| Message ID | 2f353cb69600740308c196dcfa8f570ba6420bfa.1591087232.git.jstancek@redhat.com |
|---|---|
| State | Accepted, archived |
| Headers | show |
| Series | [v2] syscalls/ioctl_loop05: set loop blksize to bdev blksize | expand |
Hi Jan Looks good to me, acked. > Test is failing on s390, where default loop blksize is less than > backing dev's blksize (4096): > tst_test.c:1247: INFO: Timeout per run is 0h 05m 00s > tst_device.c:88: INFO: Found free device 0 '/dev/loop0' > ioctl_loop05.c:116: INFO: /dev/loop0 default logical_block_size is 512 > ioctl_loop05.c:62: INFO: Without setting lo_offset or sizelimit > ioctl_loop05.c:63: BROK: ioctl(3,LOOP_SET_DIRECT_IO,...) failed: EINVAL (22) > > Per kernel comment at __loop_update_dio(), direct io is supported > when ".. logical block size of loop is bigger than the backing device's". > > Set loop blksize to one of backing device. Retry is there to avoid > EAGAIN warning "loop0 (test.img) has still dirty pages". > > Signed-off-by: Jan Stancek <jstancek@redhat.com> > --- > .../kernel/syscalls/ioctl/ioctl_loop05.c | 26 +++++++++++++++++-- > 1 file changed, 24 insertions(+), 2 deletions(-) > > Changes in v2: > - check if LOOP_SET_BLOCK_SIZE is supported in setup() > > diff --git a/testcases/kernel/syscalls/ioctl/ioctl_loop05.c b/testcases/kernel/syscalls/ioctl/ioctl_loop05.c > index 6c9ea2802981..2a3c127959aa 100644 > --- a/testcases/kernel/syscalls/ioctl/ioctl_loop05.c > +++ b/testcases/kernel/syscalls/ioctl/ioctl_loop05.c > @@ -96,6 +96,9 @@ static void verify_ioctl_loop(void) > > static void setup(void) > { > + int fd; > + struct stat buf; > + > if (tst_fs_type(".") == TST_TMPFS_MAGIC) > tst_brk(TCONF, "tmpfd doesn't support O_DIRECT flag"); > > @@ -105,6 +108,14 @@ static void setup(void) > > sprintf(sys_loop_diopath, "/sys/block/loop%d/loop/dio", dev_num); > tst_fill_file("test.img", 0, 1024, 1024); > + > + fd = SAFE_OPEN("test.img", O_RDONLY); > + SAFE_FSTAT(fd, &buf); > + SAFE_CLOSE(fd); > + > + logical_block_size = buf.st_blksize; > + tst_res(TINFO, "backing dev logical_block_size is %d", logical_block_size); > + > tst_attach_device(dev_path, "test.img"); > attach_flag = 1; > dev_fd = SAFE_OPEN(dev_path, O_RDWR); > @@ -112,8 +123,19 @@ static void setup(void) > if (ioctl(dev_fd, LOOP_SET_DIRECT_IO, 0) && errno == EINVAL) > tst_brk(TCONF, "LOOP_SET_DIRECT_IO is not supported"); > > - SAFE_IOCTL(dev_fd, BLKSSZGET, &logical_block_size); > - tst_res(TINFO, "%s default logical_block_size is %d", dev_path, logical_block_size); > + /* > + * from __loop_update_dio(): > + * We support direct I/O only if lo_offset is aligned with the > + * logical I/O size of backing device, and the logical block > + * size of loop is bigger than the backing device's and the loop > + * needn't transform transfer. > + */ > + if (ioctl(dev_fd, LOOP_SET_BLOCK_SIZE, logical_block_size) != 0) { > + if (errno == EINVAL) > + tst_brk(TCONF, "LOOP_SET_BLOCK_SIZE is not supported"); > + TST_RETRY_FUNC(ioctl(dev_fd, LOOP_SET_BLOCK_SIZE, > + logical_block_size), TST_RETVAL_EQ0); > + } > } > > static void cleanup(void) >
----- Original Message ----- > Hi Jan > > Looks good to me, acked. Pushed.
diff --git a/testcases/kernel/syscalls/ioctl/ioctl_loop05.c b/testcases/kernel/syscalls/ioctl/ioctl_loop05.c index 6c9ea2802981..2a3c127959aa 100644 --- a/testcases/kernel/syscalls/ioctl/ioctl_loop05.c +++ b/testcases/kernel/syscalls/ioctl/ioctl_loop05.c @@ -96,6 +96,9 @@ static void verify_ioctl_loop(void) static void setup(void) { + int fd; + struct stat buf; + if (tst_fs_type(".") == TST_TMPFS_MAGIC) tst_brk(TCONF, "tmpfd doesn't support O_DIRECT flag"); @@ -105,6 +108,14 @@ static void setup(void) sprintf(sys_loop_diopath, "/sys/block/loop%d/loop/dio", dev_num); tst_fill_file("test.img", 0, 1024, 1024); + + fd = SAFE_OPEN("test.img", O_RDONLY); + SAFE_FSTAT(fd, &buf); + SAFE_CLOSE(fd); + + logical_block_size = buf.st_blksize; + tst_res(TINFO, "backing dev logical_block_size is %d", logical_block_size); + tst_attach_device(dev_path, "test.img"); attach_flag = 1; dev_fd = SAFE_OPEN(dev_path, O_RDWR); @@ -112,8 +123,19 @@ static void setup(void) if (ioctl(dev_fd, LOOP_SET_DIRECT_IO, 0) && errno == EINVAL) tst_brk(TCONF, "LOOP_SET_DIRECT_IO is not supported"); - SAFE_IOCTL(dev_fd, BLKSSZGET, &logical_block_size); - tst_res(TINFO, "%s default logical_block_size is %d", dev_path, logical_block_size); + /* + * from __loop_update_dio(): + * We support direct I/O only if lo_offset is aligned with the + * logical I/O size of backing device, and the logical block + * size of loop is bigger than the backing device's and the loop + * needn't transform transfer. + */ + if (ioctl(dev_fd, LOOP_SET_BLOCK_SIZE, logical_block_size) != 0) { + if (errno == EINVAL) + tst_brk(TCONF, "LOOP_SET_BLOCK_SIZE is not supported"); + TST_RETRY_FUNC(ioctl(dev_fd, LOOP_SET_BLOCK_SIZE, + logical_block_size), TST_RETVAL_EQ0); + } } static void cleanup(void)
Test is failing on s390, where default loop blksize is less than backing dev's blksize (4096): tst_test.c:1247: INFO: Timeout per run is 0h 05m 00s tst_device.c:88: INFO: Found free device 0 '/dev/loop0' ioctl_loop05.c:116: INFO: /dev/loop0 default logical_block_size is 512 ioctl_loop05.c:62: INFO: Without setting lo_offset or sizelimit ioctl_loop05.c:63: BROK: ioctl(3,LOOP_SET_DIRECT_IO,...) failed: EINVAL (22) Per kernel comment at __loop_update_dio(), direct io is supported when ".. logical block size of loop is bigger than the backing device's". Set loop blksize to one of backing device. Retry is there to avoid EAGAIN warning "loop0 (test.img) has still dirty pages". Signed-off-by: Jan Stancek <jstancek@redhat.com> --- .../kernel/syscalls/ioctl/ioctl_loop05.c | 26 +++++++++++++++++-- 1 file changed, 24 insertions(+), 2 deletions(-) Changes in v2: - check if LOOP_SET_BLOCK_SIZE is supported in setup()