[PATCH] arm64: entry: simplify trampoline data page
Ard Biesheuvel
ardb at kernel.org
Wed Jun 22 08:27:35 PDT 2022
On Wed, 22 Jun 2022 at 17:05, Mark Rutland <mark.rutland at arm.com> wrote:
>
> On Wed, Jun 22, 2022 at 04:41:41PM +0200, Ard Biesheuvel wrote:
> > On Wed, 27 Apr 2022 at 12:23, Ard Biesheuvel <ardb at kernel.org> wrote:
> > >
> > > Get rid of some clunky open coded arithmetic on section addresses, by
> > > emitting the trampoline data variables into a separate, dedicated r/o
> > > data section, and putting it at the next page boundary. This way, we can
> > > access the literals via single LDR instruction.
> > >
> > > While at it, get rid of other, implicit literals, and use ADRP/ADD or
> > > MOVZ/MOVK sequences, as appropriate. Note that the latter are only
> > > supported for CONFIG_RELOCATABLE=n (which is usually the case if
> > > CONFIG_RANDOMIZE_BASE=n), so update the CPP conditionals to reflect
> > > this.
> > >
> > > Cc: James Morse <james.morse at arm.com>
> > > Signed-off-by: Ard Biesheuvel <ardb at kernel.org>
> >
> > Ping, in case this one slipped behind the desk.
>
> Sorry for the delay. THis has been on my queue of things to look at with a
> bunch of other stuff, and I've had some difficulty prioritizing all that.
>
Yeah, no worries.
> >
> > > ---
> > > arch/arm64/include/asm/fixmap.h | 4 +-
> > > arch/arm64/kernel/entry.S | 45 ++++++--------------
> > > arch/arm64/kernel/vmlinux.lds.S | 3 +-
> > > arch/arm64/mm/mmu.c | 10 ++---
> > > 4 files changed, 22 insertions(+), 40 deletions(-)
> > >
> > > diff --git a/arch/arm64/include/asm/fixmap.h b/arch/arm64/include/asm/fixmap.h
> > > index daff882883f9..71ed5fdf718b 100644
> > > --- a/arch/arm64/include/asm/fixmap.h
> > > +++ b/arch/arm64/include/asm/fixmap.h
> > > @@ -62,10 +62,12 @@ enum fixed_addresses {
> > > #endif /* CONFIG_ACPI_APEI_GHES */
> > >
> > > #ifdef CONFIG_UNMAP_KERNEL_AT_EL0
> > > +#ifdef CONFIG_RELOCATABLE
> > > + FIX_ENTRY_TRAMP_TEXT4, /* one extra slot for the data page */
> > > +#endif
> > > FIX_ENTRY_TRAMP_TEXT3,
> > > FIX_ENTRY_TRAMP_TEXT2,
> > > FIX_ENTRY_TRAMP_TEXT1,
> > > - FIX_ENTRY_TRAMP_DATA,
> > > #define TRAMP_VALIAS (__fix_to_virt(FIX_ENTRY_TRAMP_TEXT1))
> > > #endif /* CONFIG_UNMAP_KERNEL_AT_EL0 */
> > > __end_of_permanent_fixed_addresses,
> > > diff --git a/arch/arm64/kernel/entry.S b/arch/arm64/kernel/entry.S
> > > index ede028dee81b..aed2b41e05aa 100644
> > > --- a/arch/arm64/kernel/entry.S
> > > +++ b/arch/arm64/kernel/entry.S
> > > @@ -636,18 +636,20 @@ alternative_else_nop_endif
> > > */
> > > .endm
> > >
> > > - .macro tramp_data_page dst
> > > - adr_l \dst, .entry.tramp.text
> > > - sub \dst, \dst, PAGE_SIZE
> > > - .endm
> > > -
> > > - .macro tramp_data_read_var dst, var
> > > -#ifdef CONFIG_RANDOMIZE_BASE
> > > - tramp_data_page \dst
> > > - add \dst, \dst, #:lo12:__entry_tramp_data_\var
> > > - ldr \dst, [\dst]
> > > + .macro tramp_data_read_var dst, var
> > > +#ifdef CONFIG_RELOCATABLE
> > > + ldr \dst, .L__tramp_data_\var
> > > + .ifndef .L__tramp_data_\var
> > > + .pushsection ".entry.tramp.rodata", "a", %progbits
> > > + .align 3
> > > +.L__tramp_data_\var:
> > > + .quad \var
> > > + .popsection
> > > + .endif
> > > #else
> > > - ldr \dst, =\var
> > > + movz \dst, :abs_g2_s:\var
> > > + movk \dst, :abs_g1_nc:\var
> > > + movk \dst, :abs_g0_nc:\var
> > > #endif
> > > .endm
>
> Given the lack of a g3 reloc, I assumme `var` is always an address, and we're
> assuming it's always in the upper 48-bits? I think it'd be worth a comment as
> to why this is safe, or just use a g3 reloc since then it's always good per
> inspection.
>
Upper 47 bits, yes. This is because since the 52-bit VA address space
overhaul, the kernel, fixmap and anything else we may want to address
statically here will always be in the upper 47-bit addressable part of
the address space. The abs_g2_s relocation sign extends that into the
bits above.
I opted for as few instructions as required, as these sequences are
emitted into the vector table.
> I'm a bit confused that we've put the var into the literal; I thought the idea
> here was that it was secret and needed to be placed in a page not mapped during
> userspace. Is the assumption there that it's pointless for !RELOCATABLE kernels
> since it can be known anyway, have I misunderstood, or something else?
>
Basically, yes. !RELOCATABLE implies !RANDOMIZE_BASE, and so the
kernel will be running from a known address anyway. So if you are
using KPTI without KASLR, there is no need to use a literal load here.
> Otherwise this all looks good superficially; I just haven't had the time to
> page it all in.
>
> Mark.
>
> > >
> > > @@ -695,7 +697,7 @@ alternative_else_nop_endif
> > > msr vbar_el1, x30
> > > isb
> > > .else
> > > - ldr x30, =vectors
> > > + adr_l x30, vectors
> > > .endif // \kpti == 1
> > >
> > > .if \bhb == BHB_MITIGATION_FW
> > > @@ -764,24 +766,7 @@ SYM_CODE_END(tramp_exit_native)
> > > SYM_CODE_START(tramp_exit_compat)
> > > tramp_exit 32
> > > SYM_CODE_END(tramp_exit_compat)
> > > -
> > > - .ltorg
> > > .popsection // .entry.tramp.text
> > > -#ifdef CONFIG_RANDOMIZE_BASE
> > > - .pushsection ".rodata", "a"
> > > - .align PAGE_SHIFT
> > > -SYM_DATA_START(__entry_tramp_data_start)
> > > -__entry_tramp_data_vectors:
> > > - .quad vectors
> > > -#ifdef CONFIG_ARM_SDE_INTERFACE
> > > -__entry_tramp_data___sdei_asm_handler:
> > > - .quad __sdei_asm_handler
> > > -#endif /* CONFIG_ARM_SDE_INTERFACE */
> > > -__entry_tramp_data_this_cpu_vector:
> > > - .quad this_cpu_vector
> > > -SYM_DATA_END(__entry_tramp_data_start)
> > > - .popsection // .rodata
> > > -#endif /* CONFIG_RANDOMIZE_BASE */
> > > #endif /* CONFIG_UNMAP_KERNEL_AT_EL0 */
> > >
> > > /*
> > > @@ -932,7 +917,6 @@ NOKPROBE(call_on_irq_stack)
> > > * This clobbers x4, __sdei_handler() will restore this from firmware's
> > > * copy.
> > > */
> > > -.ltorg
> > > .pushsection ".entry.tramp.text", "ax"
> > > SYM_CODE_START(__sdei_asm_entry_trampoline)
> > > mrs x4, ttbr1_el1
> > > @@ -967,7 +951,6 @@ SYM_CODE_START(__sdei_asm_exit_trampoline)
> > > 1: sdei_handler_exit exit_mode=x2
> > > SYM_CODE_END(__sdei_asm_exit_trampoline)
> > > NOKPROBE(__sdei_asm_exit_trampoline)
> > > - .ltorg
> > > .popsection // .entry.tramp.text
> > > #endif /* CONFIG_UNMAP_KERNEL_AT_EL0 */
> > >
> > > diff --git a/arch/arm64/kernel/vmlinux.lds.S b/arch/arm64/kernel/vmlinux.lds.S
> > > index edaf0faf766f..17e554be9198 100644
> > > --- a/arch/arm64/kernel/vmlinux.lds.S
> > > +++ b/arch/arm64/kernel/vmlinux.lds.S
> > > @@ -117,7 +117,8 @@ jiffies = jiffies_64;
> > > __entry_tramp_text_start = .; \
> > > *(.entry.tramp.text) \
> > > . = ALIGN(PAGE_SIZE); \
> > > - __entry_tramp_text_end = .;
> > > + __entry_tramp_text_end = .; \
> > > + *(.entry.tramp.rodata)
> > > #else
> > > #define TRAMP_TEXT
> > > #endif
> > > diff --git a/arch/arm64/mm/mmu.c b/arch/arm64/mm/mmu.c
> > > index 626ec32873c6..be4d6c3f5692 100644
> > > --- a/arch/arm64/mm/mmu.c
> > > +++ b/arch/arm64/mm/mmu.c
> > > @@ -665,13 +665,9 @@ static int __init map_entry_trampoline(void)
> > > __set_fixmap(FIX_ENTRY_TRAMP_TEXT1 - i,
> > > pa_start + i * PAGE_SIZE, prot);
> > >
> > > - if (IS_ENABLED(CONFIG_RANDOMIZE_BASE)) {
> > > - extern char __entry_tramp_data_start[];
> > > -
> > > - __set_fixmap(FIX_ENTRY_TRAMP_DATA,
> > > - __pa_symbol(__entry_tramp_data_start),
> > > - PAGE_KERNEL_RO);
> > > - }
> > > + if (IS_ENABLED(CONFIG_RELOCATABLE))
> > > + __set_fixmap(FIX_ENTRY_TRAMP_TEXT1 - i,
> > > + pa_start + i * PAGE_SIZE, PAGE_KERNEL_RO);
> > >
> > > return 0;
> > > }
> > > --
> > > 2.30.2
> > >
More information about the linux-arm-kernel
mailing list