linux-kernel - Re: [PATCH] kdb: Adopt scheduler's task clasification

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: <CAD=FV=Xri+J2=iQzCHLxB+ksT41V6Rexp+BXWi6Fe7=jq3oTFg@mail.gmail.com>
Date:   Thu, 16 Sep 2021 09:28:22 -0700
From:   Doug Anderson <dianders@...omium.org>
To:     Daniel Thompson <daniel.thompson@...aro.org>
Cc:     Jason Wessel <jason.wessel@...driver.com>,
        Xiang wangx <wangxiang@...rlc.com>,
        jing yangyang <jing.yangyang@....com.cn>,
        kgdb-bugreport@...ts.sourceforge.net,
        LKML <linux-kernel@...r.kernel.org>,
        Patch Tracking <patches@...aro.org>
Subject: Re: [PATCH] kdb: Adopt scheduler's task clasification

Hi,

On Thu, Sep 16, 2021 at 8:43 AM Daniel Thompson
<daniel.thompson@...aro.org> wrote:
>
> Currently kdb contains some open-coded routines to generate a summary
> character for each task. This code currently issues warnings, is
> almost certainly broken and won't make any sense to any kernel dev who
> has ever used /proc to examine tasks (D means uninterruptible?).
>
> Fix both the warning and the potential for confusion but adopting the
> scheduler's task clasification. Whilst doing this we also simplify the

s/clasification/classification/


> filtering by using mask strings directly (this means we don't have to
> guess all the characters the scheduler might give us).
>
> Unfortunately we can't quite adopt the scheudler classification it in

s/scheudler/scheduler/


> its entirity because, whilst we can tolerate some changes to the filter

s/entirity/entirety/


> characters, we need to keep I as a means to identify idle CPUs rather than
> system daemons that don't contribute to the load average! Naturally there
> is quite a large comment discussing this.

I'm a bit curious why we're OK with changing other characters but not
'I'. Even if the scheduler use of the character 'I' is a bit
confusing, it still seems like it might be nice to match it just to
avoid confusion. Couldn't we use lowercase 'i' for idle CPUs?
Alternatively beef up the commit message justifying why exactly we
need to keep 'I' as-is.


> Signed-off-by: Daniel Thompson <daniel.thompson@...aro.org>

Worth having a "Fixes" for the patch that introduced the warning?


> @@ -74,7 +74,7 @@ static void kdb_show_stack(struct task_struct *p, void *addr)
>   */
>
>  static int
> -kdb_bt1(struct task_struct *p, unsigned long mask, bool btaprompt)
> +kdb_bt1(struct task_struct *p, const char *mask, bool btaprompt)

In the comment above this function there is still a reference to
"DRSTCZEUIMA". Update that?


> @@ -2300,7 +2298,7 @@ void kdb_ps_suppressed(void)
>  /*
>   * kdb_ps - This function implements the 'ps' command which shows a
>   *     list of the active processes.
> - *             ps [DRSTCZEUIMA]   All processes, optionally filtered by state
> + *             ps [RSDTtXZPIMA]   All processes, optionally filtered by state

What about "U"? What about "E"?


> @@ -2742,7 +2741,7 @@ static kdbtab_t maintab[] = {
>         },
>         {       .name = "bta",
>                 .func = kdb_bt,
> -               .usage = "[D|R|S|T|C|Z|E|U|I|M|A]",
> +               .usage = "[R|S|D|T|t|X|Z|P|I|M|A]",

What about "U"? What about "E"?


> @@ -559,7 +484,6 @@ unsigned long kdb_task_state_string(const char *s)
>   */
>  char kdb_task_state_char (const struct task_struct *p)
>  {
> -       unsigned int p_state;
>         unsigned long tmp;
>         char state;
>         int cpu;
> @@ -568,16 +492,20 @@ char kdb_task_state_char (const struct task_struct *p)
>             copy_from_kernel_nofault(&tmp, (char *)p, sizeof(unsigned long)))
>                 return 'E';
>
> -       cpu = kdb_process_cpu(p);

Don't you still need this? You still have the `cpu` variable and you
still use it in the idle task case.


> -       p_state = READ_ONCE(p->__state);
> -       state = (p_state == 0) ? 'R' :
> -               (p_state < 0) ? 'U' :
> -               (p_state & TASK_UNINTERRUPTIBLE) ? 'D' :
> -               (p_state & TASK_STOPPED) ? 'T' :
> -               (p_state & TASK_TRACED) ? 'C' :
> -               (p->exit_state & EXIT_ZOMBIE) ? 'Z' :
> -               (p->exit_state & EXIT_DEAD) ? 'E' :
> -               (p_state & TASK_INTERRUPTIBLE) ? 'S' : '?';
> +       state = task_state_to_char((struct task_struct *) p);

Casting away constness is fine for now and likely makes this easier to
land, but maybe you can send a patch up to change the API to have
"const" in it?


-Doug