lkml.org 
[lkml]   [2020]   [Mar]   [6]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH v2 3/9] rcu,tracing: Create trace_rcu_{enter,exit}()
On Sat, Mar 07, 2020 at 03:01:49AM +0900, Masami Hiramatsu wrote:
> Hi,
>
> On Wed, 19 Feb 2020 11:45:10 +0900
> Masami Hiramatsu <mhiramat@kernel.org> wrote:
>
> > On Tue, 18 Feb 2020 12:18:06 -0800
> > "Paul E. McKenney" <paulmck@kernel.org> wrote:
> >
> > > On Tue, Feb 18, 2020 at 12:46:09PM -0500, Steven Rostedt wrote:
> > > > On Tue, 18 Feb 2020 13:33:35 +0900
> > > > Masami Hiramatsu <mhiramat@kernel.org> wrote:
> > > >
> > > > > On Mon, 17 Feb 2020 08:31:12 -0800
> > > > > "Paul E. McKenney" <paulmck@kernel.org> wrote:
> > > > > >
> > > > > > > BTW, if you consider the x86 specific code is in the generic file,
> > > > > > > we can move NOKPROBE_SYMBOL() in arch/x86/kernel/traps.c.
> > > > > > > (Sorry, I've hit this idea right now)
> > > > > >
> > > > > > Might this affect other architectures with NMIs and probe-like things?
> > > > > > If so, it might make sense to leave it where it is.
> > > > >
> > > > > Yes, git grep shows that arm64 is using rcu_nmi_enter() in
> > > > > debug_exception_enter().
> > > > > OK, let's keep it, but maybe it is good to update the comment for
> > > > > arm64 too. What about following?
> > > > >
> > > > > +/*
> > > > > + * All functions in do_int3() on x86, do_debug_exception() on arm64 must be
> > > > > + * marked NOKPROBE before kprobes handler is called.
> > > > > + * ist_enter() on x86 and debug_exception_enter() on arm64 which is called
> > > > > + * before kprobes handle happens to call rcu_nmi_enter() which means
> > > > > + * that rcu_nmi_enter() must be marked NOKRPOBE.
> > > > > + */
> > > > >
> > > >
> > > > Ah, why don't we just say...
> > > >
> > > > /*
> > > > * All functions called in the breakpoint trap handler (e.g. do_int3()
> > > > * on x86), must not allow kprobes until the kprobe breakpoint handler
> > > > * is called, otherwise it can cause an infinite recursion.
> > > > * On some archs, rcu_nmi_enter() is called in the breakpoint handler
> > > > * before the kprobe breakpoint handler is called, thus it must be
> > > > * marked as NOKPROBE.
> > > > */
> > > >
> > > > And that way we don't make this an arch specific comment.
> > >
> > > That looks good to me. Masami, does this work for you?
> >
> > Yes, that looks good to me too :)
>
> Oops, I'm guilty!
> Sorry *rcu_nmi_exit()* also must be NOKPROBE, since even if we could catch
> a recursive kprobe call, we can only skip the kprobe handler, but we must
> exit from do_int3() and hit rcu_nmi_exit() again!
>
> [45235.497591] Unrecoverable kprobe detected.
> [45235.501400] Dumping kprobe:
> [45235.502433] Name: (null)
> [45235.502433] Offset: 0
> [45235.502433] Address: rcu_nmi_exit+0x0/0x290
> [45235.504044] ------------[ cut here ]------------
> [45235.504855] kernel BUG at arch/x86/kernel/kprobes/core.c:646!
> [45235.505816] invalid opcode: 0000 [#1] PREEMPT SMP PTI
> [45235.506615] CPU: 7 PID: 143 Comm: sh Not tainted 5.6.0-rc3+ #143
> [45235.507662] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.12.1-0-ga5cab58e9a3f-prebuilt.qemu.org 04/01/2014
> [45235.509764] RIP: 0010:reenter_kprobe.cold+0x14/0x16
> [45235.510630] Code: 48 8b 75 10 48 c7 c7 f0 70 0e 82 48 8b 56 28 e8 22 91 08 00 0f 0b 48 c7 c7 20 71 0e 82 e8 14 91 08 00 48 89 ef e8 23 ee 0f 00 <0f> 0b 48 89 ee 48 c7 c7 48 71 0e 82 e8 fb 90 08 00 e9 c3 fc ff ff
> [45235.513948] RSP: 0018:ffffc90000347bf8 EFLAGS: 00010046
> [45235.514906] RAX: 0000000000000036 RBX: 0000000000017f20 RCX: 0000000000000000
> [45235.516109] RDX: 0000000000000000 RSI: 0000000000000000 RDI: 0000000000000001
> [45235.517278] RBP: ffff88807c9820c0 R08: 0000000000000000 R09: 0000000000000001
> [45235.518415] R10: 0000000000000000 R11: ffff88807c9d1f18 R12: ffff88807d9d7f20
> [45235.519609] R13: ffffc90000347c68 R14: ffffffff810e8a60 R15: ffffffff810e8a61
> [45235.520787] FS: 0000000001d9a8c0(0000) GS:ffff88807d9c0000(0000) knlGS:0000000000000000
> [45235.522198] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
> [45235.523172] CR2: 0000000001da9000 CR3: 000000007a880000 CR4: 00000000000006a0
> [45235.524288] Call Trace:
> [45235.524825] kprobe_int3_handler+0x74/0x150
> [45235.525627] do_int3+0x36/0xf0
> [45235.526244] int3+0x42/0x50
> [45235.526767] RIP: 0010:rcu_nmi_exit+0x1/0x290
> [45235.527551] Code: a2 0d 82 be c2 01 00 00 48 c7 c7 d5 44 0f 82 c6 05 e7 ac 24 01 01 e8 1f ba fd ff eb b8 66 66 2e 0f 1f 84 00 00 00 00 00 90 cc <57> 41 56 41 55 41 54 55 48 c7 c5 40 c2 02 00 53 48 89 eb e8 77 75
> [45235.530898] RSP: 0018:ffffc90000347d40 EFLAGS: 00000046
> [45235.531816] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
> [45235.533001] RDX: 0000000000000001 RSI: ffffffff8101e1fe RDI: ffffffff8101e1fe
> [45235.534252] RBP: 0000000000000000 R08: 0000000000000001 R09: 0000000000000000
> [45235.535516] R10: 0000000000000000 R11: 0000000000000000 R12: 0000000000000000
> [45235.536759] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
> [45235.537945] ? ist_exit+0xe/0x20
> [45235.538593] ? ist_exit+0xe/0x20
> [45235.539239] ? rcu_nmi_exit+0x1/0x290
> [45235.541182] int3+0x42/0x50
> [45235.541687] RIP: 0010:0xffffffffa000005a
> [45235.542363] Code: 2e 16 13 e1 00 00 00 00 00 00 00 00 89 f8 e9 1f 16 13 e1 00 00 00 00 00 00 00 00 89 f8 e9 20 16 13 e1 00 00 00 00 00 00 00 00 <41> 57 e9 01 8a 0e e1 00 00 00 00 00 00 00 00 41 57 e9 f2 22 26 e1
> [45235.545628] RSP: 0018:ffffc90000347e20 EFLAGS: 00000146
> [45235.546596] RAX: 0000000000000000 RBX: 0000000000000000 RCX: 0000000000000000
> [45235.547989] RDX: 0000000000000001 RSI: ffffffff8101e1fe RDI: ffffffff8101e1fe
> [45235.550183] RBP: 0000000000000000 R08: 0000000000000001 R09: ffff88807d2aa000
> [45235.551591] R10: 0000000000000a4c R11: ffff88807bfec600 R12: 0000000000000000
> [45235.552893] R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
> [45235.554633] ? ist_exit+0xe/0x20
> [45235.555537] ? ist_exit+0xe/0x20
> [45235.556565] ? rcu_nmi_exit+0x1/0x290
> [45235.557909] ? int3+0x42/0x50
> [45235.559156] ? 0xffffffffa0000069
> [45235.560547] ? vfs_read+0x1/0x150
> [45235.561522] ? ksys_read+0x60/0xe0
> [45235.562458] ? do_syscall_64+0x4b/0x1e0
> [45235.563404] ? entry_SYSCALL_64_after_hwframe+0x49/0xbe
> [45235.564705] Modules linked in:
> [45235.565556] ---[ end trace 870af8724dba9ac8 ]---
>
> So all functions called from do_int3() must be NOKPROBE.

Makes sense, I am curious though why you thought before that the breakpoint
recursion was not possible *after* kprobe_int3_handler(). In the below thread
you said it is to be marked only for functions *before*
kprobe_int3_handler(). Was there a reason?
https://lore.kernel.org/lkml/154998793571.31052.11301258949601150994.stgit@devbox/

thanks,

- Joel


>
> Thank you,
>
> --
> Masami Hiramatsu <mhiramat@kernel.org>

\
 
 \ /
  Last update: 2020-03-06 20:12    [W:0.085 / U:6.928 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site