Re: [PATCH v3] x86: efi: Turn off efi_enabled after setup on mixed fw/kernel

[Matt Fleming <matt-HNK1S37rvNbeXh+fF434Mdi2O/JbrIOy@public.gmane.org>]

On Wed, 2012-10-24 at 10:00 -0700, Olof Johansson wrote:
> When 32-bit EFI is used with 64-bit kernel (or vice versa), turn off
> efi_enabled once setup is done. Beyond setup, it is normally used to
> determine if runtime services are available and we will have none.
> 
> This will resolve issues stemming from efivars modprobe panicking on a
> 32/64-bit setup, as well as some reboot issues on similar setups.
> 
> Bugzilla: https://bugzilla.kernel.org/show_bug.cgi?id=45991
> 
> Reported-by: Marko Kohtala <marko.kohtala-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> Reported-by: Maxim Kammerer <mk-HZxBN897WPs@public.gmane.org>
> Signed-off-by: Olof Johansson <olof-nZhT3qVonbNeoWH0uzbU5w@public.gmane.org>
> Acked-by: Maarten Lankhorst <maarten.lankhorst-Z7WLFzj8eWMS+FvcfC7Uqw@public.gmane.org>
> Cc: stable-DgEjT+Ai2ygdnm+yROfE0A@public.gmane.org # 3.4 - 3.6
> Cc: Matthew Garrett <mjg-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
> ---

Running with this patch results in the following,

[    3.486943] ------------[ cut here ]------------
[    3.488455] WARNING: at /home/matt/src/kernels/linux-2.6/arch/x86/mm/ioremap.c:586 check_early_ioremap_leak+0x3b/0x50()
[    3.490043] Hardware name: MacBook2,1
[    3.491600] Debug warning: early ioremap leak of 1 areas detected.
[    3.493156] Modules linked in:
[    3.493973] usb 1-4: new high-speed USB device number 3 using ehci_hcd
[    3.496340] Pid: 1, comm: swapper/0 Not tainted 3.7.0-rc1+ #26
[    3.497949] Call Trace:
[    3.499558]  [<ffffffff81d022ab>] ? check_early_ioremap_leak+0x3b/0x50
[    3.501217]  [<ffffffff81032970>] warn_slowpath_common+0x85/0x9d
[    3.502872]  [<ffffffff81d02270>] ? early_ioremap_debug_setup+0x12/0x12
[    3.504516]  [<ffffffff81032a2b>] warn_slowpath_fmt+0x46/0x48
[    3.506161]  [<ffffffff81cf89df>] ? mcheck_debugfs_init+0x3b/0x3b
[    3.507790]  [<ffffffff81d022ab>] check_early_ioremap_leak+0x3b/0x50
[    3.509418]  [<ffffffff8100023b>] do_one_initcall+0x7f/0x134
[    3.511040]  [<ffffffff8163cc3c>] kernel_init+0x10c/0x275
[    3.512638]  [<ffffffff81cef4c7>] ? loglevel+0x31/0x31
[    3.514241]  [<ffffffff8163cb30>] ? rest_init+0x74/0x74
[    3.515824]  [<ffffffff8165ca9c>] ret_from_fork+0x7c/0xb0
[    3.517393]  [<ffffffff8163cb30>] ? rest_init+0x74/0x74
[    3.518958] ---[ end trace afc7d1c216dac6e1 ]---
[    3.520479] please boot with early_ioremap_debug and report the dmesg.

because we're now leaking the mapping from efi_memmap_init(). 

Maybe it should be handled like so?

diff --git a/arch/x86/include/asm/efi.h b/arch/x86/include/asm/efi.h
index c9dcc18..029189d 100644
--- a/arch/x86/include/asm/efi.h
+++ b/arch/x86/include/asm/efi.h
@@ -98,6 +98,7 @@ extern void efi_set_executable(efi_memory_desc_t *md, bool executable);
 extern int efi_memblock_x86_reserve_range(void);
 extern void efi_call_phys_prelog(void);
 extern void efi_call_phys_epilog(void);
+extern void efi_unmap_memmap(void);
 
 #ifndef CONFIG_EFI
 /*
diff --git a/arch/x86/kernel/setup.c b/arch/x86/kernel/setup.c
index 392dae3..6fd7752 100644
--- a/arch/x86/kernel/setup.c
+++ b/arch/x86/kernel/setup.c
@@ -1043,6 +1043,7 @@ void __init setup_arch(char **cmdline_p)
 	 */
 	if (efi_enabled && IS_ENABLED(CONFIG_X86_64) != efi_64bit) {
 		pr_info("efi: Setup done, disabling due to 32/64-bit mismatch\n");
+		efi_unmap_memmap();
 		efi_enabled = 0;
 	}
 #endif
diff --git a/arch/x86/platform/efi/efi.c b/arch/x86/platform/efi/efi.c
index 6e620f1..9bae3b7 100644
--- a/arch/x86/platform/efi/efi.c
+++ b/arch/x86/platform/efi/efi.c
@@ -424,7 +424,7 @@ void __init efi_reserve_boot_services(void)
 	}
 }
 
-static void __init efi_unmap_memmap(void)
+void __init efi_unmap_memmap(void)
 {
 	if (memmap.map) {
 		early_iounmap(memmap.map, memmap.nr_map * memmap.desc_size);


-- 
Matt Fleming, Intel Open Source Technology Center