APPLIED: [SRU][Xenial][PATCH] Fix for LP:#1775326

[Khaled Elmously]

Applied to Xenial


On 2018-06-07 10:58:36 , Gavin Guo wrote:
> BugLink: https://bugs.launchpad.net/bugs/1775326
> 
> [Impact]
> 
> In function cpuacct_charge(), the NULL pointer dereference happens
> with the stack pointer being zero inside the task_struct when the
> task_cpu() is trying to access the member cpu of the struct
> thread_info inside the stack. It's a use-after-free corruption
> happening in the situation that the task_struct is released almost
> concurrently before accessing the task_struct->stack.
> 
> void cpuacct_charge(struct task_struct *tsk, u64 cputime)
>  {
>         struct cpuacct *ca;
>         int cpu;
>  
>         cpu = task_cpu(tsk);
>  
>         rcu_read_lock();
>  
>         ca = task_ca(tsk);
>  
>         while (true) {
>                 u64 *cpuusage = per_cpu_ptr(ca->cpuusage, cpu);
>                 *cpuusage += cputime;
>  
>                 ca = parent_ca(ca);
>                 if (!ca)
>                         break;
>         }
> 
> 	rcu_read_unlock();
> }
> 
> 
> BUG: unable to handle kernel NULL pointer dereference at 0000000000000010
> IP: [<ffffffff810c3ff4>] cpuacct_charge+0x14/0x40
> PGD 0 
> Oops: 0000 [#1] SMP  
> CPU: 10 PID: 148614 Comm: qemu-system-x86 Tainted: P        W  OE   4.4.0-45-generic #66~14.04.1-Ubuntu
> Hardware name: Dell Inc. PowerEdge R630/02C2CP, BIOS 2.1.7 06/16/2016
> task: ffff881ff0f01b80 ti: ffff88018fd70000 task.ti: ffff88018fd70000
> RIP: 0010:[<ffffffff810c3ff4>]  [<ffffffff810c3ff4>] cpuacct_charge+0x14/0x40
> RSP: 0018:ffff88018fd73d10  EFLAGS: 00010246
> RAX: 0000000000000000 RBX: ffff8801931e8000 RCX: ffff88010caff200
> RDX: ffff880124508000 RSI: 0066f757398831d6 RDI: ffff8801931e7fa0
> RBP: ffff88018fd73d10 R08: ffffffffc04b8320 R09: 0000000000000001
> R10: 0000000000000001 R11: 0000000000000000 R12: 0066f757398831d6
> R13: 0066f757398b8997 R14: ffff8801931e7fa0 R15: 0000000000000001
> FS:  00007f162aaf7700(0000) GS:ffff881ffe740000(0000) knlGS:0000000000000000
> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> CR2: 0000000000000010 CR3: 000000011d86e000 CR4: 00000000003426e0
> DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
> DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
> Stack:
>  ffff88018fd73d28 ffffffff810b1a9f ffff8801931e8000 ffff88018fd73d40
>  ffffffffc069df72 ffff8801931e8000 ffff88018fd73da8 ffffffffc069f121
>  ffff881ff0f01b80 0000000000000000 ffff881ff0f01b80 ffffffff810bddc0
> Call Trace:
>  [<ffffffff810b1a9f>] update_curr+0xdf/0x170
>  [<ffffffffc069df72>] kvm_vcpu_check_block+0x12/0x60 [kvm]
>  [<ffffffffc069f121>] kvm_vcpu_block+0x191/0x2d0 [kvm]
>  [<ffffffff810bddc0>] ? prepare_to_wait_event+0xf0/0xf0
>  [<ffffffffc06bb9ee>] kvm_arch_vcpu_ioctl_run+0x17e/0x3d0 [kvm]
>  [<ffffffffc06a1f8b>] kvm_vcpu_ioctl+0x2ab/0x640 [kvm]
>  [<ffffffff81174517>] ? perf_event_context_sched_in+0x87/0xa0
>  [<ffffffff81210d6d>] do_vfs_ioctl+0x2dd/0x4c0
>  [<ffffffff8111fa1f>] ? __audit_syscall_entry+0xaf/0x100
>  [<ffffffff81003176>] ? do_audit_syscall_entry+0x66/0x70
>  [<ffffffff81210fc9>] SyS_ioctl+0x79/0x90
>  [<ffffffff817fa4f6>] entry_SYSCALL_64_fastpath+0x16/0x75
> Code: 9a 11 00 5b 48 c7 c0 f4 ff ff ff 5d eb df 66 0f 1f 84 00 00 00 00 00 0f 1f 44 00 00 55 48 8b 47 08 48 8b 97 78 07 00 00 48 89 e5 <48> 63 48 10 48 8b 52 60 48 8b 82 b8 00 00 00 48 03 04 cd c0 7a
> RIP  [<ffffffff810c3ff4>] cpuacct_charge+0x14/0x40
>  RSP <ffff88018fd73d10>
> CR2: 0000000000000010
> ---[ end trace 419a30375d0e4622 ]---
> 
> 
> 
> [Fix]
> 
> The patch uses this_cpu_ptr() instead of getting the cpu number by 
> task_cpu() and proceeds to get the cpu_usage by per_cpu_ptr(). And
> that can avoid accessing the thread_info inside the stack. 
> 
> commit 73e6aafd9ea81498d31361f01db84a0118da2d1c
> Author: Zhao Lei <zhaolei at cn.fujitsu.com>
> Date:   Thu Mar 17 12:19:43 2016 +0800
> 
>     sched/cpuacct: Simplify the cpuacct code
>     
>      - Use for() instead of while() loop in some functions
>        to make the code simpler.
>     
>      - Use this_cpu_ptr() instead of per_cpu_ptr() to make the code
>        cleaner and a bit faster.
>     
>     Suggested-by: Peter Zijlstra <peterz at infradead.org>
>     Signed-off-by: Zhao Lei <zhaolei at cn.fujitsu.com>
>     Signed-off-by: Peter Zijlstra (Intel) <peterz at infradead.org>
>     Cc: Linus Torvalds <torvalds at linux-foundation.org>
>     Cc: Tejun Heo <htejun at gmail.com>
>     Cc: Thomas Gleixner <tglx at linutronix.de>
>     Link: http://lkml.kernel.org/r/d8a7ef9592f55224630cb26dea239f05b6398a4e.1458187654.git.zhaolei@cn.fujitsu.com
>     Signed-off-by: Ingo Molnar <mingo at kernel.org>
> 
> 
> 
> [Test]
> The kernel has been tested by the Qemu and went well.
> 
> Zhao Lei (1):
>   sched/cpuacct: Simplify the cpuacct code
> 
>  kernel/sched/cpuacct.c | 28 +++++-----------------------
>  kernel/sched/cpuacct.h |  4 ++--
>  2 files changed, 7 insertions(+), 25 deletions(-)
> 
> -- 
> 2.7.4
> 
> 
> -- 
> kernel-team mailing list
> kernel-team at lists.ubuntu.com
> https://lists.ubuntu.com/mailman/listinfo/kernel-team