There are two convenient ways to report errors to userspace
1) retun error to original syscall for example write(2)
2) mark mapping with error flag and return it on later fsync(2)
Second one is broken if (mapping->nrpages == 0)
This is real-life situation because after error pages are likey
to be truncated or invalidated.
We have to return an error regardless to number of pages in the mapping.
#Original testcase: [email protected]:dmonakhov/xfstests.git
MOUNT_OPTIONS="-b1024"
./check shared/305
Signed-off-by: Dmitry Monakhov <[email protected]>
---
mm/filemap.c | 29 +++++++++++++++++++++--------
1 files changed, 21 insertions(+), 8 deletions(-)
diff --git a/mm/filemap.c b/mm/filemap.c
index e1979fd..ca47f0b 100644
--- a/mm/filemap.c
+++ b/mm/filemap.c
@@ -184,6 +184,17 @@ static int sleep_on_page_killable(void *word)
return fatal_signal_pending(current) ? -EINTR : 0;
}
+static int filemap_check_errors(struct address_space *mapping)
+{
+ int ret = 0;
+ /* Check for outstanding write errors */
+ if (test_and_clear_bit(AS_ENOSPC, &mapping->flags))
+ ret = -ENOSPC;
+ if (test_and_clear_bit(AS_EIO, &mapping->flags))
+ ret = -EIO;
+ return ret;
+}
+
/**
* __filemap_fdatawrite_range - start writeback on mapping dirty pages in range
* @mapping: address space structure to write
@@ -265,10 +276,10 @@ int filemap_fdatawait_range(struct address_space *mapping, loff_t start_byte,
pgoff_t end = end_byte >> PAGE_CACHE_SHIFT;
struct pagevec pvec;
int nr_pages;
- int ret = 0;
+ int ret2, ret = 0;
if (end_byte < start_byte)
- return 0;
+ goto out;
pagevec_init(&pvec, 0);
while ((index <= end) &&
@@ -291,12 +302,10 @@ int filemap_fdatawait_range(struct address_space *mapping, loff_t start_byte,
pagevec_release(&pvec);
cond_resched();
}
-
- /* Check for outstanding write errors */
- if (test_and_clear_bit(AS_ENOSPC, &mapping->flags))
- ret = -ENOSPC;
- if (test_and_clear_bit(AS_EIO, &mapping->flags))
- ret = -EIO;
+out:
+ ret2 = filemap_check_errors(mapping);
+ if (!ret)
+ ret = ret2;
return ret;
}
@@ -337,6 +346,8 @@ int filemap_write_and_wait(struct address_space *mapping)
if (!err)
err = err2;
}
+ } else {
+ err = filemap_check_errors(mapping);
}
return err;
}
@@ -368,6 +379,8 @@ int filemap_write_and_wait_range(struct address_space *mapping,
if (!err)
err = err2;
}
+ } else {
+ err = filemap_check_errors(mapping);
}
return err;
}
--
1.7.1
On Thu 11-04-13 16:47:59, Dmitry Monakhov wrote:
> There are two convenient ways to report errors to userspace
> 1) retun error to original syscall for example write(2)
> 2) mark mapping with error flag and return it on later fsync(2)
>
> Second one is broken if (mapping->nrpages == 0)
> This is real-life situation because after error pages are likey
> to be truncated or invalidated.
>
> We have to return an error regardless to number of pages in the mapping.
>
> #Original testcase: [email protected]:dmonakhov/xfstests.git
> MOUNT_OPTIONS="-b1024"
> ./check shared/305
Makes sence to me. You can add:
Reviewed-by: Jan Kara <[email protected]>
BTW, this more belongs to the mm area so Andrew Morton might be the right
person to merge this. Added to CC.
Honza
>
> Signed-off-by: Dmitry Monakhov <[email protected]>
> ---
> mm/filemap.c | 29 +++++++++++++++++++++--------
> 1 files changed, 21 insertions(+), 8 deletions(-)
>
> diff --git a/mm/filemap.c b/mm/filemap.c
> index e1979fd..ca47f0b 100644
> --- a/mm/filemap.c
> +++ b/mm/filemap.c
> @@ -184,6 +184,17 @@ static int sleep_on_page_killable(void *word)
> return fatal_signal_pending(current) ? -EINTR : 0;
> }
>
> +static int filemap_check_errors(struct address_space *mapping)
> +{
> + int ret = 0;
> + /* Check for outstanding write errors */
> + if (test_and_clear_bit(AS_ENOSPC, &mapping->flags))
> + ret = -ENOSPC;
> + if (test_and_clear_bit(AS_EIO, &mapping->flags))
> + ret = -EIO;
> + return ret;
> +}
> +
> /**
> * __filemap_fdatawrite_range - start writeback on mapping dirty pages in range
> * @mapping: address space structure to write
> @@ -265,10 +276,10 @@ int filemap_fdatawait_range(struct address_space *mapping, loff_t start_byte,
> pgoff_t end = end_byte >> PAGE_CACHE_SHIFT;
> struct pagevec pvec;
> int nr_pages;
> - int ret = 0;
> + int ret2, ret = 0;
>
> if (end_byte < start_byte)
> - return 0;
> + goto out;
>
> pagevec_init(&pvec, 0);
> while ((index <= end) &&
> @@ -291,12 +302,10 @@ int filemap_fdatawait_range(struct address_space *mapping, loff_t start_byte,
> pagevec_release(&pvec);
> cond_resched();
> }
> -
> - /* Check for outstanding write errors */
> - if (test_and_clear_bit(AS_ENOSPC, &mapping->flags))
> - ret = -ENOSPC;
> - if (test_and_clear_bit(AS_EIO, &mapping->flags))
> - ret = -EIO;
> +out:
> + ret2 = filemap_check_errors(mapping);
> + if (!ret)
> + ret = ret2;
>
> return ret;
> }
> @@ -337,6 +346,8 @@ int filemap_write_and_wait(struct address_space *mapping)
> if (!err)
> err = err2;
> }
> + } else {
> + err = filemap_check_errors(mapping);
> }
> return err;
> }
> @@ -368,6 +379,8 @@ int filemap_write_and_wait_range(struct address_space *mapping,
> if (!err)
> err = err2;
> }
> + } else {
> + err = filemap_check_errors(mapping);
> }
> return err;
> }
> --
> 1.7.1
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
--
Jan Kara <[email protected]>
SUSE Labs, CR