Re: [PATCH V16 09/12] MIPS: Loongson: Add Loongson-3 Kconfig options

[Aurelien Jarno <aurelien@aurel32.net>]

On Mon, Jan 13, 2014 at 06:15:26PM +0800, Huacai Chen wrote:
> On Mon, Jan 13, 2014 at 12:30 PM, Aurelien Jarno <aurelien@aurel32.net> wrote:
> > On Mon, Jan 13, 2014 at 11:02:19AM +0800, Huacai Chen wrote:
> >> On Sat, Jan 11, 2014 at 11:24 PM, Aurelien Jarno <aurelien@aurel32.net> wrote:
> >> > On Thu, Jan 09, 2014 at 01:07:30PM +0000, Alex Smith wrote:
> >> >> On 08/01/14 02:44, Huacai Chen wrote:
> >> >> >Added Kconfig options include: Loongson-3 CPU and machine definition,
> >> >> >CPU cache features, UEFI-like firmware interface (LEFI), HT-linked PCI,
> >> >> >and big memory support.
> >> >> >
> >> >> >Signed-off-by: Huacai Chen <chenhc@lemote.com>
> >> >> >Signed-off-by: Hongliang Tao <taohl@lemote.com>
> >> >> >Signed-off-by: Hua Yan <yanh@lemote.com>
> >> >> >---
> >> >> >  arch/mips/Kconfig           |   29 +++++++++++++++++++++++++++-
> >> >> >  arch/mips/loongson/Kconfig  |   44 +++++++++++++++++++++++++++++++++++++++++++
> >> >> >  arch/mips/loongson/Platform |    1 +
> >> >> >  3 files changed, 73 insertions(+), 1 deletions(-)
> >> >> >
> >> >> >diff --git a/arch/mips/Kconfig b/arch/mips/Kconfig
> >> >> >index 17cc7ff..513e941 100644
> >> >> >--- a/arch/mips/Kconfig
> >> >> >+++ b/arch/mips/Kconfig
> >> >> >@@ -1487,6 +1487,18 @@ config CPU_LOONGSON2
> >> >> >     select CPU_SUPPORTS_HIGHMEM
> >> >> >     select CPU_SUPPORTS_HUGEPAGES
> >> >> >
> >> >> >+config CPU_LOONGSON3
> >> >> >+    bool "Loongson 3 CPU"
> >> >> >+    depends on SYS_HAS_CPU_LOONGSON3
> >> >> >+    select CPU_SUPPORTS_64BIT_KERNEL
> >> >> >+    select CPU_SUPPORTS_HIGHMEM
> >> >> >+    select CPU_SUPPORTS_HUGEPAGES
> >> >> >+    select WEAK_ORDERING
> >> >> >+    select WEAK_REORDERING_BEYOND_LLSC
> >> >> >+    help
> >> >> >+            The Loongson 3 processor implements the MIPS III instruction set
> >> >> >+            with many extensions.
> >> >> >+
> >> >>
> >> >> This should be moved into the "CPU type" choice block. Currently
> >> >> this appears as an option outside of that choice, and so it is
> >> >> possible to build a kernel without any CPU type selected.
> >> >>
> >> >> Also, as Aurelien said on the previous version, shouldn't the
> >> >> comment be changed to MIPS64R2 rather than MIPS III even if you
> >> >> aren't selecting the MIPS64 CPU types for the time being? Note, if
> >> >> you replied to that comment I may not have seen it - I think the
> >> >> list is filtering your replies because they're HTML emails.
> >> >>
> >> >> >  config CPU_LOONGSON1
> >> >> >     bool
> >> >> >     select CPU_MIPS32
> >> >> >@@ -1513,6 +1526,10 @@ config SYS_HAS_CPU_LOONGSON2F
> >> >> >     select CPU_SUPPORTS_ADDRWINCFG if 64BIT
> >> >> >     select CPU_SUPPORTS_UNCACHED_ACCELERATED
> >> >> >
> >> >> >+config SYS_HAS_CPU_LOONGSON3
> >> >> >+    bool
> >> >> >+    select CPU_SUPPORTS_CPUFREQ
> >> >> >+
> >> >> >  config SYS_HAS_CPU_LOONGSON1B
> >> >> >     bool
> >> >> >
> >> >> >@@ -1703,7 +1720,7 @@ choice
> >> >> >
> >> >> >  config PAGE_SIZE_4KB
> >> >> >     bool "4kB"
> >> >> >-    depends on !CPU_LOONGSON2
> >> >> >+    depends on !CPU_LOONGSON2 && !CPU_LOONGSON3
> >> >> >     help
> >> >> >      This option select the standard 4kB Linux page size.  On some
> >> >> >      R3000-family processors this is the only available page size.  Using
> >> >> >@@ -2373,6 +2390,16 @@ config PCI
> >> >> >       your box. Other bus systems are ISA, EISA, or VESA. If you have PCI,
> >> >> >       say Y, otherwise N.
> >> >> >
> >> >> >+config HT_PCI
> >> >> >+    bool "Support for HT-linked PCI"
> >> >> >+    depends on CPU_LOONGSON3
> >> >> >+    select PCI_DOMAINS
> >> >> >+    help
> >> >> >+      Loongson family machines use Hyper-Transport bus for inter-core
> >> >> >+      connection and device connection. The PCI bus is a subordinate
> >> >> >+      linked at HT. Choose Y unless you are using Loongson 2E/2F based
> >> >> >+      machines.
> >> >> >+
> >> >>
> >> >> Should this default to y, given that it is selected in the
> >> >> defconfig? Also, the comment referring to 2E/2F is redundant given
> >> >> that it depends on CPU_LOONGSON3 and cannot be selected on those
> >> >> machines.
> >> >
> >> > I am still not sure we need such an option. This option only changes an
> >> > address in arch/mips/include/asm/mach-loongson/loongson.h:
> >> >
> >> > | #if defined(CONFIG_HT_PCI)
> >> > | #define LOONGSON_PCIIO_BASE     loongson_sysconf.pci_io_base
> >> > | #else
> >> > | #define LOONGSON_PCIIO_BASE     0x1fd00000
> >> > | #endif
> >> >
> >> > If both options are working on a Loongson 3 machine, it's definitely a
> >> > good idea to leave the choice to the user. That said if only the address
> >> > from loongson_sysconf.pci_io_base works on a Loongson 3 machine, this
> >> > option must always be selected and is therefore useless.
> >> HT_PCI is needed, because Loongson-3 device structure looks like this:
> >> Loongson3-Core -----　　　　　  ---- MEM
> >>　　　　　　　　　 |--- Cache ---|
> >> HT1-controller  -------　　　　　  ---- PCI-Controller --- PCI-Devices (PATH-B)
> >>     |
> >>     |
> >>  PCI-Controller
> >>     |
> >>     |
> >>  PCI-Devices (PATH-A)
> >>
> >> If PCI devices are linked via PATH-A, we should select HT_PCI;
> >> otherwise we should unselected HT_PCI.
> >
> > Thanks for the explanation, it looks now more clear why we need this
> > option. Both cases are valid, it just depends on where the PCI
> > controller is connected.
> >
> > That said it means that we won't be able to use the same kernel on
> > machines using the PATH-A and on machines using the PATH-B. Is there
> > a way to determine this at runtime through LEFI?
> At present we can't determine PATH-A/B through LEFI, and LEFI is
> surely need to extend at some time. Most of machines use PATH-A, since
> in this way DMA coherency is maintained by hardware. If use PATH-B,
> both HT_PCI and LEFI should be disabled.

As far as I understand, the current patchset assumes DMA coherency in
all the patches. It means that even if this option is disabled to use
the PATH-B instead, it won't work correctly.

Therefore I think that this option should not be given to the user,
either LEMOTE_MACH3A should automatically select it, or it should simply
be removed. Of course it could be reintroduced in a later patchset
adding full support for machines having their PCI connected through
PATH-B.

Aurelien

-- 
Aurelien Jarno	                        GPG: 1024D/F1BCDB73
aurelien@aurel32.net                 http://www.aurel32.net