[GIT PULL] pidfd and clone3 fixes

* [GIT PULL] pidfd and clone3 fixes
@ 2019-07-15 15:15 Christian Brauner
  2019-07-16 19:40 ` pr-tracker-bot
  0 siblings, 1 reply; 3+ messages in thread
From: Christian Brauner @ 2019-07-15 15:15 UTC (permalink / raw)
  To: torvalds; +Cc: linux-kernel, ldv

Hi Linus,

This contains a bugfix for CLONE_PIDFD when used with the legacy clone
syscall. two fixes to ensure that syscall numbering and clone3
entrypoint implementations will stay consistent, and an update for the
maintainers file:

The following changes since commit 964a4eacef67503a1154f7e0a75f52fbdce52022:

  Merge tag 'dlm-5.3' of git://git.kernel.org/pub/scm/linux/kernel/git/teigland/linux-dlm (2019-07-12 17:37:53 -0700)

are available in the Git repository at:

  git@gitolite.kernel.org:pub/scm/linux/kernel/git/brauner/linux tags/for-linus-20190715

for you to fetch changes up to 69b53720e92c1bdea854a2fc204477ddabfa902b:

  MAINTAINERS: add new entry for pidfd api (2019-07-15 12:59:44 +0200)

/* Summary */
The addition of clone3 broke CLONE_PIDFD for legacy clone on all
architectures that use do_fork() directly instead of calling the clone
syscall itself. (Fwiw, cleaning do_fork() up is on my todo.)
The reason this happened was that during conversion of  _do_fork() to use
struct kernel_clone_args we missed that do_fork() is called directly by
various architectures. This is fixed by making sure that the pidfd argument
in struct kernel_clone_args is correctly initialized with the parent_tidptr
argument passed down from do_fork(). Additionally, do_fork() missed a check
to make CLONE_PIDFD and CLONE_PARENT_SETTID mutually exclusive just a
clone() does. This is now fixed too.

When clone3() was introduced we skipped architectures that require special
handling for fork-like syscalls. Their syscall tables did not contain any
mention of clone3(). To make sure that Arnd's work to make syscall numbers
on all architectures identical (minus alpha) was not for naught we are
placing a comment in all syscall tables that do not yet implement clone3().
The comment makes it clear that 435 is reserved for clone3 and should not
be used.

Also, this PR contains a patch to make the clone3() syscall definition in
asm-generic/unist.h conditional on __ARCH_WANT_SYS_CLONE3. This lets us
catch new architectures that implicitly make use of clone3 without setting
__ARCH_WANT_SYS_CLONE3 which is a good indicator that they did not check
whether it needs special treatment or not.

Last, this contains a patch to add me as maintainer for pidfd stuff so
people can start blaming me (more).

Please consider pulling from the signed for-linus-20190715 tag.

Thanks!
Christian

----------------------------------------------------------------
for-linus-20190715

----------------------------------------------------------------
Christian Brauner (3):
      arch: mark syscall number 435 reserved for clone3
      unistd: protect clone3 via __ARCH_WANT_SYS_CLONE3
      MAINTAINERS: add new entry for pidfd api

Dmitry V. Levin (1):
      clone: fix CLONE_PIDFD support

 MAINTAINERS                               | 11 +++++++++++
 arch/alpha/kernel/syscalls/syscall.tbl    |  1 +
 arch/ia64/kernel/syscalls/syscall.tbl     |  1 +
 arch/m68k/kernel/syscalls/syscall.tbl     |  1 +
 arch/mips/kernel/syscalls/syscall_n32.tbl |  1 +
 arch/mips/kernel/syscalls/syscall_n64.tbl |  1 +
 arch/mips/kernel/syscalls/syscall_o32.tbl |  1 +
 arch/parisc/kernel/syscalls/syscall.tbl   |  1 +
 arch/powerpc/kernel/syscalls/syscall.tbl  |  1 +
 arch/s390/kernel/syscalls/syscall.tbl     |  1 +
 arch/sh/kernel/syscalls/syscall.tbl       |  1 +
 arch/sparc/kernel/syscalls/syscall.tbl    |  1 +
 arch/x86/ia32/sys_ia32.c                  |  4 ++++
 include/linux/sched/task.h                |  1 +
 include/uapi/asm-generic/unistd.h         |  2 ++
 kernel/fork.c                             | 17 +++++++++++++++--
 16 files changed, 44 insertions(+), 2 deletions(-)

^ permalink raw reply	[flat|nested] 3+ messages in thread