linux-kernel - Re: [RFC 4/4] x86/arch_prctl: add ARCH_{GET,SET}_TASK

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Fri, 30 Dec 2016 18:02:15 -0800
From:   Andy Lutomirski <luto@...capital.net>
To:     Dmitry Safonov <dsafonov@...tuozzo.com>
Cc:     "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Dmitry Safonov <0x7f454c46@...il.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>,
        "H. Peter Anvin" <hpa@...or.com>,
        Andy Lutomirski <luto@...nel.org>,
        "Kirill A. Shutemov" <kirill.shutemov@...ux.intel.com>,
        X86 ML <x86@...nel.org>
Subject: Re: [RFC 4/4] x86/arch_prctl: add ARCH_{GET,SET}_TASK_SIZE

On Fri, Dec 30, 2016 at 7:56 AM, Dmitry Safonov <dsafonov@...tuozzo.com> wrote:
> Add arch_prctl getters/setters for size of virtual address space of task.
> This adds ability to change task's virtual address space limit.
> I need this for correctly restore virtual address space limits in CRIU.
> Currently, on x86 there are three task sizes: 3GB for some old 32 bit java
> apps, 4Gb for ordinary 32-bit compatible apps and 47-bits for native
> x86_64 processes.
> 32-bit applications are restored by CRIU with the help of 64-bit clone()-d
> child, and on restore we need to place correct address space limitations
> back - otherwise 32-bit restored application may mmap() address over
> 4Gb space and as this address will not fit into 4-byte pointer, it
> will silently reuse/corrupt the pointer that has the same lower 4-bytes.
>

I agree we need something like this, but this particular justification
is a bit bogus.  If 32-bit mmap() returns an address above 2^32, then
I think it's a straight-up bug.  The address space limit shouldn't
have anything to do with  it -- the kernel *knows* that it's a
"compat" syscall.