[PATCH 0/2] arm64: don't call early_*map() post paging_init()

[PATCH 0/2] arm64: don't call early_*map() post paging_init()

[Leif Lindholm]

This (tiny) series resolves a fairly serious problem with
early_ioremap/iounmap/memremap/memunmap on arm64. These functions
cannot safely be called after paging_init(), but the sanity check
was not triggering.

As a result, a fixmap entry was incorrectly cleared during
early_initcalls on arm64 UEFI systems.

1/2 reworks the arm64 UEFI support code to not attempt these calls
and
2/2 enables the sanity check

Leif Lindholm (2):
  arm64: don't make early_*map() calls post paging_init()
  arm64: call early_ioremap_reset() in paging_init()

 arch/arm64/kernel/efi.c   | 12 +++---------
 arch/arm64/kernel/setup.c |  2 --
 arch/arm64/mm/mmu.c       |  1 +
 3 files changed, 4 insertions(+), 11 deletions(-)

-- 
2.1.3

[PATCH 0/2] arm64: don't call early_*map() post paging_init()

[Leif Lindholm]

This (tiny) series resolves a fairly serious problem with
early_ioremap/iounmap/memremap/memunmap on arm64. These functions
cannot safely be called after paging_init(), but the sanity check
was not triggering.

As a result, a fixmap entry was incorrectly cleared during
early_initcalls on arm64 UEFI systems.

1/2 reworks the arm64 UEFI support code to not attempt these calls
and
2/2 enables the sanity check

Leif Lindholm (2):
  arm64: don't make early_*map() calls post paging_init()
  arm64: call early_ioremap_reset() in paging_init()

 arch/arm64/kernel/efi.c   | 12 +++---------
 arch/arm64/kernel/setup.c |  2 --
 arch/arm64/mm/mmu.c       |  1 +
 3 files changed, 4 insertions(+), 11 deletions(-)

-- 
2.1.3

[PATCH 1/2] arm64: don't make early_*map() calls post paging_init()

[Leif Lindholm]

arm64 early_ioremap/iounmap/memremap/memunmap are not supported beyond
the call to paging_init(), but arm64_enter_virtual_mode() (an early
initcall) makes one call to unmap the UEFI memory map.

Rearrange the code to unmap this region before paging_init().

Signed-off-by: Leif Lindholm <leif.lindholm@linaro.org>
---
 arch/arm64/kernel/efi.c   | 12 +++---------
 arch/arm64/kernel/setup.c |  2 --
 2 files changed, 3 insertions(+), 11 deletions(-)

diff --git a/arch/arm64/kernel/efi.c b/arch/arm64/kernel/efi.c
index 6fac253..790adb5 100644
--- a/arch/arm64/kernel/efi.c
+++ b/arch/arm64/kernel/efi.c
@@ -317,15 +317,8 @@ void __init efi_init(void)
 		return;
 
 	reserve_regions();
-}
-
-void __init efi_idmap_init(void)
-{
-	if (!efi_enabled(EFI_BOOT))
-		return;
 
-	/* boot time idmap_pg_dir is incomplete, so fill in missing parts */
-	efi_setup_idmap();
+	early_memunmap(memmap.map, params.mmap_size);
 }
 
 static int __init remap_region(efi_memory_desc_t *md, void **new)
@@ -380,7 +373,6 @@ static int __init arm64_enter_virtual_mode(void)
 	}
 
 	mapsize = memmap.map_end - memmap.map;
-	early_memunmap(memmap.map, mapsize);
 
 	if (efi_runtime_disabled()) {
 		pr_info("EFI runtime services will be disabled.\n");
@@ -393,6 +385,8 @@ static int __init arm64_enter_virtual_mode(void)
 						   mapsize);
 	memmap.map_end = memmap.map + mapsize;
 
+	efi_setup_idmap();
+
 	efi.memmap = &memmap;
 
 	/* Map the runtime regions */
diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
index b809911..c7eb3c5 100644
--- a/arch/arm64/kernel/setup.c
+++ b/arch/arm64/kernel/setup.c
@@ -401,8 +401,6 @@ void __init setup_arch(char **cmdline_p)
 	paging_init();
 	request_standard_resources();
 
-	efi_idmap_init();
-
 	unflatten_device_tree();
 
 	psci_init();
-- 
2.1.3

[PATCH 1/2] arm64: don't make early_*map() calls post paging_init()

[Leif Lindholm]

arm64 early_ioremap/iounmap/memremap/memunmap are not supported beyond
the call to paging_init(), but arm64_enter_virtual_mode() (an early
initcall) makes one call to unmap the UEFI memory map.

Rearrange the code to unmap this region before paging_init().

Signed-off-by: Leif Lindholm <leif.lindholm@linaro.org>
---
 arch/arm64/kernel/efi.c   | 12 +++---------
 arch/arm64/kernel/setup.c |  2 --
 2 files changed, 3 insertions(+), 11 deletions(-)

diff --git a/arch/arm64/kernel/efi.c b/arch/arm64/kernel/efi.c
index 6fac253..790adb5 100644
--- a/arch/arm64/kernel/efi.c
+++ b/arch/arm64/kernel/efi.c
@@ -317,15 +317,8 @@ void __init efi_init(void)
 		return;
 
 	reserve_regions();
-}
-
-void __init efi_idmap_init(void)
-{
-	if (!efi_enabled(EFI_BOOT))
-		return;
 
-	/* boot time idmap_pg_dir is incomplete, so fill in missing parts */
-	efi_setup_idmap();
+	early_memunmap(memmap.map, params.mmap_size);
 }
 
 static int __init remap_region(efi_memory_desc_t *md, void **new)
@@ -380,7 +373,6 @@ static int __init arm64_enter_virtual_mode(void)
 	}
 
 	mapsize = memmap.map_end - memmap.map;
-	early_memunmap(memmap.map, mapsize);
 
 	if (efi_runtime_disabled()) {
 		pr_info("EFI runtime services will be disabled.\n");
@@ -393,6 +385,8 @@ static int __init arm64_enter_virtual_mode(void)
 						   mapsize);
 	memmap.map_end = memmap.map + mapsize;
 
+	efi_setup_idmap();
+
 	efi.memmap = &memmap;
 
 	/* Map the runtime regions */
diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
index b809911..c7eb3c5 100644
--- a/arch/arm64/kernel/setup.c
+++ b/arch/arm64/kernel/setup.c
@@ -401,8 +401,6 @@ void __init setup_arch(char **cmdline_p)
 	paging_init();
 	request_standard_resources();
 
-	efi_idmap_init();
-
 	unflatten_device_tree();
 
 	psci_init();
-- 
2.1.3

[PATCH 2/2] arm64: call early_ioremap_reset() in paging_init()

[Leif Lindholm]

arm64 does not support early_memremap/memunmap/ioremap/iounmap after
paging_init() has been called. The core early_*remap code handles this
via the after_paging_init variable, which is set by a call to
early_ioremap_reset().

However, arm64 currently does not call early_ioremap_reset(), which
has made it possible to poke around in the fixmap region after kmap
is enabled. Add the required call.

Signed-off-by: Leif Lindholm <leif.lindholm-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
---
 arch/arm64/mm/mmu.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index 6032f3e..506544f 100644
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c
@@ -348,6 +348,7 @@ void __init paging_init(void)
 {
 	void *zero_page;
 
+	early_ioremap_reset();
 	map_mem();
 
 	/*
-- 
2.1.3

[PATCH 2/2] arm64: call early_ioremap_reset() in paging_init()

[Leif Lindholm]

arm64 does not support early_memremap/memunmap/ioremap/iounmap after
paging_init() has been called. The core early_*remap code handles this
via the after_paging_init variable, which is set by a call to
early_ioremap_reset().

However, arm64 currently does not call early_ioremap_reset(), which
has made it possible to poke around in the fixmap region after kmap
is enabled. Add the required call.

Signed-off-by: Leif Lindholm <leif.lindholm@linaro.org>
---
 arch/arm64/mm/mmu.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
index 6032f3e..506544f 100644
--- a/arch/arm64/mm/mmu.c
+++ b/arch/arm64/mm/mmu.c
@@ -348,6 +348,7 @@ void __init paging_init(void)
 {
 	void *zero_page;
 
+	early_ioremap_reset();
 	map_mem();
 
 	/*
-- 
2.1.3

Re: [PATCH 1/2] arm64: don't make early_*map() calls post paging_init()

[Ard Biesheuvel]

On 5 January 2015 at 18:33, Leif Lindholm <leif.lindholm@linaro.org> wrote:
> arm64 early_ioremap/iounmap/memremap/memunmap are not supported beyond
> the call to paging_init(), but arm64_enter_virtual_mode() (an early
> initcall) makes one call to unmap the UEFI memory map.
>
> Rearrange the code to unmap this region before paging_init().
>
> Signed-off-by: Leif Lindholm <leif.lindholm@linaro.org>
> ---
>  arch/arm64/kernel/efi.c   | 12 +++---------
>  arch/arm64/kernel/setup.c |  2 --
>  2 files changed, 3 insertions(+), 11 deletions(-)
>
> diff --git a/arch/arm64/kernel/efi.c b/arch/arm64/kernel/efi.c
> index 6fac253..790adb5 100644
> --- a/arch/arm64/kernel/efi.c
> +++ b/arch/arm64/kernel/efi.c
> @@ -317,15 +317,8 @@ void __init efi_init(void)
>                 return;
>
>         reserve_regions();
> -}
> -
> -void __init efi_idmap_init(void)
> -{
> -       if (!efi_enabled(EFI_BOOT))
> -               return;
>
> -       /* boot time idmap_pg_dir is incomplete, so fill in missing parts */
> -       efi_setup_idmap();
> +       early_memunmap(memmap.map, params.mmap_size);
>  }
>
>  static int __init remap_region(efi_memory_desc_t *md, void **new)
> @@ -380,7 +373,6 @@ static int __init arm64_enter_virtual_mode(void)
>         }
>
>         mapsize = memmap.map_end - memmap.map;
> -       early_memunmap(memmap.map, mapsize);
>
>         if (efi_runtime_disabled()) {
>                 pr_info("EFI runtime services will be disabled.\n");
> @@ -393,6 +385,8 @@ static int __init arm64_enter_virtual_mode(void)
>                                                    mapsize);
>         memmap.map_end = memmap.map + mapsize;
>
> +       efi_setup_idmap();
> +

I think this is incorrect: we may call early_alloc() here which is
only allowed before paging_init()

>         efi.memmap = &memmap;
>
>         /* Map the runtime regions */
> diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
> index b809911..c7eb3c5 100644
> --- a/arch/arm64/kernel/setup.c
> +++ b/arch/arm64/kernel/setup.c
> @@ -401,8 +401,6 @@ void __init setup_arch(char **cmdline_p)
>         paging_init();
>         request_standard_resources();
>
> -       efi_idmap_init();
> -
>         unflatten_device_tree();
>
>         psci_init();
> --
> 2.1.3
>

[PATCH 1/2] arm64: don't make early_*map() calls post paging_init()

[Ard Biesheuvel]

On 5 January 2015 at 18:33, Leif Lindholm <leif.lindholm@linaro.org> wrote:
> arm64 early_ioremap/iounmap/memremap/memunmap are not supported beyond
> the call to paging_init(), but arm64_enter_virtual_mode() (an early
> initcall) makes one call to unmap the UEFI memory map.
>
> Rearrange the code to unmap this region before paging_init().
>
> Signed-off-by: Leif Lindholm <leif.lindholm@linaro.org>
> ---
>  arch/arm64/kernel/efi.c   | 12 +++---------
>  arch/arm64/kernel/setup.c |  2 --
>  2 files changed, 3 insertions(+), 11 deletions(-)
>
> diff --git a/arch/arm64/kernel/efi.c b/arch/arm64/kernel/efi.c
> index 6fac253..790adb5 100644
> --- a/arch/arm64/kernel/efi.c
> +++ b/arch/arm64/kernel/efi.c
> @@ -317,15 +317,8 @@ void __init efi_init(void)
>                 return;
>
>         reserve_regions();
> -}
> -
> -void __init efi_idmap_init(void)
> -{
> -       if (!efi_enabled(EFI_BOOT))
> -               return;
>
> -       /* boot time idmap_pg_dir is incomplete, so fill in missing parts */
> -       efi_setup_idmap();
> +       early_memunmap(memmap.map, params.mmap_size);
>  }
>
>  static int __init remap_region(efi_memory_desc_t *md, void **new)
> @@ -380,7 +373,6 @@ static int __init arm64_enter_virtual_mode(void)
>         }
>
>         mapsize = memmap.map_end - memmap.map;
> -       early_memunmap(memmap.map, mapsize);
>
>         if (efi_runtime_disabled()) {
>                 pr_info("EFI runtime services will be disabled.\n");
> @@ -393,6 +385,8 @@ static int __init arm64_enter_virtual_mode(void)
>                                                    mapsize);
>         memmap.map_end = memmap.map + mapsize;
>
> +       efi_setup_idmap();
> +

I think this is incorrect: we may call early_alloc() here which is
only allowed before paging_init()

>         efi.memmap = &memmap;
>
>         /* Map the runtime regions */
> diff --git a/arch/arm64/kernel/setup.c b/arch/arm64/kernel/setup.c
> index b809911..c7eb3c5 100644
> --- a/arch/arm64/kernel/setup.c
> +++ b/arch/arm64/kernel/setup.c
> @@ -401,8 +401,6 @@ void __init setup_arch(char **cmdline_p)
>         paging_init();
>         request_standard_resources();
>
> -       efi_idmap_init();
> -
>         unflatten_device_tree();
>
>         psci_init();
> --
> 2.1.3
>

Re: [PATCH 2/2] arm64: call early_ioremap_reset() in paging_init()

[Ard Biesheuvel]

On 5 January 2015 at 18:33, Leif Lindholm <leif.lindholm-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org> wrote:
> arm64 does not support early_memremap/memunmap/ioremap/iounmap after
> paging_init() has been called. The core early_*remap code handles this
> via the after_paging_init variable, which is set by a call to
> early_ioremap_reset().
>
> However, arm64 currently does not call early_ioremap_reset(), which
> has made it possible to poke around in the fixmap region after kmap
> is enabled. Add the required call.
>
> Signed-off-by: Leif Lindholm <leif.lindholm-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>
Acked-by: Ard Biesheuvel <ard.biesheuvel-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>

Perhaps add 'Fixes' and 'Cc: stable' tags as well?


> ---
>  arch/arm64/mm/mmu.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
> index 6032f3e..506544f 100644
> --- a/arch/arm64/mm/mmu.c
> +++ b/arch/arm64/mm/mmu.c
> @@ -348,6 +348,7 @@ void __init paging_init(void)
>  {
>         void *zero_page;
>
> +       early_ioremap_reset();
>         map_mem();
>
>         /*
> --
> 2.1.3
>

[PATCH 2/2] arm64: call early_ioremap_reset() in paging_init()

[Ard Biesheuvel]

On 5 January 2015 at 18:33, Leif Lindholm <leif.lindholm@linaro.org> wrote:
> arm64 does not support early_memremap/memunmap/ioremap/iounmap after
> paging_init() has been called. The core early_*remap code handles this
> via the after_paging_init variable, which is set by a call to
> early_ioremap_reset().
>
> However, arm64 currently does not call early_ioremap_reset(), which
> has made it possible to poke around in the fixmap region after kmap
> is enabled. Add the required call.
>
> Signed-off-by: Leif Lindholm <leif.lindholm@linaro.org>
Acked-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>

Perhaps add 'Fixes' and 'Cc: stable' tags as well?


> ---
>  arch/arm64/mm/mmu.c | 1 +
>  1 file changed, 1 insertion(+)
>
> diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
> index 6032f3e..506544f 100644
> --- a/arch/arm64/mm/mmu.c
> +++ b/arch/arm64/mm/mmu.c
> @@ -348,6 +348,7 @@ void __init paging_init(void)
>  {
>         void *zero_page;
>
> +       early_ioremap_reset();
>         map_mem();
>
>         /*
> --
> 2.1.3
>

Re: [PATCH 1/2] arm64: don't make early_*map() calls post paging_init()

[Leif Lindholm]

On Tue, Jan 06, 2015 at 09:38:24AM +0000, Ard Biesheuvel wrote:
> > diff --git a/arch/arm64/kernel/efi.c b/arch/arm64/kernel/efi.c
> > index 6fac253..790adb5 100644
> > --- a/arch/arm64/kernel/efi.c
> > +++ b/arch/arm64/kernel/efi.c

> > @@ -393,6 +385,8 @@ static int __init arm64_enter_virtual_mode(void)
> >                                                    mapsize);
> >         memmap.map_end = memmap.map + mapsize;
> >
> > +       efi_setup_idmap();
> > +
> 
> I think this is incorrect: we may call early_alloc() here which is
> only allowed before paging_init()

Hmm...
Technically, the existing code is already breaking this restriction
then - but my patch moved that _way_ later in the boot process (early
initcall instead of setup_arch()).

I'll resend with an alternative solution that leaves the
efi_setup_idmap() ordering vs paging init() in its current state.
Since efi_setup_idmap() goes away with your "stable UEFI virtual
mappings for kexec" series, I don't want to make changes here and for
potential backports too invasive.

/
    Leif

[PATCH 1/2] arm64: don't make early_*map() calls post paging_init()

[Leif Lindholm]

On Tue, Jan 06, 2015 at 09:38:24AM +0000, Ard Biesheuvel wrote:
> > diff --git a/arch/arm64/kernel/efi.c b/arch/arm64/kernel/efi.c
> > index 6fac253..790adb5 100644
> > --- a/arch/arm64/kernel/efi.c
> > +++ b/arch/arm64/kernel/efi.c

> > @@ -393,6 +385,8 @@ static int __init arm64_enter_virtual_mode(void)
> >                                                    mapsize);
> >         memmap.map_end = memmap.map + mapsize;
> >
> > +       efi_setup_idmap();
> > +
> 
> I think this is incorrect: we may call early_alloc() here which is
> only allowed before paging_init()

Hmm...
Technically, the existing code is already breaking this restriction
then - but my patch moved that _way_ later in the boot process (early
initcall instead of setup_arch()).

I'll resend with an alternative solution that leaves the
efi_setup_idmap() ordering vs paging init() in its current state.
Since efi_setup_idmap() goes away with your "stable UEFI virtual
mappings for kexec" series, I don't want to make changes here and for
potential backports too invasive.

/
    Leif