public inbox for linux-kernel@vger.kernel.org
 help / color / mirror / Atom feed
From: Tom Rini <trini@kernel.crashing.org>
To: Andi Kleen <ak@suse.de>
Cc: akpm@osdl.org, linux-kernel@vger.kernel.org, bob.picco@hp.com
Subject: Re: [patch 2/3] x86_64: Run setup_per_cpu_areas and trap_init sooner
Date: Mon, 29 Aug 2005 13:03:22 -0700	[thread overview]
Message-ID: <20050829200322.GG3827@smtp.west.cox.net> (raw)
In-Reply-To: <200508292155.01252.ak@suse.de>

On Mon, Aug 29, 2005 at 09:55:00PM +0200, Andi Kleen wrote:
> 
> > +void __init early_setup_per_cpu_areas(void)
> > +{
> > +	static char cpu0[PERCPU_ENOUGH_ROOM]
> > +		__attribute__ ((aligned (SMP_CACHE_BYTES)));
> 
> This needs a __cacheline_aligned too, otherwise there could be false sharing.

Done, thanks:

CC: Andi Kleen <ak@suse.de>, Bob Picco <bob.picco@hp.com>
It can be handy in some situations to have run trap_init() sooner than the
generic code does.  In order to do this on x86_64 we need to add a custom
early_setup_per_cpu_areas() call as well.

---

 linux-2.6.13-trini/arch/x86_64/kernel/setup.c   |    3 +++
 linux-2.6.13-trini/arch/x86_64/kernel/setup64.c |   16 +++++++++++++++-
 linux-2.6.13-trini/arch/x86_64/kernel/traps.c   |    4 ++++
 linux-2.6.13-trini/include/asm-x86_64/proto.h   |    2 ++
 4 files changed, 24 insertions(+), 1 deletion(-)

diff -puN arch/x86_64/kernel/setup.c~x86_64-early_funcs arch/x86_64/kernel/setup.c
--- linux-2.6.13/arch/x86_64/kernel/setup.c~x86_64-early_funcs	2005-08-29 12:39:49.000000000 -0700
+++ linux-2.6.13-trini/arch/x86_64/kernel/setup.c	2005-08-29 12:39:49.000000000 -0700
@@ -525,6 +525,9 @@ void __init setup_arch(char **cmdline_p)
 {
 	unsigned long kernel_end;
 
+	early_setup_per_cpu_areas();
+	early_trap_init();
+
  	ROOT_DEV = old_decode_dev(ORIG_ROOT_DEV);
  	drive_info = DRIVE_INFO;
  	screen_info = SCREEN_INFO;
diff -puN arch/x86_64/kernel/setup64.c~x86_64-early_funcs arch/x86_64/kernel/setup64.c
--- linux-2.6.13/arch/x86_64/kernel/setup64.c~x86_64-early_funcs	2005-08-29 12:39:49.000000000 -0700
+++ linux-2.6.13-trini/arch/x86_64/kernel/setup64.c	2005-08-29 13:01:52.000000000 -0700
@@ -77,7 +77,19 @@ static int __init nonx32_setup(char *str
 }
 __setup("noexec32=", nonx32_setup);
 
+void __init early_setup_per_cpu_areas(void)
+{
+	static char cpu0[PERCPU_ENOUGH_ROOM] __cacheline_aligned
+		__attribute__ ((aligned (SMP_CACHE_BYTES)));
+	char *ptr = cpu0;
+
+	cpu_pda[0].data_offset = ptr - __per_cpu_start;
+	memcpy(ptr, __per_cpu_start, __per_cpu_end - __per_cpu_start);
+}
+
 /*
+ * We run this a bit sooner than the normal code, so provide a dummy
+ * function as well.
  * Great future plan:
  * Declare PDA itself and support (irqstack,tss,pgd) as per cpu data.
  * Always point %gs to its beginning
@@ -97,7 +109,9 @@ void __init setup_per_cpu_areas(void)
 	for (i = 0; i < NR_CPUS; i++) { 
 		char *ptr;
 
-		if (!NODE_DATA(cpu_to_node(i))) {
+		if (cpu_pda[i].data_offset)
+			continue;
+		else if (!NODE_DATA(cpu_to_node(i))) {
 			printk("cpu with no node %d, num_online_nodes %d\n",
 			       i, num_online_nodes());
 			ptr = alloc_bootmem(size);
diff -puN arch/x86_64/kernel/traps.c~x86_64-early_funcs arch/x86_64/kernel/traps.c
--- linux-2.6.13/arch/x86_64/kernel/traps.c~x86_64-early_funcs	2005-08-29 12:39:49.000000000 -0700
+++ linux-2.6.13-trini/arch/x86_64/kernel/traps.c	2005-08-29 12:39:49.000000000 -0700
@@ -904,6 +904,10 @@ void do_call_debug(struct pt_regs *regs)
 
 void __init trap_init(void)
 {
+}
+
+void __init early_trap_init(void)
+{
 	set_intr_gate(0,&divide_error);
 	set_intr_gate_ist(1,&debug,DEBUG_STACK);
 	set_intr_gate_ist(2,&nmi,NMI_STACK);
diff -puN include/asm-x86_64/proto.h~x86_64-early_funcs include/asm-x86_64/proto.h
--- linux-2.6.13/include/asm-x86_64/proto.h~x86_64-early_funcs	2005-08-29 12:39:49.000000000 -0700
+++ linux-2.6.13-trini/include/asm-x86_64/proto.h	2005-08-29 12:39:49.000000000 -0700
@@ -10,6 +10,8 @@ struct pt_regs;
 
 extern void get_cpu_vendor(struct cpuinfo_x86*);
 extern void start_kernel(void);
+extern void early_trap_init(void);
+extern void early_setup_per_cpu_areas(void);
 extern void pda_init(int); 
 
 extern void early_idt_handler(void);

-- 
Tom Rini
http://gate.crashing.org/~trini/

  reply	other threads:[~2005-08-29 20:03 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
     [not found] <1.2982005.trini@kernel.crashing.org>
2005-08-29 16:08 ` [patch 01/16] Add a KGDB core Tom Rini
2005-08-29 16:09   ` [patch 02/16] Add support for i386 platforms to KGDB Tom Rini
2005-08-29 16:09     ` [patch 03/16] Add support for PowerPC32 " Tom Rini
2005-08-29 16:09       ` [patch 04/16] I/O driver for 8250-compatible UARTs Tom Rini
2005-08-29 16:09         ` [patch 05/16] Add support for MIPS platforms to KGDB Tom Rini
2005-08-29 16:10           ` [patch 06/16] Add support for IA64 " Tom Rini
2005-08-29 16:10             ` [patch 07/16] x86_64: Rename KDB_VECTOR to DEBUGGER_VECTOR Tom Rini
2005-08-29 16:10               ` [patch 08/16] Add support for X86_64 platforms to KGDB Tom Rini
2005-08-29 16:10                 ` [patch 09/16] Add support for SuperH " Tom Rini
2005-08-29 16:10                   ` [patch 10/16] Add support for ARM " Tom Rini
2005-08-29 16:11                     ` [patch 11/16] Add support for PowerPC64 " Tom Rini
2005-08-29 16:11                       ` [patch 12/16] KGDBoE I/O driver Tom Rini
2005-08-29 16:11                         ` [patch 13/16] Add CFI DWARF2 annotation support Tom Rini
2005-08-29 16:11                           ` [patch 14/16] Minor SysRq keyboard bugfix for KGDB Tom Rini
2005-08-29 16:11                             ` [patch 15/16] Allow KGDB to work well with loaded modules Tom Rini
2005-08-29 16:12                               ` [patch 16/16] Add hardware breakpoint support for i386 Tom Rini
2005-08-29 21:23                                 ` Andi Kleen
2005-08-31 14:39                                   ` Tom Rini
2005-08-30  1:06                                 ` Keith Owens
2005-08-29 17:13                 ` [patch 08/16] Add support for X86_64 platforms to KGDB Andi Kleen
2005-08-29 17:45                   ` Tom Rini
2005-08-29 18:46                     ` Andi Kleen
2005-08-29 18:49                       ` Tom Rini
2005-08-29 16:18         ` [patch 04/16] I/O driver for 8250-compatible UARTs Russell King
2005-08-29 16:28           ` Tom Rini
2005-08-31 19:38         ` Bjorn Helgaas
2005-08-31 20:10           ` Tom Rini
2005-08-31 21:03             ` Russell King
2005-08-31 21:23               ` Tom Rini
2005-08-31 21:19             ` Bjorn Helgaas
2005-08-31 22:15               ` Tom Rini
2005-08-29 19:55     ` [patch 2/3] x86_64: Run setup_per_cpu_areas and trap_init sooner Andi Kleen
2005-08-29 20:03       ` Tom Rini [this message]
2005-08-30  7:33   ` [patch 1/3] x86_64: Add a notify_die() call to the "no context" part of do_page_fault() George Anzinger
2005-08-30 14:06     ` Tom Rini
2005-08-30 14:50       ` George Anzinger
2005-08-30 19:53         ` Tom Rini

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20050829200322.GG3827@smtp.west.cox.net \
    --to=trini@kernel.crashing.org \
    --cc=ak@suse.de \
    --cc=akpm@osdl.org \
    --cc=bob.picco@hp.com \
    --cc=linux-kernel@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox