Skip to contentSkip to conversation

2.6.17: scan DMI early

PATCH In review Jan Beulich 2 participants 3 messages Feb 14, 2007 – Mar 5, 2007

lore
Share & tools
lore
#0 lore raw

PATCH 8/8 2.6.17: scan DMI early

While shuffling quite a few things around, this gets us closer to native, which clearly had a reason to do the DMI scan early.

arch/i386/mm/ioremap-xen.c +0 -4
Index: head-2007-02-08/arch/i386/mm/ioremap-xen.c
===================================================================
@@ -374,8 +374,6 @@ void iounmap(volatile void __iomem *addr
 }
 EXPORT_SYMBOL(iounmap);
 
-#ifdef __i386__
-
 void __init *bt_ioremap(unsigned long phys_addr, unsigned long size)
 {
 	unsigned long offset, last_addr;
@@ -443,5 +441,3 @@ void __init bt_iounmap(void *addr, unsig
 		--nrpages;
 	}
 }
-
-#endif /* __i386__ */
arch/x86_64/kernel/setup-xen.c +2 -8
Index: head-2007-02-08/arch/x86_64/kernel/setup-xen.c
===================================================================
@@ -681,7 +681,8 @@ void __init setup_arch(char **cmdline_p)
 
 	init_memory_mapping(0, (end_pfn_map << PAGE_SHIFT));
 
-	/* dmi_scan_machine(); */
+	if (is_initial_xendomain())
+		dmi_scan_machine();
 
 #ifdef CONFIG_ACPI_NUMA
 	/*
@@ -1630,13 +1629,6 @@ struct seq_operations cpuinfo_op = {
 	.show =	show_cpuinfo,
 };
 
-static int __init run_dmi_scan(void)
-{
-	dmi_scan_machine();
-	return 0;
-}
-core_initcall(run_dmi_scan);
-
 #if defined(CONFIG_INPUT_PCSPKR) || defined(CONFIG_INPUT_PCSPKR_MODULE)
 #include <linux/platform_device.h>
 static __init int add_pcspkr(void)
arch/x86_64/mm/init-xen.c +59 -24

160 lines not on this page · raw message

include/asm-x86_64/mach-xen/asm/io.h +4 -2
Index: head-2007-02-08/include/asm-x86_64/mach-xen/asm/io.h
===================================================================
@@ -150,8 +150,10 @@ static inline void __iomem * ioremap (un
 	return __ioremap(offset, size, 0);
 }
 
-extern void *early_ioremap(unsigned long addr, unsigned long size);
-extern void early_iounmap(void *addr, unsigned long size);
+extern void *bt_ioremap(unsigned long addr, unsigned long size);
+extern void bt_iounmap(void *addr, unsigned long size);
+#define early_ioremap bt_ioremap
+#define early_iounmap bt_iounmap
 
 /*
  * This one maps high address device memory and turns off caching for that area.
include/asm-x86_64/mach-xen/asm/fixmap.h +5 -0
Index: head-2007-02-08/include/asm-x86_64/mach-xen/asm/fixmap.h
===================================================================
@@ -53,6 +53,11 @@ enum fixed_addresses {
 #define NR_FIX_ISAMAPS	256
 	FIX_ISAMAP_END,
 	FIX_ISAMAP_BEGIN = FIX_ISAMAP_END + NR_FIX_ISAMAPS - 1,
+	__end_of_permanent_fixed_addresses,
+	/* temporary boot-time mappings, used before ioremap() is functional */
+#define NR_FIX_BTMAPS	16
+	FIX_BTMAP_END = __end_of_permanent_fixed_addresses,
+	FIX_BTMAP_BEGIN = FIX_BTMAP_END + NR_FIX_BTMAPS - 1,
 	__end_of_fixed_addresses
 };
#1 lore raw
Quoted 4 lines · Jan Beulich

While shuffling quite a few things around, this gets us closer to native, which clearly had a reason to do the DMI scan early.

Taken, but a few comments:

- could we get rid of alloc_static_page() and use spp_getpage() everywhere?

- the double-staged find_early_table_space(), where table_end gets updated
halfway through init_memory_mapping(), is pretty skanky. I suppose we get to keep the BUG_ON(start_pfn != table_end) this way, but perhaps it would be nicer just to set table_end once and for all in find_early_table_space().

- the estimate of number of fixmap pagetables required in
find_early_table_space() is conservative. Is there any risk that spp_getpage() may end up with start_pfn<table_end for long enough that we run into concurrency issues? Perhaps we could do a dummy set_fixmap for every fixmap slot to force population of all fixmap slots in init_memory_mapping(), and then BUG_ON(start_pfn != table_end)?

-- Keir

#2 lore raw
Quoted 6 lines · Keir Fraser
Quoted 4 lines

While shuffling quite a few things around, this gets us closer to native, which clearly had a reason to do the DMI scan early.

Taken, but a few comments:

I had a go at cleaning this up myself and it isn't as straightforward as I first thought (can't get rid of alloc_static_page() for example!). But I did simplify the fixmap pagetable allocation and initialisation a bit, at least.

-- Keir