| 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
| ||
|
Date: Mon, 24 Jul 2017 15:41:11 +0530
From: Gautham R Shenoy <ego.lkml@...il.com>
To: Seeteena Thoufeek <s1seetee@...ux.vnet.ibm.com>
Cc: Linux PM list <linux-pm@...r.kernel.org>,
LKML <linux-kernel@...r.kernel.org>, trenn@...e.com,
ego@...ux.vnet.ibm.com,
Vaidyanathan Srinivasan <svaidy@...ux.vnet.ibm.com>
Subject: Re: [PATCH] Fix header-name to read full state name and adjust column width
Hello Seeteena,
Thanks for this useful patch. Couple of suggestions below.
On Thu, Jul 6, 2017 at 12:44 PM, Seeteena Thoufeek
<s1seetee@...ux.vnet.ibm.com> wrote:
> ---Steps to Reproduce---
>
> 1.Execute the cpupower monitor command to fetch all the Idle_Stats values.
>
> root:~# cpupower monitor
> |Idle_Stats
> PKG |CORE|CPU | snoo | stop | stop
> 0| 8| 0| 0.00| 0.00| 2.79
> 0| 8| 1| 0.00| 0.00| 70.68
>
> User interpret it as 2 columns for stop Idle_stats.
>
> root:/sys/devices/system/cpu/cpu8/cpuidle# cat */name
> snooze
> stop0_lite
> stop1_lite
>
> cpupower monitor now shows full state name and adjust colomn width.
>
> root:~/linux/tools/power/cpupower# cpupower monitor
> |Idle_Stats
> PKG |CORE|CPU |snooze|stop0_lite|stop1_lite
> 0| 8| 0| 0.00| 0.00| 99.83
> 0| 8| 1| 0.00| 0.00| 43.82
> 0| 8| 2| 0.00| 0.00| 102.1
> 0| 8| 3| 0.00| 0.00| 96.56
>
> tested on x86 as well.
Can you rewrite the commit log describing the problem and what you do
in this patch to mitigate the problem ?
Something along the lines of the following
"The names of the idle states in the output of cpupower monitor
command are truncated to 4 characters.
Hence, On POWER9, since the states are named "stop0, stop1, stop2,
stop4, stop11" , this output is ambiguous
root:~# cpupower monitor
|Idle_Stats
PKG |CORE|CPU | snoo | stop | stop
0| 8| 0| 0.00| 0.00| 2.79
0| 8| 1| 0.00| 0.00| 70.68
In this patch, we modify the output to print the full state name and
adjust the column width, that results in a
legible output.
root:~/linux/tools/power/cpupower# cpupower monitor
|Idle_Stats
PKG |CORE|CPU |snooze|stop0_lite|stop1_lite
0| 8| 0| 0.00| 0.00| 99.83
0| 8| 1| 0.00| 0.00| 43.82
0| 8| 2| 0.00| 0.00| 102.1
0| 8| 3| 0.00| 0.00| 96.56
Tested on POWER8, POWER9 and x86."
> Signed-off-by: Seeteena Thoufeek <s1seetee@...ux.vnet.ibm.com>
> ---
> .../cpupower/utils/idle_monitor/cpupower-monitor.c | 25 ++++++++++++++--------
> .../cpupower/utils/idle_monitor/cpupower-monitor.h | 2 +-
> 2 files changed, 17 insertions(+), 10 deletions(-)
>
> diff --git a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c
> index 05f953f..5e708d5 100644
> --- a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c
> +++ b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.c
> @@ -88,7 +88,7 @@ void print_header(int topology_depth)
> int state, need_len;
> cstate_t s;
> char buf[128] = "";
> - int percent_width = 4;
> + unsigned int percent_width = 6;
While you are at it, can we have these as #defines? i.e.
#define COL_WIDTH_MIN 6
>
> fill_string_with_spaces(buf, topology_depth * 5 - 1);
> printf("%s|", buf);
> @@ -116,17 +116,18 @@ void print_header(int topology_depth)
> for (mon = 0; mon < avail_monitors; mon++) {
> if (mon != 0)
> printf("|| ");
> - else
> - printf(" ");
> for (state = 0; state < monitors[mon]->hw_states_num; state++) {
> if (state != 0)
> - printf(" | ");
> + printf("|");
> s = monitors[mon]->hw_states[state];
> sprintf(buf, "%s", s.name);
> - fill_string_with_spaces(buf, percent_width);
> + if (strlen(s.name) > percent_width)
> + fill_string_with_spaces(buf, strlen(s.name));
When we have many states (and on POWER9 we can have upto 8 idle
states), printing the full name
may not be advisable. We can instead define COL_WIDTH_MAX along with
COL_WIDTH_MIN and set
the percent_width to the length of the state's name only when that
length is in the interval [COL_WIDTH_MIN, COL_WIDTH_MAX]
Something like the following?
#define COL_WIDTH_MIN 4
#define COL_WIDTH_MAX 8
name_len = strlen(s.name);
if (name_len < COL_WIDTH_MIN)
percent_width = COL_WIDTH_MIN;
else if (name_len > COL_WIDTH_MAX )
percent_width = COL_WIDTH_MAX;
else
percent_width = name_len;
fill_string_with_spaces(buf, percent_width)
> + else
> + fill_string_with_spaces(buf, percent_width);
> +
> printf("%s", buf);
> - }
> - printf(" ");
> + }
> }
> printf("\n");
> }
> @@ -139,7 +140,9 @@ void print_results(int topology_depth, int cpu)
> double percent;
> unsigned long long result;
> cstate_t s;
> -
> + char buf[128] = "";
> + unsigned int percent_width = 6;
> + unsigned int width;
> /* Be careful CPUs may got resorted for pkg value do not just use cpu */
> if (!bitmask_isbitset(cpus_chosen, cpu_top.core_info[cpu].cpu))
> return;
> @@ -163,7 +166,11 @@ void print_results(int topology_depth, int cpu)
> printf("|");
>
> s = monitors[mon]->hw_states[state];
> -
> + if (strlen(s.name) > percent_width) {
> + width = strlen(s.name) - percent_width;
> + fill_string_with_spaces(buf, width);
> + printf("%s", buf);
> + }
> if (s.get_count_percent) {
> ret = s.get_count_percent(s.id, &percent,
> cpu_top.core_info[cpu].cpu);
> diff --git a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h
> index 9e43f33..c9179c6 100644
> --- a/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h
> +++ b/tools/power/cpupower/utils/idle_monitor/cpupower-monitor.h
> @@ -15,7 +15,7 @@
>
> #define MONITORS_MAX 20
> #define MONITOR_NAME_LEN 20
> -#define CSTATE_NAME_LEN 5
> +#define CSTATE_NAME_LEN 16
So, this is based on the CPUIDLE_NAME_LEN I suppose. You might want to
add a comment
about it here.
> #define CSTATE_DESC_LEN 60
>
> int cpu_count;
> --
> 1.8.3.1
>
Otherwise the patch looks good.
Reviewed-by: Gautham R. Shenoy <ego@...ux.vnet.ibm.com>
--
Thanks and Regards
gautham.
Powered by blists - more mailing lists