Thread (5 messages) flat view 5 messages, 3 authors, 2007-01-26

Re: [PATCH 6/14] ps3: smp interrupt fixes

From: Geoff Levand <hidden>
Date: 2007-01-26 02:47:43

Arnd Bergmann wrote:
On Thursday 25 January 2007 03:40, Geoff Levand wrote:
quoted
--- ps3-linux-dev.orig/arch/powerpc/platforms/ps3/interrupt.c
+++ ps3-linux-dev/arch/powerpc/platforms/ps3/interrupt.c
@@ -317,22 +317,23 @@ struct ps3_bmp {
=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=
=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BDu=
nsigned long unused_1[3];
quoted
=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=
=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BDu=
nsigned long mask;
quoted
=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=
=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BDu=
nsigned long unused_2[3];
quoted
-=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD} __at=
tribute__ ((packed));
quoted
+=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD} __at=
tribute__ ((aligned (64)));
=20
Is that meant to be a 64 _byte_ alignment? By default, it is
already aligned to 64 bit.
If 64 byte is some magic requirement by the hypervisor, you
might want to make that explicit by using a constant definition
for it.

The status + mask (512 bits) can't cross a page boundary, as required
by the HV.  That is the intent of the aligned (64) attribute.

Sorry, I don't quite understand what you mean by 'using a constant defini=
tion
for it'.  Could you elaborate?

As you suggested I have already changed this to:

struct ps3_bmp {
	struct {
		u64 status;
		u64 unused_1[3];
		u64 mask;
		u64 unused_2[3];
	};
	u64 ipi_debug_brk_mask;
	spinlock_t lock;
};

struct ps3_private {
	struct ps3_bmp bmp __attribute__ ((aligned (64)));
	u64 node;
	unsigned int cpu;
};

quoted
+=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BD=EF=BF=BDcur_cp=
u_spec->cpu_features |=3D CPU_FTR_SMT;
quoted
=20
CPU_FTRS_CELL already contains CPU_FTR_SMT, why do you need to enable
it explicitly?

Yes, I questioned whether that was needed or not.  I'll remove it.

-Geoff
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help