/* * netsniff-ng - the packet sniffing beast * Copyright 2011 - 2013 Daniel Borkmann. * Subject to the GPL, version 2. */ #include #include #include #include #include #include #include "pcap_io.h" #include "xmalloc.h" #include "built_in.h" #include "iosched.h" #include "ioops.h" static struct iovec iov[1024] __cacheline_aligned; static off_t iov_off_rd = 0, iov_slot = 0; static ssize_t pcap_sg_write(int fd, pcap_pkthdr_t *phdr, enum pcap_type type, const uint8_t *packet, size_t len) { ssize_t ret, hdrsize = pcap_get_hdr_length(phdr, type); if (unlikely(iov_slot == array_size(iov))) { ret = writev(fd, iov, array_size(iov)); if (ret < 0) panic("Writev I/O error: %s!\n", strerror(errno)); iov_slot = 0; } fmemcpy(iov[iov_slot].iov_base, &phdr->raw, hdrsize); iov[iov_slot].iov_len = hdrsize; fmemcpy(iov[iov_slot].iov_base + iov[iov_slot].iov_len, packet, len); ret = (iov[iov_slot].iov_len += len); iov_slot++; return ret; } static ssize_t __pcap_sg_inter_iov_hdr_read(int fd, pcap_pkthdr_t *phdr, size_t hdrsize) { int ret; size_t offset = 0; ssize_t remainder; offset = iov[iov_slot].iov_len - iov_off_rd; remainder = hdrsize - offset; if (remainder < 0) remainder = 0; bug_on(offset + remainder != hdrsize); fmemcpy(&phdr->raw, iov[iov_slot].iov_base + iov_off_rd, offset); iov_off_rd = 0; iov_slot++; if (iov_slot == array_size(iov)) { iov_slot = 0; ret = readv(fd, iov, array_size(iov)); if (unlikely(ret <= 0)) return -EIO; } fmemcpy(&phdr->raw + offset, iov[iov_slot].iov_base + iov_off_rd, remainder); iov_off_rd += remainder; return hdrsize; } static ssize_t __pcap_sg_inter_iov_data_read(int fd, uint8_t *packet, size_t hdrlen) { int ret; size_t offset = 0; ssize_t remainder; offset = iov[iov_slot].iov_len - iov_off_rd; remainder = hdrlen - offset; if (remainder < 0) remainder = 0; bug_on(offset + remainder != hdrlen); fmemcpy(packet, iov[iov_slot].iov_base + iov_off_rd, offset); iov_off_rd = 0; iov_slot++; if (iov_slot == array_size(iov)) { iov_slot = 0; ret = readv(fd, iov, array_size(iov)); if (unlikely(ret <= 0)) return -EIO; } fmemcpy(packet + offset, iov[iov_slot].iov_base + iov_off_rd, remainder); iov_off_rd += remainder; return hdrlen; } static ssize_t pcap_sg_read(int fd, pcap_pkthdr_t *phdr, enum pcap_type type, uint8_t *packet, size_t len) { ssize_t ret = 0; size_t hdrsize = pcap_get_hdr_length(phdr, type), hdrlen; if (likely(iov[iov_slot].iov_len - iov_off_rd >= hdrsize)) { fmemcpy(&phdr->raw, iov[iov_slot].iov_base + iov_off_rd, hdrsize); iov_off_rd += hdrsize; } else { ret = __pcap_sg_inter_iov_hdr_read(fd, phdr, hdrsize); if (unlikely(ret < 0)) return ret; } hdrlen = pcap_get_length(phdr, type); if (unlikely(hdrlen == 0 || hdrlen > len)) return -EINVAL; if (likely(iov[iov_slot].iov_len - iov_off_rd >= hdrlen)) { fmemcpy(packet, iov[iov_slot].iov_base + iov_off_rd, hdrlen); iov_off_rd += hdrlen; } else { ret = __pcap_sg_inter_iov_data_read(fd, packet, hdrlen); if (unlikely(ret < 0)) return ret; } return hdrsize + hdrlen; } static void pcap_sg_fsync(int fd) { ssize_t ret = writev(fd, iov, iov_slot); if (ret < 0) panic("Writev I/O error: %s!\n", strerror(errno)); iov_slot = 0; fdatasync(fd); } static void pcap_sg_init_once(void) { set_ioprio_rt(); } static int pcap_sg_prepare_access(int fd, enum pcap_mode mode, bool jumbo) { int ret; size_t i, len = 0; iov_slot = 0; len = jumbo ? (PAGE_SIZE * 16) /* 64k max */ : (PAGE_SIZE * 3) /* 12k max */; for (i = 0; i < array_size(iov); ++i) { iov[i].iov_base = xzmalloc_aligned(len, 64); iov[i].iov_len = len; } if (mode == PCAP_MODE_RD) { ret = readv(fd, iov, array_size(iov)); if (ret <= 0) return -EIO; iov_off_rd = 0; iov_slot = 0; } return 0; } static void pcap_sg_prepare_close(int fd __maybe_unused, enum pcap_mode mode __maybe_unused) { size_t i; for (i = 0; i < array_size(iov); ++i) xfree(iov[i].iov_base); } const struct pcap_file_ops pcap_sg_ops = { .init_once_pcap = pcap_sg_init_once, .pull_fhdr_pcap = pcap_generic_pull_fhdr, .push_fhdr_pcap = pcap_generic_push_fhdr, .prepare_access_pcap = pcap_sg_prepare_access, .prepare_close_pcap = pcap_sg_prepare_close, .read_pcap = pcap_sg_read, .write_pcap = pcap_sg_write, .fsync_pcap = pcap_sg_fsync, }; commit05c974669ecec510a85d8534099bb75404e82c41 (patch) tree8f35a5f3dfa7f9b98781c1026e17b26c5e716097 parent78d506e1b7071b24850fd5ac22b896c459b0a04c (diff)
xprtrdma: Fix receive buffer accounting
An RPC can terminate before its reply arrives, if a credential problem or a soft timeout occurs. After this happens, xprtrdma reports it is out of Receive buffers. A Receive buffer is posted before each RPC is sent, and returned to the buffer pool when a reply is received. If no reply is received for an RPC, that Receive buffer remains posted. But xprtrdma tries to post another when the next RPC is sent. If this happens a few dozen times, there are no receive buffers left to be posted at send time. I don't see a way for a transport connection to recover at that point, and it will spit warnings and unnecessarily delay RPCs on occasion for its remaining lifetime. Commit 1e465fd4ff47 ("xprtrdma: Replace send and receive arrays") removed a little bit of logic to detect this case and not provide a Receive buffer so no more buffers are posted, and then transport operation continues correctly. We didn't understand what that logic did, and it wasn't commented, so it was removed as part of the overhaul to support backchannel requests. Restore it, but be wary of the need to keep extra Receives posted to deal with backchannel requests. Fixes: 1e465fd4ff47 ("xprtrdma: Replace send and receive arrays") Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Reviewed-by: Anna Schumaker <Anna.Schumaker@Netapp.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
Diffstat