set up autorepeat usb: core: Add a helper function to check the validity of EP type in URB ALSA: line6: Perform sanity check for each URB creation ALSA: usb-audio: Fix race against the error recovery URB submission USB: c67x00: fix use after free in c67x00_giveback_urb usb: chipidea: core: add wakeup support for extcon usb: gadget: function: fix missing spinlock in f_uac1_legacy USB: serial: iuu_phoenix: fix memory corruption USB: serial: cypress_m8: enable Simply Automated UPB PIM USB: serial: ch341: add new Product ID for CH340 USB: serial: option: add GosunCn GM500 series USB: serial: option: add Quectel EG95 LTE modem virtio: virtio_console: add missing MODULE_DEVICE_TABLE() for rproc serial fuse: Fix parameter for FS_IOC_{GET,SET}FLAGS mei: bus: don't clean driver pointer Input: i8042 - add Lenovo XiaoXin Air 12 to i8042 nomux list uio_pdrv_genirq: fix use without device tree and no interrupt MIPS: Fix build for LTS kernel caused by backporting lpj adjustment hwmon: (emc2103) fix unable to change fan pwm1_enable attribute dmaengine: fsl-edma: Fix NULL pointer exception in fsl_edma_tx_handler misc: atmel-ssc: lock with mutex instead of spinlock sched/fair: handle case of task_h_load() returning 0 Linux 4.4.231 Signed-off-by: Greg Kroah-Hartman <gregkh@google.com> Change-Id: Ibbd0a09b0d8d49101c674cf304c3df711e940b77
Diffstat (limited to 'fs/btrfs/extent_io.c')
-rw-r--r--fs/btrfs/extent_io.c40
1 files changed, 24 insertions, 16 deletions
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 2c86c472f670..42b7409d4cc5 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -4861,25 +4861,28 @@ err:
static void check_buffer_tree_ref(struct extent_buffer *eb)
{
int refs;
- /* the ref bit is tricky. We have to make sure it is set
- * if we have the buffer dirty. Otherwise the
- * code to free a buffer can end up dropping a dirty
- * page
+ /*
+ * The TREE_REF bit is first set when the extent_buffer is added
+ * to the radix tree. It is also reset, if unset, when a new reference
+ * is created by find_extent_buffer.
*
- * Once the ref bit is set, it won't go away while the
- * buffer is dirty or in writeback, and it also won't
- * go away while we have the reference count on the
- * eb bumped.
+ * It is only cleared in two cases: freeing the last non-tree
+ * reference to the extent_buffer when its STALE bit is set or
+ * calling releasepage when the tree reference is the only reference.
*
- * We can't just set the ref bit without bumping the
- * ref on the eb because free_extent_buffer might
- * see the ref bit and try to clear it. If this happens
- * free_extent_buffer might end up dropping our original
- * ref by mistake and freeing the page before we are able
- * to add one more ref.
+ * In both cases, care is taken to ensure that the extent_buffer's
+ * pages are not under io. However, releasepage can be concurrently
+ * called with creating new references, which is prone to race
+ * conditions between the calls to check_buffer_tree_ref in those
+ * codepaths and clearing TREE_REF in try_release_extent_buffer.
*
- * So bump the ref count first, then set the bit. If someone
- * beat us to it, drop the ref we added.
+ * The actual lifetime of the extent_buffer in the radix tree is
+ * adequately protected by the refcount, but the TREE_REF bit and
+ * its corresponding reference are not. To protect against this
+ * class of races, we call check_buffer_tree_ref from the codepaths
+ * which trigger io after they set eb->io_pages. Note that once io is
+ * initiated, TREE_REF can no longer be cleared, so that is the
+ * moment at which any such race is best fixed.
*/
refs = atomic_read(&eb->refs);
if (refs >= 2 && test_bit(EXTENT_BUFFER_TREE_REF, &eb->bflags))
@@ -5346,6 +5349,11 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
clear_bit(EXTENT_BUFFER_READ_ERR, &eb->bflags);
eb->read_mirror = 0;
atomic_set(&eb->io_pages, num_reads);
+ /*
+ * It is possible for releasepage to clear the TREE_REF bit before we
+ * set io_pages. See check_buffer_tree_ref for a more detailed comment.
+ */
+ check_buffer_tree_ref(eb);
for (i = start_i; i < num_pages; i++) {
page = eb->pages[i];
if (!PageUptodate(page)) {