| Message ID | 20200825160735.24602-4-mdoucha@suse.cz |
|---|---|
| State | Accepted |
| Headers | show |
| Series | Improve reliability of ioctl_sg01 | expand |
Hi Martin, Reviewed-by: Petr Vorel <pvorel@suse.cz> Kind regards, Petr > The test wasn't reliable if most of available memory was full of zeroes. > Pollute free memory to increase the chance of detecting data leak. > Signed-off-by: Martin Doucha <mdoucha@suse.cz> > --- > Changes since v1: > - Split patch > - Use tst_pollute_memory() instead of allocating and pre-polluting > a fixed-size block of memory in setup(). > testcases/kernel/syscalls/ioctl/ioctl_sg01.c | 11 +++++++---- > 1 file changed, 7 insertions(+), 4 deletions(-) > diff --git a/testcases/kernel/syscalls/ioctl/ioctl_sg01.c b/testcases/kernel/syscalls/ioctl/ioctl_sg01.c > index daaa96be5..8c9fd0dae 100644 > --- a/testcases/kernel/syscalls/ioctl/ioctl_sg01.c > +++ b/testcases/kernel/syscalls/ioctl/ioctl_sg01.c > @@ -7,9 +7,7 @@ > * CVE-2018-1000204 > * > * Test ioctl(SG_IO) and check that kernel doesn't leak data. Requires > - * a read-accessible SCSI-compatible device (e.g. SATA disk). Running oom* > - * test program before this one may increase the chance of successfully > - * reproducing the bug. > + * a read-accessible generic SCSI device (e.g. a DVD drive). > * > * Leak fixed in: > * > @@ -29,8 +27,9 @@ > #include <sys/ioctl.h> > #include <stdio.h> > #include "tst_test.h" > +#include "tst_memutils.h" > -#define BUF_SIZE 128 * 4096 > +#define BUF_SIZE (128 * 4096) > #define CMD_SIZE 6 > static int devfd = -1; > @@ -80,6 +79,10 @@ static void setup(void) > tst_brk(TCONF, "Could not find any usable SCSI device"); > tst_res(TINFO, "Found SCSI device %s", devpath); > + > + /* Pollute some memory to avoid false negatives */ > + tst_pollute_memory(0, 0x42); > + > devfd = SAFE_OPEN(devpath, O_RDONLY); > query.interface_id = 'S'; > query.dxfer_direction = SG_DXFER_FROM_DEV;
diff --git a/testcases/kernel/syscalls/ioctl/ioctl_sg01.c b/testcases/kernel/syscalls/ioctl/ioctl_sg01.c index daaa96be5..8c9fd0dae 100644 --- a/testcases/kernel/syscalls/ioctl/ioctl_sg01.c +++ b/testcases/kernel/syscalls/ioctl/ioctl_sg01.c @@ -7,9 +7,7 @@ * CVE-2018-1000204 * * Test ioctl(SG_IO) and check that kernel doesn't leak data. Requires - * a read-accessible SCSI-compatible device (e.g. SATA disk). Running oom* - * test program before this one may increase the chance of successfully - * reproducing the bug. + * a read-accessible generic SCSI device (e.g. a DVD drive). * * Leak fixed in: * @@ -29,8 +27,9 @@ #include <sys/ioctl.h> #include <stdio.h> #include "tst_test.h" +#include "tst_memutils.h" -#define BUF_SIZE 128 * 4096 +#define BUF_SIZE (128 * 4096) #define CMD_SIZE 6 static int devfd = -1; @@ -80,6 +79,10 @@ static void setup(void) tst_brk(TCONF, "Could not find any usable SCSI device"); tst_res(TINFO, "Found SCSI device %s", devpath); + + /* Pollute some memory to avoid false negatives */ + tst_pollute_memory(0, 0x42); + devfd = SAFE_OPEN(devpath, O_RDONLY); query.interface_id = 'S'; query.dxfer_direction = SG_DXFER_FROM_DEV;
The test wasn't reliable if most of available memory was full of zeroes. Pollute free memory to increase the chance of detecting data leak. Signed-off-by: Martin Doucha <mdoucha@suse.cz> --- Changes since v1: - Split patch - Use tst_pollute_memory() instead of allocating and pre-polluting a fixed-size block of memory in setup(). testcases/kernel/syscalls/ioctl/ioctl_sg01.c | 11 +++++++---- 1 file changed, 7 insertions(+), 4 deletions(-)