It appears to be technically possible to use PCMCIA cards on POWER8/9
systems through a PCI Express to PCI adapter and a PCI to
PCMCIA/CardBus adapter. But I can't believe anyone would want to.
So rather than adding a pcmcia-modules package or excluding the
drivers from udebs, disable PCMCIA altogether.
From: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com>
- The value of ppc64 is incorrect for ppc64el (specifically, console=tty0),
rendering the console useless (blank and unresponsive) if no kernel parameters
are specified (e.g., debian-installer netboot).
- It's not needed currently: the kernel correctly autodetects the console
on ppc64el platforms (hvc0 on pseries, hvc0 and hvc1 on powernv).
svn path=/dists/sid/linux/; revision=21943
This config has options for little-endian PowerPC64 systems.
It shares most options with big-endian PowerPC64 systems.
The differences are:
- choice: Endianness selection
Build a little endian kernel.
- choice: Page size
64k pages have benefits (performance et al) over 4k pages on
IBM POWER processors.
The Debian ppc64el port primarily runs on this sort of hardware
and chances are it will also run on hardware based on it (i.e.,
OpenPOWER) [1] [2].
- Maximum number of CPUs
This was increased to 2048 (following pseries_le_defconfig).
For the currently announced systems, the number of CPUs range
between 80-192 (1 or 2 processor module(s) * 10 or 12 cores
per module * 8 threads per core) [3]. This is enough to have
to diverge from CONFIG_NR_CPUS=32 in the other powerpc ports.
For future systems, it's likely larger ones will be announced.
The rationale: consider the announced systems are classified
as 'scale-out' and 'entry-level', plus larger ones have been
historically made available for addressing other markets; and
notice the largest POWER7 server has 1024 CPUs (threads) [4],
and that the threads-per-core doubled from POWER7 to POWER8.
So, the 2048 value is a reasonable 'max' in that projection.
Certainly it is greater than what would be required for most
systems, but I belive 'max' makes sense in that case, if we
are not looking for kernel rebuild and/with different config
for the larger systems (although I would be ok with flavours).
- choice: Default CPUFreq governor
As other architectures, we would prefer the default cpufreq
governor to be 'ondemand'.
The currently available cpufreq driver is for the PowerNV
(non-virtualized) platform, where all processors are available.
In that scenario, statically running at the highest frequency
(specially on idle processors) is not very desireable for the
hardware around (servers), and it is not unlikely for future
hardware (possibly non-servers) to benefit too, considering
that energy savings have been increasingly important on most
environments.
(Note: the powernv-cpufreq driver was introduced only in 3.15;
so, this option has no effect in 3.14; it is harmless.
I can put in patches for enabling this on 3.14 soon.)
- Apple PowerMac based machines
This is being disabled temporarily, until a patch makes upstream
(restricting it to 'depends on !CPU_LITTLE_ENDIAN').
This hardware line has no (known) support for little endian
mode currently, and disabling it has the useful effect of also
disabling a lot of config options which 'depends on PPC_PMAC',
thus saving tens of lines from changing config files.
It indeed has to be disabled because it's enabled by default
('depends on BOOK3S', 'default y'), so even changing it from
config files would not be sufficient.
[1] 'OpenPOWER Foundation Unveils First Innovations and Roadmap'
http://openpowerfoundation.org/press-releases/openpower-foundation-unveils-first-innovations-and-roadmap/
[2] 'POWER8 Reference Board now available for Development!'
http://openpowerfoundation.org/technical/related-links/
[3] 'IBM Power System S812L and S822L'
http://www-03.ibm.com/systems/power/hardware/s812l-s822l/specs.html
[4] 'IBM Power 795 server'
http://www-03.ibm.com/systems/power/hardware/795/perfdata.html
Signed-off-by: Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com>
svn path=/dists/sid/linux/; revision=21422
Ben Hutchings
Ben Hutchings
Bastian Blank