gmane.linux.ports.ppc64.devel

Re: [PATCH] powerpc/kvm: Handle transparent hugepage in KVM

Michael Neuling — 2013-06-19T23:59:22

OK, maybe you can update to reflect that.


How about something like this then?

while (1) {
      if (unlikely(old_pte & _PAGE_BUSY))
    continue;
.....
      if cmpxchg(foo)
       break;
}






Sure.

Mikey

Re: DEBUG_PAGEALLOC on PPC not working (kernels 2.6-25, 3.0-34)

Scott Wood — 2013-06-19T21:00:46

This is not supported on booke -- the tlbcam mapping is required for  
all lowmem.


For e300, I think I had it working at one point a few years ago (see  
commit bde6c6e16aa489ea76c762fb7ffb0abb48660dd8).

The reason we can do it on e300 and not on booke is because e300 takes  
exceptions in real mode.  On e500 the MMU is always enabled, so we need  
bolted TLB1 entries that cover at least all exception code (up to the  
point where a TLB miss could safely be taken) and all page tables (in  
practice, we just bolt all lowmem) and other data that can be  
referenced from said exception code.  There are not enough TLB1 entries  
to do this on a fine-grained basis.

-Scott

Re: Pull request: scottwood/linux.git for-3.10

Scott Wood — 2013-06-19T16:50:30

That's the patch I'm asking you to pull. :-P

-Scott

Re: [PATCH 1/2] perf tools: fix a typo of a Power7 event name

Sukadev Bhattiprolu — 2013-06-19T16:13:12

| In the Power7 PMU guide:
| https://www.power.org/documentation/commonly-used-metrics-for-performance-analysis/
| PM_BRU_MPRED is referred to as PM_BR_MPRED.
| 
| This patch fix the typo by changing the name of the event in kernel and
| documentation accordingly.
| 
| Signed-off-by: Runzhen Wang <runzhen< at >linux.vnet.ibm.com>

Reviewed-by: Sukadev Bhattiprolu <sukadev< at >linux.vnet.ibm.com>

Re: Pull request: scottwood/linux.git for-3.10

Kumar Gala — 2013-06-19T15:06:38

On Jun 18, 2013, at 3:14 PM, Scott Wood wrote:


What about Rohit's patch: powerpc/pci: Fix setup of Freescale PCI / PCIe controllers?  Seems like also a fix for 3.10

- k

Re: Pull request: scottwood/linux.git for-3.10

Kumar Gala — 2013-06-19T15:06:38

On Jun 18, 2013, at 3:14 PM, Scott Wood wrote:


What about Rohit's patch: powerpc/pci: Fix setup of Freescale PCI / PCIe controllers?  Seems like also a fix for 3.10

- k

DEBUG_PAGEALLOC on PPC not working (kernels 2.6-25, 3.0-34)

saikia.partha — 2013-06-19T13:09:44

Hi,

We have two Freescale PowerPC boards on which we're trying to enable
DEBUG_PAGEALLOC with the hope that we'll see an exception whenever some
code tries to modify a page that's been already freed. To test it, we wrote
this sample code -

===
#define BUF_SIZE    4096
void
pagealloc_test(void)
{
    char *buf = kmalloc(BUF_SIZE, GFP_KERNEL);

    if (!buf) {
        printk("%s[%d] - alloc failed!\n", __func__, __LINE__);
        return;
    }
    printk("%s[%d] - alloc'd\n", __func__, __LINE__);
    memset(&buf[0], 0, BUF_SIZE);
    printk("%s[%d] - memset'd\n", __func__, __LINE__);
    kfree(buf);
    printk("%s[%d] - free'd\n", __func__, __LINE__);
    memset(&buf[0], 1, BUF_SIZE);
    printk("%s[%d] - memset'd after free!\n", __func__, __LINE__);
}
===

Here, the last memset() should generate an exception if PAGEALLOC code
correctly unmapped the page during kfree(). However, kernel is happily
running after the memset post-free. Any clue?
Also, the 2nd board has Book-E which has a different MMU architect

DEBUG_PAGEALLOC on PPC not working (kernels 2.6-25, 3.0-34)

saikia.partha — 2013-06-19T12:58:27

Hi,

We have two Freescale PowerPC boards on which we're trying to enable
DEBUG_PAGEALLOC with the hope that we'll see an exception whenever some
code tries to modify a page that's been already freed. To test it, we wrote
this sample code -

===
#define BUF_SIZE    4096
void
pagealloc_test(void)
{
    char *buf = kmalloc(BUF_SIZE, GFP_KERNEL);

    if (!buf) {
        printk("%s[%d] - alloc failed!\n", __func__, __LINE__);
        return;
    }
    printk("%s[%d] - alloc'd\n", __func__, __LINE__);
    memset(&buf[0], 0, BUF_SIZE);
    printk("%s[%d] - memset'd\n", __func__, __LINE__);
    kfree(buf);
    printk("%s[%d] - free'd\n", __func__, __LINE__);
    memset(&buf[0], 1, BUF_SIZE);
    printk("%s[%d] - memset'd after free!\n", __func__, __LINE__);
}
===

Here, the last memset() should generate an exception if PAGEALLOC code
correctly unmapped the page during kfree(). However, kernel is happily
running after the memset post-free. Any clue?
Also, the 2nd board has Book-E which has a different MMU architect

Re: [PATCH] powerpc/kvm: Handle transparent hugepage in KVM

Aneesh Kumar K.V — 2013-06-19T12:30:43


In a way yes. Instead of lock and read as it was before, it is now done
via cmpxchg which still use ldarx/stdcx



No that would be wrong. (I did that in an earlier version :).We really
don't want the below cmpxchg to run if we find _PAGE_BUSY.


it is actually the shift bits represending hugepage size. We set it to 0
if we don't find hugepage in find_linux_pte_or_hugepte. May be something
like hugepage_shift is better ?

-aneesh

Re: [PATCH 05/31] powerpc/eeh: Trace PCI bus from PE

Gavin Shan — 2013-06-19T10:20:47

Please use the updated [5/31] in the attachment.

Thanks,
Gavin

_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev< at >lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev

Re: [PATCH] powerpc/THP: Wait for all hash_page calls to finish before invalidating HPTE entries

Benjamin Herrenschmidt — 2013-06-19T10:18:33


Right, this is expected. This fix is a result of me asking him to do it
that way on ST and should eventually be folded in the series (or
re-ordered to be in there before the patch that enables THP).

Cheers,
Ben.

[PATCH 4/4] powerpc: make the kernel bootable from non 0 address for 6xx

Kevin Hao — 2013-06-19T09:20:04

Add the support to boot the kernel from a non 0 address for 6xx.
Setup the exception trampoline if the physical start address is
not 0.

For a kdump kernel, enable the relocatable support implicitly.
Since the memstart_adddr of the kdump is not 0, we definitely
should regard this when setting up the BAT map.

Signed-off-by: Kevin Hao <haokexin< at >gmail.com>
---
 arch/powerpc/Kconfig                            |  2 +-
 arch/powerpc/include/asm/exception_trampoline.h |  4 ++--
 arch/powerpc/kernel/Makefile                    |  3 ++-
 arch/powerpc/kernel/exception_trampoline.c      | 18 +++++++++++++++---
 arch/powerpc/mm/ppc_mmu_32.c                    |  7 +------
 5 files changed, 21 insertions(+), 13 deletions(-)

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index 8fe2792..6e03028 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
< at >< at > -382,7 +382,7 < at >< at > config KEXEC
 config CRASH_DUMP
 bool "Build a kdump crash kernel"
 depends on PPC64 || 6xx || FSL_BOOKE || (44x && !SMP)
-select RELOCA

[PATCH 3/4] powerpc: s/kdump/exception/ for the exception trampoline functions

Kevin Hao — 2013-06-19T09:20:03

These functions are not just kdump specific. Replace the 'kdump' with
the 'exception' to make them more general.

Signed-off-by: Kevin Hao <haokexin< at >gmail.com>
---
 arch/powerpc/include/asm/exception_trampoline.h | 20 ++++++++++----------
 arch/powerpc/kernel/exception_trampoline.c      | 14 +++++++-------
 arch/powerpc/kernel/prom.c                      |  2 +-
 arch/powerpc/kernel/setup_32.c                  |  2 +-
 arch/powerpc/kernel/setup_64.c                  |  2 +-
 5 files changed, 20 insertions(+), 20 deletions(-)

diff --git a/arch/powerpc/include/asm/exception_trampoline.h b/arch/powerpc/include/asm/exception_trampoline.h
index 707ad6c..88281c9 100644
--- a/arch/powerpc/include/asm/exception_trampoline.h
+++ b/arch/powerpc/include/asm/exception_trampoline.h
< at >< at > -1,10 +1,10 < at >< at >
 #ifndef _EXCEPTION_TRAMPOLINE_H
 #define _EXCEPTION_TRAMPOLINE_H
 
-/* How many bytes to reserve at zero for kdump. The reserve limit should
+/* How many bytes to reserve at zero for exception. The reserve limit should
  *

[PATCH 2/4] powerpc: move the exception trampoline helper functions to a separate file

Kevin Hao — 2013-06-19T09:20:02

For some platform such as 6xx we need to setup the exception
trampoline for a relocatable kernel even through the CONFIG_CRASH_DUMP
is disabled. So move these functions to a separate file so they can
be used by non dump kernel.  This patch doesn't introduce any function
change.

Signed-off-by: Kevin Hao <haokexin< at >gmail.com>
---
 arch/powerpc/include/asm/exception_trampoline.h | 35 +++++++++++++
 arch/powerpc/include/asm/kdump.h                | 32 -----------
 arch/powerpc/kernel/Makefile                    |  2 +-
 arch/powerpc/kernel/crash_dump.c                | 41 ---------------
 arch/powerpc/kernel/exception_trampoline.c      | 70 +++++++++++++++++++++++++
 arch/powerpc/kernel/prom.c                      |  2 +-
 arch/powerpc/kernel/setup_32.c                  |  1 +
 arch/powerpc/kernel/setup_64.c                  |  2 +-
 8 files changed, 109 insertions(+), 76 deletions(-)
 create mode 100644 arch/powerpc/include/asm/exception_trampoline.h
 create mode 100644 arch/powerpc/kernel/exception_trampoline.

[PATCH 1/4] powerpc: enable relocatable support for 6xx

Kevin Hao — 2013-06-19T09:20:01

This is based on the codes in head_44x.S. With this patch the kernel
can only boot from 0 with CONFIG_RELOCATABLE enabled. We will add the
support to boot from a non 0 address in the following patches.

Signed-off-by: Kevin Hao <haokexin< at >gmail.com>
---
 arch/powerpc/Kconfig                   |   2 +-
 arch/powerpc/include/asm/page.h        |   2 +-
 arch/powerpc/kernel/head_32.S          | 103 +++++++++++++++++++++++++++++++++
 arch/powerpc/kernel/prom_init_check.sh |   2 +-
 4 files changed, 106 insertions(+), 3 deletions(-)

diff --git a/arch/powerpc/Kconfig b/arch/powerpc/Kconfig
index c33e3ad..8fe2792 100644
--- a/arch/powerpc/Kconfig
+++ b/arch/powerpc/Kconfig
< at >< at > -866,7 +866,7 < at >< at > config DYNAMIC_MEMSTART
 
 config RELOCATABLE
 bool "Build a relocatable kernel"
-depends on ADVANCED_OPTIONS && FLATMEM && 44x
+depends on ADVANCED_OPTIONS && FLATMEM && (44x || 6xx)
 select NONSTATIC_KERNEL
 help
   This builds a kernel image that is capable of running at the
diff --git a/arch/powerpc/include/asm/page.h

[PATCH 0/4] powerpc: enable relocatable support for 6xx

Kevin Hao — 2013-06-19T09:20:00

This patch series enables the relocatable support for 6xx boards.
With these patches:
  * the kernel can boot from any address between 0x10000 ~ 0x2000000
  * kdump is workable
  * a single kernel image can be used as boot or kdump kernel

Boot test on a mpc8260 board. Also passed the build test for the
following configurations:
ppc40x_defconfig
        ppc64e_defconfig
        ppc64_defconfig
        corenet32_smp_defconfig
        corenet64_smp_defconfig
        ppc44x_defconfig
pmac32_defconfig
pq2fads_defconfig
mpc5200_defconfig
        pseries_defconfig

---
Kevin Hao (4):
  powerpc: enable relocatable support for 6xx
  powerpc: move the exception trampoline helper functions to a separate
    file
  powerpc: s/kdump/exception/ for the exception trampoline functions
  powerpc: make the kernel bootable from non 0 address for 6xx

 arch/powerpc/Kconfig                            |   4 +-
 arch/powerpc/include/asm/exception_trampoline.h |  35 ++++++++
 arch/powerpc/include/asm/kdump.h                |

Re: [PATCH 05/31] powerpc/eeh: Trace PCI bus from PE

Gavin Shan — 2013-06-19T08:48:52

.../...


Thanks, Mike. Good catch actually, it won't have side-effect and
possiblly avoid problems during hot-plug: For PCI bus with only
child PCI device, we won't return the PCI bus (without fix) and
the EEH core doesn't do hot-plug on the affected PCI bus then. 

I'm testing on the updated patch and will send that out soon if it
works well.

Thanks,
Gavin

Re: [PATCH 01/31] powerpc/eeh: Move common part to kernel directory

Gavin Shan — 2013-06-19T07:29:54

Please use the updated [1/31] in the attachment.

Thanks,
Gavin

_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev< at >lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev

Re: [PATCH 05/31] powerpc/eeh: Trace PCI bus from PE

Mike Qiu — 2013-06-19T07:21:07

于 2013/6/18 16:33, Gavin Shan 写道:
Hi Gavin

I have qestion, can we keep pe->bus for a device pe ? the value is the 
bus which edev belongs to.

so that we can make the code more efficient for device pe.

I have no idea of whether this will cause side effect

Thanks
Mike

_______________________________________________
Linuxppc-dev mailing list
Linuxppc-dev< at >lists.ozlabs.org
https://lists.ozlabs.org/listinfo/linuxppc-dev

Re: [PATCH] powerpc/THP: Wait for all hash_page calls to finish before invalidating HPTE entries

Michael Neuling — 2013-06-19T07:14:56





Can you make this more obvious in the changelog (as well as making it 80
col).  I don't see 'bug' mentioned anywhere.  'Fix' is mentioned
somewhere in the middle of the changelog.


OK, but V10 THP is not in yet, right?  So why not roll it into that
series rather than pushing broken stuff and fixing it?

Mikey

Re: [PATCH] powerpc/kvm: Handle transparent hugepage in KVM

Michael Neuling — 2013-06-19T07:11:42


Can you expand on this more please.  There are a lot of details below
like removing a ldarx/stdcx loop that should be better described here.


This is comment still valid now the ldarx/stdcx is gone?  


continue here?  Please don't create looping primitives.
  

Comment looks much like the code... seems redundant.


Whitespace


'shift' goes into the new 'hugepage' parameter?  Doesn't seem logical?
Can we harmonise the name to make it less confusing?

Mikey