Re: [PATCH] x86_64: simplify the memtest parameter setting

[Stefan Richter <stefanr@s5r6.in-berlin.de>]

> On Fri, Apr 18, 2008 at 7:15 PM, Frans Pop <elendil@planet.nl> wrote:
>>  the new description is probably too short.
>>  The phrase "allows memtest to be set" does not really mean anything to a
>>  user who does not know what memtest really does. It also does not explain
>>  that memtest is a boot parameter.
>>
>>  Also, it should probably be made more explicit that no memtest is actually
>>  performed unless the memtest boot parameter is passed at the boot prompt.
>>
>>  Looks like 4 is the maximum possible value. This should be mentioned
>>  explicitly (here as well as in Documentation/kernel-parameters.txt).

To add to this:  The help text should IMO contain
   - that a test for faulty RAM is being added if Y,
   - that this feature can be controlled by the boot parameter
     (and, per the parameter's default, is normally off),
   - how many loops with which data patterns are performed depending
     on the boot parameter value,
   - how long this might take (so that nobody who hears of this for
     the first time enables it e.g. on a server which is supposed to
     not take too long to reboot),
   - how reliable the results of this test are (I presume that some
     faults can only be found by many repeated re-runs of the memory
     tester),
   - that saying Y slightly (?) increases the size of the kernel image
     but does not influence the kernel size after boot-up is completed.

Thanks,
-- 
Stefan Richter
-=====-==--- -=-- =--==
http://arcgraph.de/sr/