On 07.10.2024 19:43, Al Viro wrote:
> ->close_on_exec[] state is maintained only for opened descriptors;
> as the result, anything that marks a descriptor opened has to
> set its cloexec state explicitly.
>
> As the result, all calls of __set_open_fd() are followed by
> __set_close_on_exec(); might as well fold it into __set_open_fd()
> so that cloexec state is defined as soon as the descriptor is
> marked opened.
>
> Signed-off-by: Al Viro <viro@xxxxxxxxxxxxxxxxxx>
This patch landed in today's linux-next as commit 218a562f273b ("make
__set_open_fd() set cloexec state as well"). In my tests I found that it
breaks booting of many of my test systems (arm 32bit, arm 64bit and
riscv64). It's hard to describe what exactly is broken, but none of the
affected boards reached the login shell. All crashed somewhere in the
userspace during systemd startup. This can be easily reproduced even
with qemu.
> ---
> fs/file.c | 9 ++++-----
> 1 file changed, 4 insertions(+), 5 deletions(-)
>
> diff --git a/fs/file.c b/fs/file.c
> index d8fccd4796a9..b63294ed85ec 100644
> --- a/fs/file.c
> +++ b/fs/file.c
> @@ -248,12 +248,13 @@ static inline void __set_close_on_exec(unsigned int fd, struct fdtable *fdt,
> }
> }
>
> -static inline void __set_open_fd(unsigned int fd, struct fdtable *fdt)
> +static inline void __set_open_fd(unsigned int fd, struct fdtable *fdt, bool set)
> {
> __set_bit(fd, fdt->open_fds);
> fd /= BITS_PER_LONG;
> if (!~fdt->open_fds[fd])
> __set_bit(fd, fdt->full_fds_bits);
> + __set_close_on_exec(fd, fdt, set);
> }
>
> static inline void __clear_open_fd(unsigned int fd, struct fdtable *fdt)
> @@ -517,8 +518,7 @@ static int alloc_fd(unsigned start, unsigned end, unsigned flags)
> if (start <= files->next_fd)
> files->next_fd = fd + 1;
>
> - __set_open_fd(fd, fdt);
> - __set_close_on_exec(fd, fdt, flags & O_CLOEXEC);
> + __set_open_fd(fd, fdt, flags & O_CLOEXEC);
> error = fd;
>
> out:
> @@ -1186,8 +1186,7 @@ __releases(&files->file_lock)
> goto Ebusy;
> get_file(file);
> rcu_assign_pointer(fdt->fd[fd], file);
> - __set_open_fd(fd, fdt);
> - __set_close_on_exec(fd, fdt, flags & O_CLOEXEC);
> + __set_open_fd(fd, fdt, flags & O_CLOEXEC);
> spin_unlock(&files->file_lock);
>
> if (tofree)
Best regards
--
Marek Szyprowski, PhD
Samsung R&D Institute Poland
