| Message ID | 20240828181534.41054-4-bfoster@redhat.com (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Headers | show |
| Series | fstests/fsx: test coverage for eof zeroing | expand |
On Wed, Aug 28, 2024 at 02:15:33PM -0400, Brian Foster wrote: > File ranges that are newly exposed via size changing operations are > expected to return zeroes until written to. This behavior tends to > be difficult to regression test as failures can be racy and > transient. fsx is probably the best tool for this type of test > coverage, but uncovering issues can require running for a > significantly longer period of time than is typically invoked > through fstests tests. As a result, these types of regressions tend > to go unnoticed for an unfortunate amount of time. > > To facilitate uncovering these problems more quickly, implement an > eof pollution mode in fsx that opportunistically injects post-eof > data prior to operations that change file size. Since data injection > occurs immediately before the size changing operation, it can be > used to detect problems in partial eof page/block zeroing associated > with each relevant operation. > > The implementation takes advantage of the fact that mapped writes > can place data beyond eof so long as the page starts within eof. The > main reason for the isolated per-operation approach (vs. something > like allowing mapped writes to write beyond eof, for example) is to > accommodate the fact that writeback zeroes post-eof data on the eof > page. The current approach is therefore not necessarily guaranteed > to detect all problems, but provides more generic and broad test > coverage than the alternative of testing explicit command sequences > and doesn't require significant changes to how fsx works. If this > proves useful long term, further enhancements can be considered that > might facilitate the presence of post-eof data across operations. > > Enable the feature with the -e command line option. It is disabled > by default because zeroing behavior is inconsistent across > filesystems. This can also be revisited in the future if zeroing > behavior is refined for the major filesystems that rely on fstests > for regression testing. > > Signed-off-by: Brian Foster <bfoster@redhat.com> I wonder what insanity this is going to shake loose, so thanks for doing putting this in a stressor program. :) Reviewed-by: Darrick J. Wong <djwong@kernel.org> --D > --- > ltp/fsx.c | 79 +++++++++++++++++++++++++++++++++++++++++++++++++++++-- > 1 file changed, 77 insertions(+), 2 deletions(-) > > diff --git a/ltp/fsx.c b/ltp/fsx.c > index 4a9be0e1..1ba1bf65 100644 > --- a/ltp/fsx.c > +++ b/ltp/fsx.c > @@ -178,6 +178,7 @@ int dedupe_range_calls = 1; /* -B flag disables */ > int copy_range_calls = 1; /* -E flag disables */ > int exchange_range_calls = 1; /* -0 flag disables */ > int integrity = 0; /* -i flag */ > +int pollute_eof = 0; /* -e flag */ > int fsxgoodfd = 0; > int o_direct; /* -Z */ > int aio = 0; > @@ -983,6 +984,63 @@ gendata(char *original_buf, char *good_buf, unsigned offset, unsigned size) > } > } > > +/* > + * Pollute the EOF page with data beyond EOF prior to size change operations. > + * This provides additional test coverage for partial EOF block/page zeroing. > + * If the upcoming operation does not correctly zero, incorrect file data will > + * be detected. > + */ > +void > +pollute_eofpage(unsigned int maxoff) > +{ > + unsigned offset = file_size; > + unsigned pg_offset; > + unsigned write_size; > + char *p; > + > + /* > + * Nothing to do if pollution disabled or we're simulating. Simulation > + * only tracks file size updates and skips syscalls, so we don't want to > + * inject file data that won't be zeroed. > + */ > + if (!pollute_eof || testcalls <= simulatedopcount) > + return; > + > + /* write up to specified max or the end of the eof page */ > + pg_offset = offset & mmap_mask; > + write_size = MIN(PAGE_SIZE - pg_offset, maxoff - offset); > + > + if (!pg_offset) > + return; > + > + if (!quiet && > + ((progressinterval && testcalls % progressinterval == 0) || > + (debug && > + (monitorstart == -1 || > + (offset + write_size > monitorstart && > + (monitorend == -1 || offset <= monitorend)))))) { > + prt("%lld pollute_eof\t0x%x thru\t0x%x\t(0x%x bytes)\n", > + testcalls, offset, offset + write_size - 1, write_size); > + } > + > + if ((p = (char *)mmap(0, PAGE_SIZE, PROT_READ | PROT_WRITE, > + MAP_FILE | MAP_SHARED, fd, > + (off_t)(offset - pg_offset))) == MAP_FAILED) { > + prterr("pollute_eofpage: mmap"); > + return; > + } > + > + /* > + * Write to a range just past EOF of the test file. Do not update the > + * good buffer because the upcoming operation is expected to zero this > + * range of the file. > + */ > + gendata(original_buf, p, pg_offset, write_size); > + > + if (munmap(p, PAGE_SIZE) != 0) > + prterr("pollute_eofpage: munmap"); > +} > + > /* > * Helper to update the tracked file size. If the offset begins beyond current > * EOF, zero the range from EOF to offset in the good buffer. > @@ -990,8 +1048,10 @@ gendata(char *original_buf, char *good_buf, unsigned offset, unsigned size) > void > update_file_size(unsigned offset, unsigned size) > { > - if (offset > file_size) > + if (offset > file_size) { > + pollute_eofpage(offset + size); > memset(good_buf + file_size, '\0', offset - file_size); > + } > file_size = offset + size; > } > > @@ -1143,6 +1203,9 @@ dotruncate(unsigned size) > > log4(OP_TRUNCATE, 0, size, FL_NONE); > > + /* pollute the current EOF before a truncate down */ > + if (size < file_size) > + pollute_eofpage(maxfilelen); > update_file_size(size, 0); > > if (testcalls <= simulatedopcount) > @@ -1305,6 +1368,9 @@ do_collapse_range(unsigned offset, unsigned length) > > log4(OP_COLLAPSE_RANGE, offset, length, FL_NONE); > > + /* pollute current eof before collapse truncates down */ > + pollute_eofpage(maxfilelen); > + > if (testcalls <= simulatedopcount) > return; > > @@ -1356,6 +1422,9 @@ do_insert_range(unsigned offset, unsigned length) > > log4(OP_INSERT_RANGE, offset, length, FL_NONE); > > + /* pollute current eof before insert truncates up */ > + pollute_eofpage(maxfilelen); > + > if (testcalls <= simulatedopcount) > return; > > @@ -2385,6 +2454,7 @@ usage(void) > -b opnum: beginning operation number (default 1)\n\ > -c P: 1 in P chance of file close+open at each op (default infinity)\n\ > -d: debug output for all operations\n\ > + -e: pollute post-eof on size changes (default 0)\n\ > -f: flush and invalidate cache after I/O\n\ > -g X: write character X instead of random generated data\n\ > -i logdev: do integrity testing, logdev is the dm log writes device\n\ > @@ -2783,7 +2853,7 @@ main(int argc, char **argv) > setvbuf(stdout, (char *)0, _IOLBF, 0); /* line buffered stdout */ > > while ((ch = getopt_long(argc, argv, > - "0b:c:dfg:i:j:kl:m:no:p:qr:s:t:w:xyABD:EFJKHzCILN:OP:RS:UWXZ", > + "0b:c:de:fg:i:j:kl:m:no:p:qr:s:t:w:xyABD:EFJKHzCILN:OP:RS:UWXZ", > longopts, NULL)) != EOF) > switch (ch) { > case 'b': > @@ -2805,6 +2875,11 @@ main(int argc, char **argv) > case 'd': > debug = 1; > break; > + case 'e': > + pollute_eof = getnum(optarg, &endp); > + if (pollute_eof < 0 || pollute_eof > 1) > + usage(); > + break; > case 'f': > flush = 1; > break; > -- > 2.45.0 > >
diff --git a/ltp/fsx.c b/ltp/fsx.c index 4a9be0e1..1ba1bf65 100644 --- a/ltp/fsx.c +++ b/ltp/fsx.c @@ -178,6 +178,7 @@ int dedupe_range_calls = 1; /* -B flag disables */ int copy_range_calls = 1; /* -E flag disables */ int exchange_range_calls = 1; /* -0 flag disables */ int integrity = 0; /* -i flag */ +int pollute_eof = 0; /* -e flag */ int fsxgoodfd = 0; int o_direct; /* -Z */ int aio = 0; @@ -983,6 +984,63 @@ gendata(char *original_buf, char *good_buf, unsigned offset, unsigned size) } } +/* + * Pollute the EOF page with data beyond EOF prior to size change operations. + * This provides additional test coverage for partial EOF block/page zeroing. + * If the upcoming operation does not correctly zero, incorrect file data will + * be detected. + */ +void +pollute_eofpage(unsigned int maxoff) +{ + unsigned offset = file_size; + unsigned pg_offset; + unsigned write_size; + char *p; + + /* + * Nothing to do if pollution disabled or we're simulating. Simulation + * only tracks file size updates and skips syscalls, so we don't want to + * inject file data that won't be zeroed. + */ + if (!pollute_eof || testcalls <= simulatedopcount) + return; + + /* write up to specified max or the end of the eof page */ + pg_offset = offset & mmap_mask; + write_size = MIN(PAGE_SIZE - pg_offset, maxoff - offset); + + if (!pg_offset) + return; + + if (!quiet && + ((progressinterval && testcalls % progressinterval == 0) || + (debug && + (monitorstart == -1 || + (offset + write_size > monitorstart && + (monitorend == -1 || offset <= monitorend)))))) { + prt("%lld pollute_eof\t0x%x thru\t0x%x\t(0x%x bytes)\n", + testcalls, offset, offset + write_size - 1, write_size); + } + + if ((p = (char *)mmap(0, PAGE_SIZE, PROT_READ | PROT_WRITE, + MAP_FILE | MAP_SHARED, fd, + (off_t)(offset - pg_offset))) == MAP_FAILED) { + prterr("pollute_eofpage: mmap"); + return; + } + + /* + * Write to a range just past EOF of the test file. Do not update the + * good buffer because the upcoming operation is expected to zero this + * range of the file. + */ + gendata(original_buf, p, pg_offset, write_size); + + if (munmap(p, PAGE_SIZE) != 0) + prterr("pollute_eofpage: munmap"); +} + /* * Helper to update the tracked file size. If the offset begins beyond current * EOF, zero the range from EOF to offset in the good buffer. @@ -990,8 +1048,10 @@ gendata(char *original_buf, char *good_buf, unsigned offset, unsigned size) void update_file_size(unsigned offset, unsigned size) { - if (offset > file_size) + if (offset > file_size) { + pollute_eofpage(offset + size); memset(good_buf + file_size, '\0', offset - file_size); + } file_size = offset + size; } @@ -1143,6 +1203,9 @@ dotruncate(unsigned size) log4(OP_TRUNCATE, 0, size, FL_NONE); + /* pollute the current EOF before a truncate down */ + if (size < file_size) + pollute_eofpage(maxfilelen); update_file_size(size, 0); if (testcalls <= simulatedopcount) @@ -1305,6 +1368,9 @@ do_collapse_range(unsigned offset, unsigned length) log4(OP_COLLAPSE_RANGE, offset, length, FL_NONE); + /* pollute current eof before collapse truncates down */ + pollute_eofpage(maxfilelen); + if (testcalls <= simulatedopcount) return; @@ -1356,6 +1422,9 @@ do_insert_range(unsigned offset, unsigned length) log4(OP_INSERT_RANGE, offset, length, FL_NONE); + /* pollute current eof before insert truncates up */ + pollute_eofpage(maxfilelen); + if (testcalls <= simulatedopcount) return; @@ -2385,6 +2454,7 @@ usage(void) -b opnum: beginning operation number (default 1)\n\ -c P: 1 in P chance of file close+open at each op (default infinity)\n\ -d: debug output for all operations\n\ + -e: pollute post-eof on size changes (default 0)\n\ -f: flush and invalidate cache after I/O\n\ -g X: write character X instead of random generated data\n\ -i logdev: do integrity testing, logdev is the dm log writes device\n\ @@ -2783,7 +2853,7 @@ main(int argc, char **argv) setvbuf(stdout, (char *)0, _IOLBF, 0); /* line buffered stdout */ while ((ch = getopt_long(argc, argv, - "0b:c:dfg:i:j:kl:m:no:p:qr:s:t:w:xyABD:EFJKHzCILN:OP:RS:UWXZ", + "0b:c:de:fg:i:j:kl:m:no:p:qr:s:t:w:xyABD:EFJKHzCILN:OP:RS:UWXZ", longopts, NULL)) != EOF) switch (ch) { case 'b': @@ -2805,6 +2875,11 @@ main(int argc, char **argv) case 'd': debug = 1; break; + case 'e': + pollute_eof = getnum(optarg, &endp); + if (pollute_eof < 0 || pollute_eof > 1) + usage(); + break; case 'f': flush = 1; break;
File ranges that are newly exposed via size changing operations are expected to return zeroes until written to. This behavior tends to be difficult to regression test as failures can be racy and transient. fsx is probably the best tool for this type of test coverage, but uncovering issues can require running for a significantly longer period of time than is typically invoked through fstests tests. As a result, these types of regressions tend to go unnoticed for an unfortunate amount of time. To facilitate uncovering these problems more quickly, implement an eof pollution mode in fsx that opportunistically injects post-eof data prior to operations that change file size. Since data injection occurs immediately before the size changing operation, it can be used to detect problems in partial eof page/block zeroing associated with each relevant operation. The implementation takes advantage of the fact that mapped writes can place data beyond eof so long as the page starts within eof. The main reason for the isolated per-operation approach (vs. something like allowing mapped writes to write beyond eof, for example) is to accommodate the fact that writeback zeroes post-eof data on the eof page. The current approach is therefore not necessarily guaranteed to detect all problems, but provides more generic and broad test coverage than the alternative of testing explicit command sequences and doesn't require significant changes to how fsx works. If this proves useful long term, further enhancements can be considered that might facilitate the presence of post-eof data across operations. Enable the feature with the -e command line option. It is disabled by default because zeroing behavior is inconsistent across filesystems. This can also be revisited in the future if zeroing behavior is refined for the major filesystems that rely on fstests for regression testing. Signed-off-by: Brian Foster <bfoster@redhat.com> --- ltp/fsx.c | 79 +++++++++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 77 insertions(+), 2 deletions(-)