| Message ID | 1445610134-20528-1-git-send-email-jeremy.linton@arm.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Jeremy, On 23 October 2015 at 16:22, Jeremy Linton <jeremy.linton@arm.com> wrote: > It appears that 64k page kernel's die early, in a somewhat random set nit: kernels > of locations when built without KVM. Most likely during memblock > manipulations (depending on kernel debug options). > It looks like fixup_executable() [with DEBUG_RODATA enabled] is the culprit here: it calls create mapping() with a size that gets rounded up to a multiple of PAGE_SIZE, which means the region that ends up getting its exec bits cleared is larger than it should. Perhaps it is better to fix that instead? > Normally when KVM is built into the kernel it has an explicit > PAGE_SIZE alignment requirement and that forces the text section to be > aligned to PAGE_SIZE. Without it, the alignment granularity is likely to > be 4k. > > This updates the linker script to assure that the the text section is > aligned to a minimum of PAGE_SIZE regardless of build options. > > Signed-off-by: Jeremy Linton <jeremy.linton@arm.com> > --- > arch/arm64/kernel/vmlinux.lds.S | 2 +- > 1 file changed, 1 insertion(+), 1 deletion(-) > > diff --git a/arch/arm64/kernel/vmlinux.lds.S b/arch/arm64/kernel/vmlinux.lds.S > index 8a5d97b..bf7b972 100644 > --- a/arch/arm64/kernel/vmlinux.lds.S > +++ b/arch/arm64/kernel/vmlinux.lds.S > @@ -92,7 +92,7 @@ SECTIONS > HEAD_TEXT > } > ALIGN_DEBUG_RO > - .text : { /* Real text segment */ > + .text ALIGN(PAGE_SIZE) : { /* Real text segment */ > _stext = .; /* Text and read-only data */ > __exception_text_start = .; > *(.exception.text) > -- > 2.4.3 > >
On 10/23/2015 10:34 AM, Ard Biesheuvel wrote: > It looks like fixup_executable() [with DEBUG_RODATA enabled] is the > culprit here: it calls create mapping() with a size that gets rounded > up to a multiple of PAGE_SIZE, which means the region that ends up > getting its exec bits cleared is larger than it should. Perhaps it is > better to fix that instead? Ard, Thanks for taking a look at this! Your probably right. But AFAIK, with just that fix text/ro sections can be partially mapped without ro. It seems to me that putting a little padding in so that they are on separate pages is a good idea, especially since it seems the remaining sections of interest are page aligned, either explicitly in the linker script or via assembly directives in assorted places. Maybe a clearer fix is an additional ifdef in the DEBUG_ALIGN_RODATA stanza which sets ALIGN_DEBUG_RO to align to pagesize if DEBUG_RODATA is set, that way it only takes affect for RODATA. Thanks, Jeremy
diff --git a/arch/arm64/kernel/vmlinux.lds.S b/arch/arm64/kernel/vmlinux.lds.S index 8a5d97b..bf7b972 100644 --- a/arch/arm64/kernel/vmlinux.lds.S +++ b/arch/arm64/kernel/vmlinux.lds.S @@ -92,7 +92,7 @@ SECTIONS HEAD_TEXT } ALIGN_DEBUG_RO - .text : { /* Real text segment */ + .text ALIGN(PAGE_SIZE) : { /* Real text segment */ _stext = .; /* Text and read-only data */ __exception_text_start = .; *(.exception.text)
It appears that 64k page kernel's die early, in a somewhat random set of locations when built without KVM. Most likely during memblock manipulations (depending on kernel debug options). Normally when KVM is built into the kernel it has an explicit PAGE_SIZE alignment requirement and that forces the text section to be aligned to PAGE_SIZE. Without it, the alignment granularity is likely to be 4k. This updates the linker script to assure that the the text section is aligned to a minimum of PAGE_SIZE regardless of build options. Signed-off-by: Jeremy Linton <jeremy.linton@arm.com> --- arch/arm64/kernel/vmlinux.lds.S | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)