summaryrefslogtreecommitdiffstats
path: root/arch/sparc/mm
diff options
context:
space:
mode:
authorBob Breuer <breuerr@mc.net>2006-03-23 22:36:19 -0800
committerDavid S. Miller <davem@davemloft.net>2006-03-23 22:36:19 -0800
commita54123e27779049d27d21e6c8adfee73aa2c0734 (patch)
tree265849e706e4ebe3b75127ebe6e3cbfe2a78850a /arch/sparc/mm
parent674a396c6d2ba0341ebdd7c1c9950f32f018e2dd (diff)
downloadop-kernel-dev-a54123e27779049d27d21e6c8adfee73aa2c0734.zip
op-kernel-dev-a54123e27779049d27d21e6c8adfee73aa2c0734.tar.gz
[SPARC]: Try to start getting SMP back into shape.
Todo items: - IRQ_INPROGRESS flag - use sparc64 irq buckets, or generic irq_desc? - sun4d - re-indent large chunks of sun4m_smp.c - some places assume sequential cpu numbering (i.e. 0,1 instead of 0,2) Last I checked (with 2.6.14), random programs segfault with dual HyperSPARC. And with SuperSPARC II's, it seems stable but will eventually die from a write lock error (wrong lock owner or something). I haven't tried the HyperSPARC + highmem combination recently, so that may still be a problem. Signed-off-by: David S. Miller <davem@davemloft.net>
Diffstat (limited to 'arch/sparc/mm')
-rw-r--r--arch/sparc/mm/srmmu.c6
1 files changed, 6 insertions, 0 deletions
diff --git a/arch/sparc/mm/srmmu.c b/arch/sparc/mm/srmmu.c
index 27b0e0b..58c65cc 100644
--- a/arch/sparc/mm/srmmu.c
+++ b/arch/sparc/mm/srmmu.c
@@ -1302,7 +1302,12 @@ void __init srmmu_paging_init(void)
flush_cache_all();
srmmu_set_ctable_ptr((unsigned long)srmmu_ctx_table_phys);
+#ifdef CONFIG_SMP
+ /* Stop from hanging here... */
+ local_flush_tlb_all();
+#else
flush_tlb_all();
+#endif
poke_srmmu();
#ifdef CONFIG_SUN_IO
@@ -1419,6 +1424,7 @@ static void __init init_vac_layout(void)
max_size = vac_cache_size;
if(vac_line_size < min_line_size)
min_line_size = vac_line_size;
+ //FIXME: cpus not contiguous!!
cpu++;
if (cpu >= NR_CPUS || !cpu_online(cpu))
break;
OpenPOWER on IntegriCloud