kernel.git/tools/testing/selftests/net/tcp_mmap.c, branch master

selftests/net: report rcv_mss in tcp_mmap

2023-08-02T10:40:49Z

tcp_mmap tests TCP_ZEROCOPY_RECEIVE. If 0% of data is received using mmap, this may be due to mss. Report rcv_mss to identify this cause. Output of a run failed due to too small mss: received 32768 MB (0 % mmap'ed) in 8.40458 s, 32.7057 Gbit cpu usage user:0.027922 sys:8.21126, 251.44 usec per MB, 3252 c-switches, rcv_mss 1428 Output on a successful run: received 32768 MB (99.9507 % mmap'ed) in 4.69023 s, 58.6064 Gbit cpu usage user:0.029172 sys:2.56105, 79.0473 usec per MB, 57591 c-switches, rcv_mss 4096 Suggested-by: Eric Dumazet Signed-off-by: Willem de Bruijn Signed-off-by: David S. Miller

selftests/net: fix typo in tcp_mmap

2023-04-06T11:19:25Z

kernel test robot reported the following warning: All warnings (new ones prefixed by >>): tcp_mmap.c: In function 'child_thread': >> tcp_mmap.c:211:61: warning: 'lu' may be used uninitialized in this function [-Wmaybe-uninitialized] 211 | zc.length = min(chunk_size, FILE_SZ - lu); We want to read FILE_SZ bytes, so the correct expression should be (FILE_SZ - total) Fixes: 5c5945dc695c ("selftests/net: Add SHA256 computation over data sent in tcp_mmap") Reported-by: kernel test robot Link: https://lore.kernel.org/oe-kbuild-all/202304042104.UFIuevBp-lkp@intel.com/ Signed-off-by: Eric Dumazet Cc: Xiaoyan Li Cc: Kuniyuki Iwashima Link: https://lore.kernel.org/r/20230405071556.1019623-1-edumazet@google.com Signed-off-by: Paolo Abeni

selftests/net: Add SHA256 computation over data sent in tcp_mmap

2023-03-22T14:34:31Z

Add option to compute and send SHA256 over data sent (-i). This is to ensure the correctness of data received. Data is randomly populated from /dev/urandom. Tested: ./tcp_mmap -s -z -i ./tcp_mmap -z -H $ADDR -i SHA256 is correct ./tcp_mmap -s -i ./tcp_mmap -H $ADDR -i SHA256 is correct Signed-off-by: Coco Li Reviewed-by: Eric Dumazet Link: https://lore.kernel.org/r/20230321081202.2370275-2-lixiaoyan@google.com Signed-off-by: Paolo Abeni

selftests: net: tcp_mmap: populate pages in send path

2023-01-24T05:24:29Z

In commit 72653ae5303c ("selftests: net: tcp_mmap: Use huge pages in send path") I made a change to use hugepages for the buffer used by the client (tx path) Today, I understood that the cause for poor zerocopy performance was that after a mmap() for a 512KB memory zone, kernel uses a single zeropage, mapped 128 times. This was really the reason for poor tx path performance in zero copy mode, because this zero page refcount is under high pressure, especially when TCP ACK packets are processed on another cpu. We need either to force a COW on all the memory range, or use MAP_POPULATE so that a zero page is not abused. Signed-off-by: Eric Dumazet Link: https://lore.kernel.org/r/20230120181136.3764521-1-edumazet@google.com Signed-off-by: Jakub Kicinski

selftests: net: tcp_mmap: Use huge pages in receive path

2020-08-20T23:14:53Z

One down side of using TCP rx zerocopy is one extra TLB miss per page after the mapping operation. While if the application is using hugepages, the non zerocopy recvmsg() will not have to pay these TLB costs. This patch allows server side to use huge pages for the non zero copy case, to allow fair comparisons when both solutions use optimal conditions. Signed-off-by: Eric Dumazet Cc: Arjun Roy Cc: Soheil Hassas Yeganeh Signed-off-by: David S. Miller

selftests: net: tcp_mmap: Use huge pages in send path

2020-08-20T23:14:53Z

There are significant gains using huge pages when available, as shown in [1]. This patch adds mmap_large_buffer() and uses it in client side (tx path of this reference tool) Following patch will use the feature for server side. [1] https://patchwork.ozlabs.org/project/netdev/patch/20200820154359.1806305-1-edumazet@google.com/ Signed-off-by: Eric Dumazet Cc: Arjun Roy Cc: Soheil Hassas Yeganeh Signed-off-by: David S. Miller

selftests: net: tcp_mmap: use madvise(MADV_DONTNEED)

2020-08-20T23:14:53Z

When TCP_ZEROCOPY_RECEIVE operation has been added, I made the mistake of automatically un-mapping prior content before mapping new pages. This has the unfortunate effect of adding potentially long MMU operations (like TLB flushes) while socket lock is held. Using madvise(MADV_DONTNEED) right after pages has been used has two benefits : 1) This releases pages sooner, allowing pages to be recycled if they were part of a page pool in a NIC driver. 2) No more long unmap operations while preventing immediate processing of incoming packets. The cost of the added system call is small enough. Arjun will submit a kernel patch allowing to opt out from the unmap attempt in tcp_zerocopy_receive() Signed-off-by: Eric Dumazet Cc: Arjun Roy Cc: Soheil Hassas Yeganeh Signed-off-by: David S. Miller

selftests/net: tcp_mmap: fix clang warning for target arch PowerPC

2020-07-27T19:56:59Z

When size_t maps to unsigned int (e.g. on 32-bit powerpc), then the comparison with 1<<35 is always true. Clang 9 threw: warning: result of comparison of constant 34359738368 with \ expression of type 'size_t' (aka 'unsigned int') is always true \ [-Wtautological-constant-out-of-range-compare] while (total < FILE_SZ) { Tested: make -C tools/testing/selftests TARGETS="net" run_tests Fixes: 192dc405f308 ("selftests: net: add tcp_mmap program") Signed-off-by: Tanner Love Acked-by: Willem de Bruijn Reviewed-by: Eric Dumazet Signed-off-by: David S. Miller

selftests: net: tcp_mmap: fix SO_RCVLOWAT setting

2020-05-06T22:01:30Z

Since chunk_size is no longer an integer, we can not use it directly as an argument of setsockopt(). This patch should fix tcp_mmap for Big Endian kernels. Fixes: 597b01edafac ("selftests: net: avoid ptl lock contention in tcp_mmap") Signed-off-by: Eric Dumazet Cc: Soheil Hassas Yeganeh Cc: Arjun Roy Acked-by: Soheil Hassas Yeganeh Signed-off-by: David S. Miller

selftests: net: tcp_mmap: clear whole tcp_zerocopy_receive struct

2020-05-06T20:45:09Z

We added fields in tcp_zerocopy_receive structure, so make sure to clear all fields to not pass garbage to the kernel. We were lucky because recent additions added 'out' parameters, still we need to clean our reference implementation, before folks copy/paste it. Fixes: c8856c051454 ("tcp-zerocopy: Return inq along with tcp receive zerocopy.") Fixes: 33946518d493 ("tcp-zerocopy: Return sk_err (if set) along with tcp receive zerocopy.") Signed-off-by: Eric Dumazet Cc: Arjun Roy Cc: Soheil Hassas Yeganeh Acked-by: Soheil Hassas Yeganeh Signed-off-by: David S. Miller