On Thu, Aug 10, 2017 at 04:57:08PM +0800, Zorro Lang wrote: > When mixing buffered reads and asynchronous direct writes, it is > possible to end up with the situation where we have stale data in > the page cache while the new data is already written to disk. Hi, thanks for writing the test. Couple of comments below. > > Signed-off-by: Zorro Lang <zlang@xxxxxxxxxx> > --- > > Hi, > > This case add a new C program into src/aio-dio-regress/. I tried to > use some existent program to make some AIO load to trigger this bug, > but I found it's very hard. > > For trigger this bug ASAP, I had to write a new program. And from > the my test results, it really can reproduce this bug quickly. > > More details about this bug please check: > https://patchwork.kernel.org/patch/9851671/ > > Thanks, > Zorro > > .gitignore | 1 + > src/aio-dio-regress/aio-dio-cycle-write.c | 254 ++++++++++++++++++++++++++++++ > tests/generic/999 | 74 +++++++++ > tests/generic/999.out | 2 + > tests/generic/group | 1 + > 5 files changed, 332 insertions(+) > create mode 100644 src/aio-dio-regress/aio-dio-cycle-write.c > create mode 100755 tests/generic/999 > create mode 100644 tests/generic/999.out > > diff --git a/.gitignore b/.gitignore > index fcbc0cd4..28fe84d5 100644 > --- a/.gitignore > +++ b/.gitignore > @@ -133,6 +133,7 @@ > /src/writemod > /src/writev_on_pagefault > /src/xfsctl > +/src/aio-dio-regress/aio-dio-cycle-write > /src/aio-dio-regress/aio-dio-extend-stat > /src/aio-dio-regress/aio-dio-fcntl-race > /src/aio-dio-regress/aio-dio-hole-filling-race > diff --git a/src/aio-dio-regress/aio-dio-cycle-write.c b/src/aio-dio-regress/aio-dio-cycle-write.c > new file mode 100644 > index 00000000..6c10f9cc > --- /dev/null > +++ b/src/aio-dio-regress/aio-dio-cycle-write.c > @@ -0,0 +1,254 @@ > +/* > + * Directly AIO re-write a file with different content again and again. > + * And check the data integrity. > + * > + * Copyright (C) 2017 Red Hat, Inc. All Rights reserved. > + * > + * This program is free software; you can redistribute it and/or modify > + * it under the terms of the GNU General Public License as published by > + * the Free Software Foundation; either version 2 of the License, or > + * (at your option) any later version. > + * > + * This program is distributed in the hope that it will be useful, > + * but WITHOUT ANY WARRANTY; without even the implied warranty of > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the > + * GNU General Public License for more details. > + * > + * You should have received a copy of the GNU General Public License > + * along with this program; if not, write to the Free Software > + * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA. > + */ > + > +#include <sys/stat.h> > +#include <sys/types.h> > +#include <errno.h> > +#include <fcntl.h> > +#include <stdio.h> > +#include <stdlib.h> > +#include <unistd.h> > +#include <ctype.h> > + > +#include <libaio.h> > + > +unsigned long buf_size = 0; > +unsigned long loop_count = 0; > +void *test_buf[2]; > +#define IO_PATTERN1 0x55 > +#define IO_PATTERN2 0xaa > + > +void usage(char *progname) > +{ > + fprintf(stderr, "usage: %s [-c loop_count] [-b bufsize] filename\n" > + "\t-c loopcount: specify how many times to test" > + "\t-b bufsize: keep writing from offset 0 to this size", > + progname); > + exit(1); > +} > + > +void dump_buffer( > + void *buf, > + off64_t offset, > + ssize_t len) > +{ > + int i, j; > + char *p; > + int new; > + > + for (i = 0, p = (char *)buf; i < len; i += 16) { > + char *s = p; > + > + if (i && !memcmp(p, p - 16, 16)) { > + new = 0; > + } else { > + if (i) > + printf("*\n"); > + new = 1; > + } > + > + if (!new) { > + p += 16; > + continue; > + } > + > + printf("%08llx ", (unsigned long long)offset + i); > + for (j = 0; j < 16 && i + j < len; j++, p++) > + printf("%02x ", *p); > + printf(" "); > + for (j = 0; j < 16 && i + j < len; j++, s++) { > + if (isalnum((int)*s)) > + printf("%c", *s); > + else > + printf("."); > + } > + printf("\n"); > + > + } > + printf("%08llx\n", (unsigned long long)offset + i); > +} > + > +int init_test(char *filename) > +{ > + int fd; > + int err = 0; > + > + fd = open(filename, O_DIRECT | O_CREAT | O_RDWR, 0600); > + if (fd == -1) { > + perror("open"); > + exit(1); > + } > + > + ftruncate(fd, buf_size); > + > + /* fill test_buf[0] with IO_PATTERN1 */ > + err = posix_memalign(&(test_buf[0]), getpagesize(), buf_size); > + if (err) { > + fprintf(stderr, "error %s during %s\n", > + strerror(err), > + "posix_memalign"); > + exit(1); > + } > + memset(test_buf[0], IO_PATTERN1, buf_size); > + > + /* fill test_buf[1] with IO_PATTERN2 */ > + err = posix_memalign(&(test_buf[1]), getpagesize(), buf_size); > + if (err) { > + fprintf(stderr, "error %s during %s\n", > + strerror(err), > + "posix_memalign"); > + exit(1); > + } > + memset(test_buf[1], IO_PATTERN2, buf_size); > + > + /* fill test file with IO_PATTERN1 */ > + if (pwrite(fd, test_buf[0], buf_size, 0) != buf_size) { > + perror("pwrite"); > + exit(1); > + } > + > + fsync(fd); > + close(fd); > + > + return 0; > +} > + > +/* > + * Read file content back, then compare with the 'buf' which > + * point to the original buffer was written to file. > + */ > +int fs_check(char *filename, void *buf) > +{ > + void *cmp_buf; > + int fd; > + > + cmp_buf = malloc(buf_size); > + if (!cmp_buf) { > + perror("malloc"); > + exit(1); > + } > + memset(cmp_buf, 0, buf_size); Do you have to do the allocation with every iteration ? Seems like a waste to me. Also you forgot to free the memory so eventually you'll run dry. It would be better to change it so you only allocate it once. > + > + fd = open(filename, O_RDONLY, 0600); > + if (fd == -1) { > + perror("open"); > + exit(1); > + } > + > + if (pread(fd, cmp_buf, buf_size, 0) != buf_size) { > + perror("pread"); > + exit(1); > + } > + close(fd); > + > + if (memcmp(buf, cmp_buf, buf_size)) { > + printf("corruption while extending\n"); Not sure why it's "corruption while extending" ? > + dump_buffer(buf, 0, buf_size); > + exit(1); First thing I'd do after the test fails would be to drop the caches and see whether the content changes. That would confirm that the problem is indeed page cache inconsistency. So maybe it would be nice to confirm it right away within the test itself ? > + } > + > + return 0; > +} > + > +int main(int argc, char *argv[]) > +{ > + struct io_context *ctx = NULL; > + struct io_event evs[1]; > + struct iocb iocb1; > + struct iocb *iocbs[] = { &iocb1 }; > + int fd, err = 0; > + int i; > + int c; > + char *filename = NULL; > + > + while ((c = getopt(argc, argv, "c:b:")) != -1) { > + char *endp; > + > + switch (c) { > + case 'c': /* the number of testing cycles */ > + loop_count = strtol(optarg, &endp, 0); > + break; > + case 'b': /* buffer size */ > + buf_size = strtol(optarg, &endp, 0); > + break; > + default: > + usage(argv[0]); > + } > + } > + > + if (loop_count == 0) > + loop_count = 1600; > + if (buf_size == 0) /* default minimum buffer size is 65536 bytes */ > + buf_size = 65536; > + > + if (optind == argc - 1) > + filename = argv[optind]; > + else > + usage(argv[0]); > + > + init_test(filename); > + > + err = io_setup(1, &ctx); > + if (err) { > + fprintf(stderr, "error %s during %s\n", > + strerror(err), > + "io_setup"); > + return 1; > + } > + > + i = 0; > + /* > + * Keep running until loop_count times, fill the file with IO_PATTERN1 > + * or IO_PATTERN2 one by one, then read the file data back to check if > + * there's stale data. > + */ > + while (loop_count--) { > + i++; > + i %= 2; > + fd = open(filename, O_DIRECT | O_CREAT | O_RDWR, 0600); > + if (fd == -1) { > + perror("open"); > + return 1; > + } > + > + io_prep_pwrite(&iocb1, fd, test_buf[i], buf_size, 0); > + err = io_submit(ctx, 1, iocbs); > + if (err != 1) { > + fprintf(stderr, "error %s during %s\n", > + strerror(err), > + "io_submit"); > + return 1; > + } > + err = io_getevents(ctx, 1, 1, evs, NULL); > + if (err != 1) { > + fprintf(stderr, "error %s during %s\n", > + strerror(err), > + "io_getevents"); > + return 1; > + } > + close(fd); > + > + fs_check(filename, test_buf[i]); > + } > + > + printf("Success, all done.\n"); > + return 0; > +} > diff --git a/tests/generic/999 b/tests/generic/999 > new file mode 100755 > index 00000000..006a7c92 > --- /dev/null > +++ b/tests/generic/999 > @@ -0,0 +1,74 @@ > +#! /bin/bash > +# FS QA Test No. 999 > +# > +# Test data integrity when mixing buffered reads and asynchronous > +# direct writes a file. > +# > +#----------------------------------------------------------------------- > +# Copyright (c) 2017 Red Hat Inc. All Rights Reserved. > +# > +# This program is free software; you can redistribute it and/or > +# modify it under the terms of the GNU General Public License as > +# published by the Free Software Foundation. > +# > +# This program is distributed in the hope that it would be useful, > +# but WITHOUT ANY WARRANTY; without even the implied warranty of > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the > +# GNU General Public License for more details. > +# > +# You should have received a copy of the GNU General Public License > +# along with this program; if not, write the Free Software Foundation, > +# Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA > +#----------------------------------------------------------------------- > +# > + > +seq=`basename $0` > +seqres=$RESULT_DIR/$seq > +echo "QA output created by $seq" > + > +here=`pwd` > +tmp=/tmp/$$ > +status=1 # failure is the default! > +trap "_cleanup; exit \$status" 0 1 2 3 15 > + > +_cleanup() > +{ > + cd / > + rm -f $tmp.* > +} > + > +# get standard environment, filters and checks > +. ./common/rc > + > +# remove previous $seqres.full before test > +rm -f $seqres.full > + > +# real QA test starts here > +_supported_fs generic > +_supported_os Linux > +_require_test > +_require_test_program "feature" > +_require_aiodio aio-dio-cycle-write > + > +TESTFILE=$TEST_DIR/tst-aio-dio-cycle-write.$seq > +FSIZE=655360 # bytes > + > +# buffered reads the file frequently > +while true; do > + $XFS_IO_PROG -f -c "pread 0 $FSIZE" $TESTFILE >/dev/null 2>&1 > +done & > +reader_pid=$! > + > +nr_cpu=`$here/src/feature -o` > +LOOPCOUNT=$((nr_cpu * 100)) > +if [ $LOOPCOUNT -gt 1000 ]; then > + LOOPCOUNT=1000 > +fi Unfortunatelly this did not work for me. I had to increase the multiplicator to 1000 (with 4 CPUs I had 4000 iterations) otherwise I was not able to hit it. Those 4000 iterations in my case took about 16s which would be perfectly fine for me. Maybe running the test for predetermined length of time might be a better idea ? Also I am not sure why nr_cpu matters here ? Thanks! -Lukas > +# start a background aio writer, which does several writing loops > +# internally and check data integrality > +$AIO_TEST -c $LOOPCOUNT -b $FSIZE $TESTFILE > +status=$? > + > +kill $reader_pid > +wait $reader_pid > +exit > diff --git a/tests/generic/999.out b/tests/generic/999.out > new file mode 100644 > index 00000000..6362b866 > --- /dev/null > +++ b/tests/generic/999.out > @@ -0,0 +1,2 @@ > +QA output created by 999 > +Success, all done. > diff --git a/tests/generic/group b/tests/generic/group > index e13b5683..08462eee 100644 > --- a/tests/generic/group > +++ b/tests/generic/group > @@ -452,3 +452,4 @@ > 447 auto quick clone > 448 auto quick rw > 449 auto quick acl enospc > +999 auto aio rw > -- > 2.13.4 > -- To unsubscribe from this list: send the line "unsubscribe fstests" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
