zero-copy

I'm trying to experiment with using splice ( man 2 splice ) to copy data from a UDP socket directly to a file. Unfortunately the first call to splice() returns EINVAL. The man page states: EINVAL Target file system doesn't support splicing; target file is opened in append mode; neither of the descriptors refers to a pipe; or offset given for nonseekable device. However, I believe none of those conditions apply. I'm using Fedora 15 (kernel 2.6.40-4) so I believe splice() is supported on all filesystems. The target file should be irrelevant in the first call to splice, but for completeness I'm

I'm seeing some strange behavior when transferring large files from file to socket using zero-copy in Java. My environments: Windows 7 64-bit JDK 1.6.0_45 and 1.7.0_79. Centos 6.6 64-bit JDK 1.6.0_35 What the program does: client copies an input file into a socket, and server copies socket to output file using zero-copy methods: transferFrom and transferTo. Not all bytes are reaching the server if file size is relatively large, 100Mb+ in case of Windows and 2GB+ in case of Centos. Client and server reside on the same machine and localhost address is used to transfer data. The behavior is

Does anyone have experience with analyzing the performance of CUDA applications utilizing the zero-copy (reference here: Default Pinned Memory Vs Zero-Copy Memory ) memory model? I have a kernel that uses the zero-copy feature and with NVVP I see the following: Running the kernel on an average problem size I get instruction replay overhead of 0.7%, so nothing major. And all of this 0.7% is global memory replay overhead. When I really jack up the problem size, I get an instruction replay overhead of 95.7%, all of which is due to global memory replay overhead. However, the global load efficiency

The man page for the splice system call says that splice may fail and set errno to EINVAL if: Target file system doesn't support splicing; neither of the descriptors refers to a pipe; or offset given for non-seekable device Which file systems support splicing? My original answer was partially incorrect, this is a major rewrite. Linux 2.6.30.10 and below In Linux 2.6.30.10 and older, splice returns EINVAL when the source or target filesystem does not support splicing. Here are the filesystems that do support splicing : in read mode : adfs, affs, afs, bfs, btrfs, coda, ecryptfs, exofs, ext2,

I'd like to know the reasons for why the linux kernel (or any other mainstream OS) does not have a feature for zero copy networking ? By zero copy I mean, that an packet/datastream does not get copied for passing to an application in userspace but e.g. uses a memory-pool type of allocator to share the memory between kernel and userspace. I've came up with 3 theory's on my own: a) I guess there are security concerns. But is there really no way of making memory shared securily between userspace and kernel when they are just used as a buffer ? b) I guess there are stability concerns. But can't we