Re: [PATCH v2] printk/nmi: avoid direct printk()-s from __printk_nmi_flush()

[Andrew Morton]

On Wed, 31 Aug 2016 10:44:41 +0900 Sergey Senozhatsky <sergey.senozhatsky.work@xxxxxxxxx> wrote:

> On (08/30/16 15:03), Andrew Morton wrote:
> > > __printk_nmi_flush() can be called from nmi_panic(), therefore it has to
> > > test whether it's executed in NMI context and thus must route the messages
> > > through deferred printk() or via direct printk().
> > 
> > Why?  What misbehaviour does the current code cause?
> 
> the reasoning behind the `if in_nmi()' in print_nmi_seq_line()
> 
>        if (in_nmi())
>                printk_deferred("%.*s", (end - start) + 1, buf);
>        else
>                printk("%.*s", (end - start) + 1, buf);
> 
> was as follows (per Petr's commit message)

OK, thanks, I altered the changelog thusly and scheduled the patch for 4.8:

--- txt/printk-nmi-avoid-direct-printk-s-from-__printk_nmi_flush.txt
+++ txt/printk-nmi-avoid-direct-printk-s-from-__printk_nmi_flush.txt
@@ -3,8 +3,13 @@
 
 __printk_nmi_flush() can be called from nmi_panic(), therefore it has to
 test whether it's executed in NMI context and thus must route the messages
-through deferred printk() or via direct printk(). Except for two places
-where __printk_nmi_flush() does unconditional direct printk() calls:
+through deferred printk() or via direct printk().  This is to avoid
+potential deadlocks, as described in cf9b1106c81c45cde ("printk/nmi: flush
+NMI messages on the system panic").
+
+However there remain two places where __printk_nmi_flush() does
+unconditional direct printk() calls:
+
  - pr_err("printk_nmi_flush: internal error ...")
  - pr_cont("\n")