sparc64: ultra 60 fails to boot with 64kB pages
From: Mariusz Kozlowski
Date: Wed Jan 09 2008 - 14:14:35 EST
Hello,
I found this kconfig info on kernel page size:
'8KB and 64KB work quite well, since Sparc ELF sections provide for up to 64KB alignment.'
so I thought I'll give it a try. I tested 2.6.24-rc7 and it stopped right after
'Remapping the kernel ... done'
At first I wanted to bisect this but soon I found there is no kernel that boots with
64k pages on that machine. I tested 2.6.24-rc7, 2.6.23, 2.6.22, 2.6.20 and 2.6.16.
2.6.24-rc7 dies right after 'Remapping the kernel ... done' text. So I added:
#define DBGRK() prom_printf("%s:%d\n", __FILE__, __LINE__);
and ivoked it a bunch of times in arch/sparc64/mm/init.c to see where it stops.
Here is main portion of the diff:
--- a/arch/sparc64/mm/init.c
+++ b/arch/sparc64/mm/init.c
@@ -1328,6 +1328,7 @@ pgd_t swapper_pg_dir[2048];
static void sun4u_pgprot_init(void);
static void sun4v_pgprot_init(void);
+#define DBGRK() prom_printf("%s:%d\n", __FILE__, __LINE__);
void __init paging_init(void)
{
unsigned long end_pfn, pages_avail, shift, phys_base;
@@ -1402,30 +1403,38 @@ void __init paging_init(void)
swapper_low_pmd_dir + (shift / sizeof(pgd_t)));
inherit_prom_mappings();
-
+
read_obp_memory("reg", &pall[0], &pall_ents);
init_kpte_bitmap();
/* Ok, we can use our TLB miss and window trap handlers safely. */
setup_tba();
+ DBGRK(); <---------- line 1413 from http://tuxland.pl/misc/sparc64/sparc64-2.6.24-rc7.jpg
__flush_tlb_all(); <--------------------------- this function fails
+ DBGRK();
if (tlb_type == hypervisor)
sun4v_ktsb_register();
+ DBGRK();
/* Setup bootmem... */
pages_avail = 0;
last_valid_pfn = end_pfn = bootmem_init(&pages_avail, phys_base);
+ DBGRK();
max_mapnr = last_valid_pfn;
+ DBGRK();
kernel_physical_mapping_init();
+ DBGRK();
real_setup_per_cpu_areas();
+ DBGRK();
prom_build_devicetree();
+ DBGRK();
if (tlb_type == hypervisor)
sun4v_mdesc_init();
@@ -1908,33 +1917,44 @@ void __flush_tlb_all(void)
unsigned long pstate;
int i;
+ DBGRK(); <------------------------ the next DBGRK() to be called afrer line 1413
__asm__ __volatile__("flushw\n\t"
"rdpr %%pstate, %0\n\t"
"wrpr %0, %1, %%pstate"
: "=r" (pstate)
: "i" (PSTATE_IE));
(Please note that line numbers differ from pure 2.6.24-rc7 as I put this debug
stuff in there.)
Now please take a look at http://tuxland.pl/misc/sparc64/sparc64-2.6.24-rc7.jpg
and you'll see that __flush_tlb_all() fails. I assume that at least first DBGRK()
in __flush_tlb_all() gets printed but then gets overwritten with the garbage
you see in screenshot).
2.6.16 produces a bit more info which you can find here:
http://tuxland.pl/misc/sparc64/sparc64-2.6.16.jpg
The modified arch/sparc64/mm/init.c from 2.6.24-rc7 (to match it with the screen attached earlier)
is available here: http://tuxland.pl/misc/sparc64/init.c
2.6.24-rc7 config: http://tuxland.pl/misc/sparc64/sparc64-64kB-config-2.6.24-rc7
Kernels 2.6.20 and newer produce the same output as 2.6.24-rc7.
So ... am I doing something wrong here or something is broken?
Regards,
Mariusz
The box is sun ultra 60 with two cpus and 1G of ram.
$ cat /proc/cpuinfo
cpu : TI UltraSparc II (BlackBird)
fpu : UltraSparc II integrated FPU
prom : OBP 3.17.0 1998/10/23 11:26
type : sun4u
ncpus probed : 2
ncpus active : 2
D$ parity tl1 : 0
I$ parity tl1 : 0
Cpu0ClkTck : 000000001ad1cbeb
Cpu2ClkTck : 000000001ad1cbeb
MMU Type : Spitfire
State:
CPU0: online
CPU2: online
$ cat /proc/zoneinfo
Node 0, zone Normal
pages free 109267
min 501
low 626
high 751
scanned 0 (a: 0 i: 0)
spanned 393114
present 125792
nr_free_pages 109267
nr_inactive 5760
nr_active 4147
nr_anon_pages 995
nr_mapped 538
nr_file_pages 8914
nr_dirty 0
nr_writeback 0
nr_slab_reclaimable 5664
nr_slab_unreclaimable 871
nr_page_table_pages 47
nr_unstable 0
nr_bounce 0
nr_vmscan_write 0
protection: (0, 0)
pagesets
cpu: 0 pcp: 0
count: 88
high: 90
batch: 15
cpu: 0 pcp: 1
count: 5
high: 30
batch: 7
vm stats threshold: 16
cpu: 2 pcp: 0
count: 31
high: 90
batch: 15
cpu: 2 pcp: 1
count: 5
high: 30
batch: 7
vm stats threshold: 16
all_unreclaimable: 0
prev_priority: 12
start_pfn: 0
$ cat /proc/buddyinfo
Node 0, zone Normal 25 5 0 16 5 2 4 1 0 0 106
$ cat /proc/interrupts
CPU0 CPU2
0: 32290 29965 <NULL> timer
1: 0 0 sun4u PSYCHO_PCIERR
2: 0 0 sun4u PSYCHO_UE
3: 0 0 sun4u PSYCHO_CE
8: 5 0 sun4u su(kbd)
9: 0 0 sun4u su(mouse)
14: 2302 0 sun4u eth0
15: 6379 0 sun4u sym53c8xx
16: 0 30 sun4u sym53c8xx
17: 0 0 sun4u PSYCHO_PCIERR
# lspci
0000:00:00.0 Host bridge: Sun Microsystems Computer Corp. Psycho PCI Bus Module
0000:00:01.0 Bridge: Sun Microsystems Computer Corp. EBUS (rev 01)
0000:00:01.1 Ethernet controller: Sun Microsystems Computer Corp. Happy Meal (rev 01)
0000:00:03.0 SCSI storage controller: LSI Logic / Symbios Logic 53c875 (rev 14)
0000:00:03.1 SCSI storage controller: LSI Logic / Symbios Logic 53c875 (rev 14)
0001:00:00.0 Host bridge: Sun Microsystems Computer Corp. Psycho PCI Bus Module
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/