On 11/6/23 19:46, Ilya Dryomov wrote:
On Mon, Nov 6, 2023 at 2:05 AM <xiubli@xxxxxxxxxx> wrote:
From: Xiubo Li <xiubli@xxxxxxxxxx>
There is no any limit for the extent array size and it's possible
that we will hit 4096 limit just after a lot of random writes to
a file and then read with a large size. In this case the messager
will fail by reseting the connection and keeps resending the inflight
IOs infinitely.
Just increase the limit to a larger number and then warn it to
let user know that allocating memory could fail with this.
URL: https://tracker.ceph.com/issues/62081
Signed-off-by: Xiubo Li <xiubli@xxxxxxxxxx>
---
V2:
- Increase the MAX_EXTENTS instead of removing it.
- Do not return an errno when hit the limit.
net/ceph/osd_client.c | 15 +++++++--------
1 file changed, 7 insertions(+), 8 deletions(-)
diff --git a/net/ceph/osd_client.c b/net/ceph/osd_client.c
index c03d48bd3aff..050dc39065fb 100644
--- a/net/ceph/osd_client.c
+++ b/net/ceph/osd_client.c
@@ -5850,7 +5850,7 @@ static inline void convert_extent_map(struct ceph_sparse_read *sr)
}
#endif
-#define MAX_EXTENTS 4096
+#define MAX_EXTENTS (16*1024*1024)
I don't think this is a sensible limit -- see my other reply.
Ilya,
As I mentioned in that thread, the sparse read could be enabled in
non-fscrypt case. If so the "64M (CEPH_MSG_MAX_DATA_LEN) / 4K = 16384"
still won't be enough.
Thanks
- Xiubo
Thanks,
Ilya