| Message ID | 20181120053642.24513-1-chandan@linux.vnet.ibm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | [RESEND] get_fs_type: Validate fs type string argument | expand |
On Tuesday, November 20, 2018 11:06:42 AM IST Chandan Rajendra wrote: > On ppc64le, When a string with PAGE_SIZE - 1 (i.e. 64k-1) length is > passed as a "filesystem type" argument to the mount(2) syscall, > copy_mount_string() ends up allocating 64k (the PAGE_SIZE on ppc64le) > worth of space for holding the string in kernel's address space. > > Later, in set_precision() (invoked by get_fs_type() -> > __request_module() -> vsnprintf()), we end up assigning > strlen(fs-type-string) i.e. 65535 as the > value to 'struct printf_spec'->precision member. This field has a width > of 16 bits and it is a signed data type. Hence an invalid value ends > up getting assigned. This causes the "WARN_ONCE(spec->precision != prec, > "precision %d too large", prec)" statement inside set_precision() to be > executed. > > This commit fixes the bug by validating the length of the "filesystem > type" argument passed to get_fs_type() function. > The following is a trivial userspace program to recreate the issue, #include <stdio.h> #include <stdlib.h> #include <string.h> #include <sys/mount.h> #define BUFSIZE 65536 char buf[BUFSIZE]; int main(int argc, char *argv[]) { int ret; if (argc != 3) { fprintf(stderr, "Usage: %s <device> <mount point>.\n", argv[0]); exit(1); } memset(buf, 1, BUFSIZE); buf[BUFSIZE-1] = '\0'; printf("strlen(buf) = %lu.\n", strlen(buf)); ret = mount(argv[1], argv[2], buf, 0, NULL); if (ret) { perror("mount"); exit(0); } exit(1); }
On Tue, Nov 20, 2018 at 6:36 AM Chandan Rajendra <chandan@linux.vnet.ibm.com> wrote: > > On ppc64le, When a string with PAGE_SIZE - 1 (i.e. 64k-1) length is > passed as a "filesystem type" argument to the mount(2) syscall, > copy_mount_string() ends up allocating 64k (the PAGE_SIZE on ppc64le) > worth of space for holding the string in kernel's address space. > > Later, in set_precision() (invoked by get_fs_type() -> > __request_module() -> vsnprintf()), we end up assigning > strlen(fs-type-string) i.e. 65535 as the > value to 'struct printf_spec'->precision member. This field has a width > of 16 bits and it is a signed data type. Hence an invalid value ends > up getting assigned. This causes the "WARN_ONCE(spec->precision != prec, > "precision %d too large", prec)" statement inside set_precision() to be > executed. > > This commit fixes the bug by validating the length of the "filesystem > type" argument passed to get_fs_type() function. > > Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com> > Reported-by: Abdul Haleem <abdhalee@linux.vnet.ibm.com> > Suggested-by: Joe Perches <joe@perches.com> Acked-by: Miklos Szeredi <mszeredi@redhat.com> Al, please pick this up, it looks like a good sanity check and lack of it is causing headaches for IBM QA. Thanks, Miklos > --- > fs/filesystems.c | 3 +++ > 1 file changed, 3 insertions(+) > > diff --git a/fs/filesystems.c b/fs/filesystems.c > index 9135646e41ac..a61caf5b6ad3 100644 > --- a/fs/filesystems.c > +++ b/fs/filesystems.c > @@ -268,6 +268,9 @@ struct file_system_type *get_fs_type(const char *name) > const char *dot = strchr(name, '.'); > int len = dot ? dot - name : strlen(name); > > + if (len >= PATH_MAX) > + return NULL; > + > fs = __get_fs_type(name, len); > if (!fs && (request_module("fs-%.*s", len, name) == 0)) { > fs = __get_fs_type(name, len); > -- > 2.19.1 >
On Mon, Jan 21, 2019 at 01:20:57PM +0100, Miklos Szeredi wrote: > On Tue, Nov 20, 2018 at 6:36 AM Chandan Rajendra > <chandan@linux.vnet.ibm.com> wrote: > > > > On ppc64le, When a string with PAGE_SIZE - 1 (i.e. 64k-1) length is > > passed as a "filesystem type" argument to the mount(2) syscall, > > copy_mount_string() ends up allocating 64k (the PAGE_SIZE on ppc64le) > > worth of space for holding the string in kernel's address space. > > > > Later, in set_precision() (invoked by get_fs_type() -> > > __request_module() -> vsnprintf()), we end up assigning > > strlen(fs-type-string) i.e. 65535 as the > > value to 'struct printf_spec'->precision member. This field has a width > > of 16 bits and it is a signed data type. Hence an invalid value ends > > up getting assigned. This causes the "WARN_ONCE(spec->precision != prec, > > "precision %d too large", prec)" statement inside set_precision() to be > > executed. > > > > This commit fixes the bug by validating the length of the "filesystem > > type" argument passed to get_fs_type() function. > > > > Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com> > > Reported-by: Abdul Haleem <abdhalee@linux.vnet.ibm.com> > > Suggested-by: Joe Perches <joe@perches.com> > > Acked-by: Miklos Szeredi <mszeredi@redhat.com> > > Al, please pick this up, it looks like a good sanity check and lack of > it is causing headaches for IBM QA. Umm... I'm not against that patch, but I wonder if that should be caught earlier... If nothing else, the same string is seen by LSM shite before get_fs_type() gets a chance to reject it, and I wouldn't bet a dime on robustness of that code. Wouldn't it be saner to have char *copy_mount_string(const void __user *data) { return data ? strndup_user(data, PAGE_SIZE) : NULL; } use a more explicit size limit? PATH_MAX, for example?
diff --git a/fs/filesystems.c b/fs/filesystems.c index 9135646e41ac..a61caf5b6ad3 100644 --- a/fs/filesystems.c +++ b/fs/filesystems.c @@ -268,6 +268,9 @@ struct file_system_type *get_fs_type(const char *name) const char *dot = strchr(name, '.'); int len = dot ? dot - name : strlen(name); + if (len >= PATH_MAX) + return NULL; + fs = __get_fs_type(name, len); if (!fs && (request_module("fs-%.*s", len, name) == 0)) { fs = __get_fs_type(name, len);
On ppc64le, When a string with PAGE_SIZE - 1 (i.e. 64k-1) length is passed as a "filesystem type" argument to the mount(2) syscall, copy_mount_string() ends up allocating 64k (the PAGE_SIZE on ppc64le) worth of space for holding the string in kernel's address space. Later, in set_precision() (invoked by get_fs_type() -> __request_module() -> vsnprintf()), we end up assigning strlen(fs-type-string) i.e. 65535 as the value to 'struct printf_spec'->precision member. This field has a width of 16 bits and it is a signed data type. Hence an invalid value ends up getting assigned. This causes the "WARN_ONCE(spec->precision != prec, "precision %d too large", prec)" statement inside set_precision() to be executed. This commit fixes the bug by validating the length of the "filesystem type" argument passed to get_fs_type() function. Signed-off-by: Chandan Rajendra <chandan@linux.vnet.ibm.com> Reported-by: Abdul Haleem <abdhalee@linux.vnet.ibm.com> Suggested-by: Joe Perches <joe@perches.com> --- fs/filesystems.c | 3 +++ 1 file changed, 3 insertions(+)