From: Jeff Layton <jlayton@xxxxxxxxxx>
Some time ago, the PostgreSQL developers mentioned that they'd like a
way to tell whether there have been any writeback errors on a given
filesystem without having to forcibly sync out all buffered writes.
Now that we have a per-sb errseq_t that tracks whether any inode on the
filesystem might have failed writeback, we can present that to userland
applications via a new interface. Add a new generic fs ioctl for that
purpose. This just reports the current state of the errseq_t counter
with the SEEN bit masked off.
Cc: Andres Freund <andres@xxxxxxxxxxx>
Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
---
fs/ioctl.c | 4 ++++
include/linux/errseq.h | 1 +
include/uapi/linux/fs.h | 1 +
lib/errseq.c | 33 +++++++++++++++++++++++++++++++--
4 files changed, 37 insertions(+), 2 deletions(-)
diff --git a/fs/ioctl.c b/fs/ioctl.c
index 7c9a5df5a597..41e991cec4c3 100644
--- a/fs/ioctl.c
+++ b/fs/ioctl.c
@@ -705,6 +705,10 @@ static int do_vfs_ioctl(struct file *filp, unsigned int fd,
case FS_IOC_FIEMAP:
return ioctl_fiemap(filp, argp);
+ case FS_IOC_GETFSERR:
+ return put_user(errseq_scrape(&inode->i_sb->s_wb_err),
+ (unsigned int __user *)argp);
+
case FIGETBSZ:
/* anon_bdev filesystems may not have a block size */
if (!inode->i_sb->s_blocksize)
diff --git a/include/linux/errseq.h b/include/linux/errseq.h
index fc2777770768..de165623fa86 100644
--- a/include/linux/errseq.h
+++ b/include/linux/errseq.h
@@ -9,6 +9,7 @@ typedef u32 errseq_t;
errseq_t errseq_set(errseq_t *eseq, int err);
errseq_t errseq_sample(errseq_t *eseq);
+errseq_t errseq_scrape(errseq_t *eseq);
int errseq_check(errseq_t *eseq, errseq_t since);
int errseq_check_and_advance(errseq_t *eseq, errseq_t *since);
#endif
diff --git a/include/uapi/linux/fs.h b/include/uapi/linux/fs.h
index 379a612f8f1d..c39b37fba7f9 100644
--- a/include/uapi/linux/fs.h
+++ b/include/uapi/linux/fs.h
@@ -214,6 +214,7 @@ struct fsxattr {
#define FS_IOC_FSSETXATTR _IOW('X', 32, struct fsxattr)
#define FS_IOC_GETFSLABEL _IOR(0x94, 49, char[FSLABEL_MAX])
#define FS_IOC_SETFSLABEL _IOW(0x94, 50, char[FSLABEL_MAX])
+#define FS_IOC_GETFSERR _IOR('e', 1, unsigned int)
/*
* Inode flags (FS_IOC_GETFLAGS / FS_IOC_SETFLAGS)
diff --git a/lib/errseq.c b/lib/errseq.c
index 81f9e33aa7e7..8ded0920eed3 100644
--- a/lib/errseq.c
+++ b/lib/errseq.c
@@ -108,7 +108,7 @@ errseq_t errseq_set(errseq_t *eseq, int err)
EXPORT_SYMBOL(errseq_set);
/**
- * errseq_sample() - Grab current errseq_t value.
+ * errseq_sample() - Grab current errseq_t value (or 0 if it hasn't been seen)
* @eseq: Pointer to errseq_t to be sampled.
*
* This function allows callers to initialise their errseq_t variable.
@@ -117,7 +117,7 @@ EXPORT_SYMBOL(errseq_set);
* see it the next time it checks for an error.
*
* Context: Any context.
- * Return: The current errseq value.
+ * Return: The current errseq value or 0 if it wasn't previously seen
*/
errseq_t errseq_sample(errseq_t *eseq)
{
@@ -130,6 +130,35 @@ errseq_t errseq_sample(errseq_t *eseq)
}
EXPORT_SYMBOL(errseq_sample);
+/**
+ * errseq_scrape() - Grab current errseq_t value
+ * @eseq: Pointer to errseq_t to be sampled.
+ *
+ * This function allows callers to scrape the current value of an errseq_t.
+ * Unlike errseq_sample, this will always return the current value with
+ * the SEEN flag unset, even when the value has not yet been seen.
+ *
+ * Context: Any context.
+ * Return: The current errseq value with ERRSEQ_SEEN masked off
+ */
+errseq_t errseq_scrape(errseq_t *eseq)
+{
+ errseq_t old = READ_ONCE(*eseq);
+
+ /*
+ * For the common case of no errors ever having been set, we can skip
+ * marking the SEEN bit. Once an error has been set, the value will
+ * never go back to zero.
+ */
+ if (old != 0) {
+ errseq_t new = old | ERRSEQ_SEEN;
+ if (old != new)
+ cmpxchg(eseq, old, new);
+ }
+ return old & ~ERRSEQ_SEEN;
+}
+EXPORT_SYMBOL(errseq_scrape);
+
/**
* errseq_check() - Has an error occurred since a particular sample point?
* @eseq: Pointer to errseq_t value to be checked.
--
2.24.1
