linux-kernel - Re: [PATCH 4.19 regression fix] printk: For early boot messages check loglevel when flushing the buffer

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]

Message-ID: <20180912074930.GA470@jagdpanzerIV>
Date:   Wed, 12 Sep 2018 16:49:30 +0900
From:   Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>
To:     Petr Mladek <pmladek@...e.com>
Cc:     Sergey Senozhatsky <sergey.senozhatsky.work@...il.com>,
        Hans de Goede <hdegoede@...hat.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Ingo Molnar <mingo@...hat.com>,
        "H . Peter Anvin" <hpa@...or.com>,
        Sergey Senozhatsky <sergey.senozhatsky@...il.com>,
        x86@...nel.org, linux-kernel@...r.kernel.org,
        Steven Rostedt <rostedt@...dmis.org>,
        Maninder Singh <maninder1.s@...sung.com>
Subject: Re: [PATCH 4.19 regression fix] printk: For early boot messages
 check loglevel when flushing the buffer

On (09/11/18 10:47), Petr Mladek wrote:
> > Oh, that was intentional. I consider repeated messages to be less
> > problematic than the missing ones.
> 
> It makes some sense. But it might be problematic with slow consoles.

Right. And this is why I brought up Jan's patch.
And I agree that we better discuss that approach in a separate thread.

> > > Most problems should probably be solved when we store console_seq
> > > before setting exclusive_console. Then we could clear
> > > exclusive_console when reaching the stored sequence number.
> > > 
> > > Can this be that simple? ;-)
> > 
> > This can work, yes.
> 
> I'll try to cook up a patch later this week.

Hmm, wouldn't the same "it might be problematic with slow consoles"
argument apply to this solution as well? If we flush_on_panic(), and
have a slow exclusive_console then flushing logbuf to that slow
exclusive_console can take quite some time, before we even begin to
flush oops messages.

So maybe we can do the following:

- store console_seq when we register exclusive console
- in flush_on_panic, if we have exclusive console set
    - clear exclusive console
    - rollback console_seq to the value it had before we set exclusive console

So we will not re-flush all logbuf messages, but only the new ones and
should get to oops messages sooner.

IOW, something like this?

---

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 53c94cbce0af..4ef199572df7 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -424,6 +424,7 @@ static u32 log_next_idx;
 /* the next printk record to write to the console */
 static u64 console_seq;
 static u32 console_idx;
+static u64 rollback_console_seq;
 
 /* the next printk record to read after the last 'clear' command */
 static u64 clear_seq;
@@ -2592,6 +2593,10 @@ void console_flush_on_panic(void)
 	 */
 	console_trylock();
 	console_may_schedule = 0;
+	if (exclusive_console) {
+		exclusive_console = NULL;
+		console_seq = rollback_console_seq;
+	}
 	console_unlock();
 }
 
@@ -2793,6 +2798,7 @@ void register_console(struct console *newcon)
 		 * for us.
 		 */
 		logbuf_lock_irqsave(flags);
+		rollback_console_seq = console_seq;
 		console_seq = syslog_seq;
 		console_idx = syslog_idx;
 		logbuf_unlock_irqrestore(flags);

---

	-ss