Re: [PATCH v6 2/3] seccomp_filters: system call filtering using BPF
From: Serge E. Hallyn
Date: Thu Feb 02 2012 - 10:32:43 EST
Quoting Will Drewry (wad@xxxxxxxxxxxx):
> [This patch depends on luto@xxxxxxx's no_new_privs patch:
> This patch adds support for seccomp mode 2. This mode enables dynamic
> enforcement of system call filtering policy in the kernel as specified
> by a userland task. The policy is expressed in terms of a Berkeley
> Packet Filter program, as is used for userland-exposed socket filtering.
> Instead of network data, the BPF program is evaluated over struct
> seccomp_filter_data at the time of the system call.
> A filter program may be installed by a userland task by calling
> prctl(PR_ATTACH_SECCOMP_FILTER, &fprog);
> where fprog is of type struct sock_fprog.
> If the first filter program allows subsequent prctl(2) calls, then
> additional filter programs may be attached. All attached programs
> must be evaluated before a system call will be allowed to proceed.
> To avoid CONFIG_COMPAT related landmines, once a filter program is
> installed using specific is_compat_task() value, it is not allowed to
> make system calls using the alternate entry point.
> Filter programs will be inherited across fork/clone and execve, however
> the installation of filters must be preceded by setting 'no_new_privs'
> to ensure that unprivileged tasks cannot attach filters that affect
> privileged tasks (e.g., setuid binary). Tasks with CAP_SYS_ADMIN
> in their namespace may install inheritable filters without setting
> the no_new_privs bit.
> There are a number of benefits to this approach. A few of which are
> as follows:
> - BPF has been exposed to userland for a long time.
> - Userland already knows its ABI: system call numbers and desired
> - No time-of-check-time-of-use vulnerable data accesses are possible.
> - system call arguments are loaded on demand only to minimize copying
> required for system call number-only policy decisions.
> This patch includes its own BPF evaluator, but relies on the
> net/core/filter.c BPF checking code. It is possible to share
> evaluators, but the performance sensitive nature of the network
> filtering path makes it an iterative optimization which (I think :) can
> be tackled separately via separate patchsets. (And at some point sharing
> BPF JIT code!)
> v6: - fix memory leak on attach compat check failure
> - require no_new_privs || CAP_SYS_ADMIN prior to filter
> installation. (luto@xxxxxxx)
> - s/seccomp_struct_/seccomp_/ for macros/functions
> - cleaned up Kconfig (amwang@xxxxxxxxxx)
> - on block, note if the call was compat (so the # means something)
> v5: - uses syscall_get_arguments
> (indan@xxxxxx,oleg@xxxxxxxxxx, mcgrathr@xxxxxxxxxxxx)
> - uses union-based arg storage with hi/lo struct to
> handle endianness. Compromises between the two alternate
> proposals to minimize extra arg shuffling and account for
> endianness assuming userspace uses offsetof().
> (mcgrathr@xxxxxxxxxxxx, indan@xxxxxx)
> - update Kconfig description
> - add include/seccomp_filter.h and add its installation
> - (naive) on-demand syscall argument loading
> - drop seccomp_t (eparis@xxxxxxxxxx)
> v4: - adjusted prctl to make room for PR_[SG]ET_NO_NEW_PRIVS
> - now uses current->no_new_privs
> - assign names to seccomp modes (rdunlap@xxxxxxxxxxxx)
> - fix style issues (rdunlap@xxxxxxxxxxxx)
> - reworded Kconfig entry (rdunlap@xxxxxxxxxxxx)
> v3: - macros to inline (oleg@xxxxxxxxxx)
> - init_task behavior fixed (oleg@xxxxxxxxxx)
> - drop creator entry and extra NULL check (oleg@xxxxxxxxxx)
> - alloc returns -EINVAL on bad sizing (serge.hallyn@xxxxxxxxxxxxx)
> - adds tentative use of "always_unprivileged" as per
> torvalds@xxxxxxxxxxxxxxxxxxxx and luto@xxxxxxx
> v2: - (patch 2 only)
> Signed-off-by: Will Drewry <wad@xxxxxxxxxxxx>
as far as I can tell based on changelog I suspect you could have
kept my Acked-by (from v3?). However, I'll wait until your next
submission (as I see there were a few change requests), and do a
final complete new review of that.
Thanks for continuing to push on this.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/