linux-kernel - Re: Why is /sys/class/power_supply/CMB1/energy

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: <20080904194058.GA24955@dhcp35.suse.cz>
Date:	Thu, 4 Sep 2008 13:40:58 -0600
From:	Michal Hocko <mstsxfx@...il.com>
To:	power@...host.org
Cc:	Alexey Starikovskiy <astarikovskiy@...e.de>, arjan@...ux.intel.com,
	linux-kernel@...r.kernel.org
Subject: Re: Why is /sys/class/power_supply/CMB1/energy_now not exported?

On Mon, Jul 21, 2008 at 09:04:13PM -0600, Michal Hocko wrote:
> [fixing bad linux kernel mailing list email address - sorry, but gmail's
> thrown mailing delivery error to the spam]
> 
> On Tue, Jul 22, 2008 at 01:03:11AM +0400, Alexey Starikovskiy wrote:
> > Hi Michal,
> 
> Hi Alexey,
> 
> >
> > the charge_now part of your patch seems to be correct, but the
> > calculation is somehow broken --
> 
> I have made some experiments and the result is quite surprising to me.
> It seems that this problem is somehow related to Debian distribution
> kernel, because I wasn't able to reproduce it with the Vanilla kernel (I
> had to screw something up last time when I have checked that and
> reported in the original message).
> 
> I have tried 3 configurations:
> * 2.6.25 - Vanilla kernel with config-2.6.25 config and result output in
>   powertop-2.6.25-good
> * 2.6.25-2 - Debian distribution kernel with config-2.6.25-2 config and
>   powertop-2.6.25-2-bad output
> * 2.6.26 - Vanilla kernel with config-2.6.26 config and result output in
>   powertop-2.6.26-good
[...]

I had some time to get back to this issue and the problem why I have
seen different behavior on distribution and vanilla kernel was in
configuration. 
While distribution kernel doesn't use CONFIG_ACPI_PROCFS_POWER=y while
my vanilla kernel configuration does! This means that the patched code
wasn't triggered in my vanilla kernel testing.

So that I have checked the print_batter_sysfs again and it seems that
the problem is on another place:
powertop.c:648
	sprintf(filename, "/sys/class/power_supply/%s/current_now", dirent->d_name);
	file = fopen(filename, "r");
	if (!file)
		continue;
	memset(line, 0, 1024);
	if (fgets(line, 1024, file) != NULL) {
		watts_drawn = strtoull(line, NULL, 10) / 1000000.0;
	}
	fclose(file);

	if (!dontcount) {
		rate += watts_drawn + voltage * amperes_drawn;
	}

current rate (later used for the remaining time calculation:
display.c:201
	sprintf(buffer, _("Power usage (ACPI estimate): %3.1fW (%3.1f hours)"), rate, cap/rate);
)

is calculated as watts_drawn + voltage which is not correct, because
current_* values are in A and not in W (according to
linux-src/Documentation/power/power_supply_class.txt).

Updated patch is attached. This one is already working for me.

-- 
Michal Hocko

View attachment "charge-read.patch" of type "text/x-diff" (1816 bytes)