Hi,
We are chasing a data corruption issue seen with our pseudo FSD - it seems
to show up only when smb signing is turned. Is there any known issue with
smb signing?
The test scenario involves a client making repeated operations on a file
over the network i.e. SRV is in the picture, and the corruption happens only
when smb signing is turned on.
Our FSD uses NTFS for storage, but does its own caching. When the corruption
happens, we find that 196 bytes at the beginning of a cached page inside the
NT Cache Manager region is zeroed out. The data is fine on disk i.e. when
accessed directly on the NTFS volume.
We have instrumented our memory allocation/manipulation code and have
confirmed that none of our data structures are 196 bytes and we are not
accidentally zeroing out the range.
Any quick thoughts/pointers?
Thanks
–suresh