[LTP] [PATCH v2] Migrating the libhugetlbfs/testcases/truncate_sigbus_versus_oom.c test

[Richard Palethorpe]

Hello,

Samir Mulani <samir@linux.vnet.ibm.com> writes:

> In this test case, we are verifying the bug fix commit that is attached as
> a part of the test case structure,
>
> Some kernel have a bug in the positioning of the test against
> i_size.  This bug means that attempting to instantiate a page
> beyond the end of a hugepage file can result in an OOM and SIGKILL
> instead of the correct SIGBUS.
>
> Signed-off-by: Samir Mulani <samir@linux.vnet.ibm.com>
> ---
> v2:
>  -Corrected typo.
>  -Fixed the make check warnings.
> ---
>  runtest/hugetlb                               |  1 +
>  testcases/kernel/mem/.gitignore               |  1 +
>  .../kernel/mem/hugetlb/hugemmap/hugemmap37.c  | 88 +++++++++++++++++++
>  3 files changed, 90 insertions(+)
>  create mode 100644 testcases/kernel/mem/hugetlb/hugemmap/hugemmap37.c
>
> diff --git a/runtest/hugetlb b/runtest/hugetlb
> index 299c07ac9..7b7c44b77 100644
> --- a/runtest/hugetlb
> +++ b/runtest/hugetlb
> @@ -35,6 +35,7 @@ hugemmap29 hugemmap29
>  hugemmap30 hugemmap30
>  hugemmap31 hugemmap31
>  hugemmap32 hugemmap32
> +hugemmap37 hugemmap37
>  hugemmap05_1 hugemmap05 -m
>  hugemmap05_2 hugemmap05 -s
>  hugemmap05_3 hugemmap05 -s -m
> diff --git a/testcases/kernel/mem/.gitignore b/testcases/kernel/mem/.gitignore
> index 7258489ed..7b923c8fd 100644
> --- a/testcases/kernel/mem/.gitignore
> +++ b/testcases/kernel/mem/.gitignore
> @@ -34,6 +34,7 @@
>  /hugetlb/hugemmap/hugemmap30
>  /hugetlb/hugemmap/hugemmap31
>  /hugetlb/hugemmap/hugemmap32
> +/hugetlb/hugemmap/hugemmap37
>  /hugetlb/hugeshmat/hugeshmat01
>  /hugetlb/hugeshmat/hugeshmat02
>  /hugetlb/hugeshmat/hugeshmat03
> diff --git a/testcases/kernel/mem/hugetlb/hugemmap/hugemmap37.c b/testcases/kernel/mem/hugetlb/hugemmap/hugemmap37.c
> new file mode 100644
> index 000000000..66990db25
> --- /dev/null
> +++ b/testcases/kernel/mem/hugetlb/hugemmap/hugemmap37.c
> @@ -0,0 +1,88 @@
> +// SPDX-License-Identifier: GPL-2.0-or-later
> +/*
> + * Copyright (C) 2005-2006 David Gibson & Adam Litke, IBM Corporation.
> + */
> +
> +/*\
> + * [Description]
> + *
> + * Test Name: Truncate_sigbus_versus_oom
> + *
> + * Some kernel have a bug in the positioning of the test against
> + * i_size.  This bug means that attempting to instantiate a page
> + * beyond the end of a hugepage file can result in an OOM and SIGKILL
> + * instead of the correct SIGBUS.
> + */
> +
> +#include "hugetlb.h"
> +
> +#define MNTPOINT "hugetlbfs/"
> +#define PTS_PASS 0
> +static int  fd = -1, fdx = -1;
> +
> +static unsigned long long hpage_size;
> +static unsigned long totpages;
> +struct sigaction sa;
> +
> +static void sigbus_handler(int signum)
> +{
> +	if (signum == SIGBUS) {
> +		tst_res(TPASS, "Test PASSED\n");
> +		_exit(PTS_PASS);

In LTP we try to do as little as possible in signal handlers. In rare
cases some library functions don't work as expected.

Here we should just set a volatile variable with the signal number that
was caught. Then inspect it later.

> +	}
> +}
> +
> +static void run_test(void)
> +{
> +	void *p, *q;
> +	unsigned long i;
> +
> +	fd = tst_creat_unlinked(MNTPOINT, 0);
> +	p = SAFE_MMAP(NULL, hpage_size, PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0);
> +	SAFE_FTRUNCATE(fd, 0);
> +
> +	fdx = tst_creat_unlinked(MNTPOINT, 0);
> +	q = SAFE_MMAP(NULL, totpages * hpage_size, PROT_READ|PROT_WRITE, MAP_SHARED, fdx, 0);
> +	/* Touch the pages to ensure they're removed from the pool */
> +	for (i = 0; i < totpages; i++) {
> +		volatile char *x = (volatile char *)q + i*hpage_size;
> +		*x = 0;
> +	}
> +	/* SIGBUS is what *should* happen */
> +	SAFE_FTRUNCATE(fdx, 0);
> +	*((volatile unsigned int *)p);
> +	tst_res(TFAIL, "Didn't SIGBUS or OOM");

Here we could check a variable we set in the signal handler.

There is nothing here to free the mapped pages or close the files. So
most likely SAFE_MMAP will fail on the second iteration (i.e. with -i 2)
and eventually we would run out of file descriptors.

> +}
> +
> +void setup(void)
> +{
> +	sa.sa_flags = SA_SIGINFO;
> +	sa.sa_handler = sigbus_handler;
> +	SAFE_SIGACTION(SIGBUS, &sa, NULL);
> +	totpages = SAFE_READ_MEMINFO(MEMINFO_HPAGE_FREE);

There is a possible TOCTOU problem here. I don't think it can be avoided
completely, but perhaps we should read the number of free pages just
before calling the second mmap?

> +	hpage_size = tst_get_hugepage_size();
> +}
> +
> +void cleanup(void)
> +{
> +	if (fd > 0)
> +		SAFE_CLOSE(fd);
> +	if (fdx > 0)
> +		SAFE_CLOSE(fdx);
> +}
> +
> +
> +static struct tst_test test = {
> +	.tags = (struct tst_tag[]) {
> +		{"linux-git", "0d59a01bc461"},

I think this is the wrong commit. It prevents the stack from growing
into a huge page region.

> +		{}
> +	},
> +	.needs_root = 1,
> +	.mntpoint = MNTPOINT,
> +	.needs_hugetlbfs = 1,
> +	.needs_tmpdir = 1,
> +	.setup = setup,
> +	.cleanup = cleanup,
> +	.test_all = run_test,
> +	.hugepages = {2, TST_NEEDS},
> +};
> -- 
> 2.39.3


-- 
Thank you,
Richard.