The code in secondary_park is currently placed in the .init section. The
kernel reclaims and clears this code when it finishes booting. That
causes the cores parked in it to go to somewhere unpredictable, so we
move this function out of init to make sure the cores stay looping there.
Signed-off-by: Greentime Hu <[email protected]>
Reviewed-by: Anup Patel <[email protected]>
---
arch/riscv/kernel/head.S | 12 +++++++-----
1 file changed, 7 insertions(+), 5 deletions(-)
diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
index f8f996916c5b..276b98f9d0bd 100644
--- a/arch/riscv/kernel/head.S
+++ b/arch/riscv/kernel/head.S
@@ -217,11 +217,6 @@ relocate:
tail smp_callin
#endif
-.align 2
-.Lsecondary_park:
- /* We lack SMP support or have too many harts, so park this hart */
- wfi
- j .Lsecondary_park
END(_start)
#ifdef CONFIG_RISCV_M_MODE
@@ -303,6 +298,13 @@ ENTRY(reset_regs)
END(reset_regs)
#endif /* CONFIG_RISCV_M_MODE */
+.section ".text", "ax",@progbits
+.align 2
+.Lsecondary_park:
+ /* We lack SMP support or have too many harts, so park this hart */
+ wfi
+ j .Lsecondary_park
+
__PAGE_ALIGNED_BSS
/* Empty zero page */
.balign PAGE_SIZE
--
2.17.1
On Thu, 9 Jan 2020, Greentime Hu wrote:
> The code in secondary_park is currently placed in the .init section. The
> kernel reclaims and clears this code when it finishes booting. That
> causes the cores parked in it to go to somewhere unpredictable, so we
> move this function out of init to make sure the cores stay looping there.
>
> Signed-off-by: Greentime Hu <[email protected]>
> Reviewed-by: Anup Patel <[email protected]>
Thanks, the following is what's been queued for v5.5-rc.
- Paul
From: Greentime Hu <[email protected]>
Date: Thu, 9 Jan 2020 11:15:16 +0800
Subject: [PATCH] riscv: make sure the cores stay looping in .Lsecondary_park
The code in secondary_park is currently placed in the .init section. The
kernel reclaims and clears this code when it finishes booting. That
causes the cores parked in it to go to somewhere unpredictable, so we
move this function out of init to make sure the cores stay looping there.
Signed-off-by: Greentime Hu <[email protected]>
Reviewed-by: Anup Patel <[email protected]>
Cc: Andreas Schwab <[email protected]>
Cc: [email protected]
Fixes: 76d2a0493a17d ("RISC-V: Init and Halt Code")
Signed-off-by: Paul Walmsley <[email protected]>
---
arch/riscv/kernel/head.S | 12 +++++++-----
1 file changed, 7 insertions(+), 5 deletions(-)
diff --git a/arch/riscv/kernel/head.S b/arch/riscv/kernel/head.S
index 797802c73dee..c9cc44ef7184 100644
--- a/arch/riscv/kernel/head.S
+++ b/arch/riscv/kernel/head.S
@@ -209,11 +209,6 @@ relocate:
tail smp_callin
#endif
-.align 2
-.Lsecondary_park:
- /* We lack SMP support or have too many harts, so park this hart */
- wfi
- j .Lsecondary_park
END(_start)
#ifdef CONFIG_RISCV_M_MODE
@@ -295,6 +290,13 @@ ENTRY(reset_regs)
END(reset_regs)
#endif /* CONFIG_RISCV_M_MODE */
+.section ".text", "ax",@progbits
+.align 2
+.Lsecondary_park:
+ /* We lack SMP support or have too many harts, so park this hart */
+ wfi
+ j .Lsecondary_park
+
__PAGE_ALIGNED_BSS
/* Empty zero page */
.balign PAGE_SIZE
--
2.25.0.rc2
Hi Greentime,
On Thu, 9 Jan 2020, Paul Walmsley wrote:
> On Thu, 9 Jan 2020, Greentime Hu wrote:
>
> > The code in secondary_park is currently placed in the .init section. The
> > kernel reclaims and clears this code when it finishes booting. That
> > causes the cores parked in it to go to somewhere unpredictable, so we
> > move this function out of init to make sure the cores stay looping there.
> >
> > Signed-off-by: Greentime Hu <[email protected]>
> > Reviewed-by: Anup Patel <[email protected]>
>
> Thanks, the following is what's been queued for v5.5-rc.
During final testing, when building the kernel with an initramfs, I hit
the following linker error:
LD .tmp_vmlinux1
arch/riscv/kernel/head.o: in function `.L0 ':(.init.text+0x5c): relocation truncated to fit: R_RISCV_JAL against `.Lsecondary_park'
make[1]: *** [Makefile:1079: vmlinux] Error 1
make: *** [Makefile:326: __build_one_by_one] Error 2
Could you take a look at this?
thanks,
- Paul
Hi Paul,
On Mon, Jan 13, 2020 at 2:12 AM Paul Walmsley <[email protected]> wrote:
>
> Hi Greentime,
>
> On Thu, 9 Jan 2020, Paul Walmsley wrote:
>
> > On Thu, 9 Jan 2020, Greentime Hu wrote:
> >
> > > The code in secondary_park is currently placed in the .init section. The
> > > kernel reclaims and clears this code when it finishes booting. That
> > > causes the cores parked in it to go to somewhere unpredictable, so we
> > > move this function out of init to make sure the cores stay looping there.
> > >
> > > Signed-off-by: Greentime Hu <[email protected]>
> > > Reviewed-by: Anup Patel <[email protected]>
> >
> > Thanks, the following is what's been queued for v5.5-rc.
>
> During final testing, when building the kernel with an initramfs, I hit
> the following linker error:
>
> LD .tmp_vmlinux1
> arch/riscv/kernel/head.o: in function `.L0 ':(.init.text+0x5c): relocation truncated to fit: R_RISCV_JAL against `.Lsecondary_park'
> make[1]: *** [Makefile:1079: vmlinux] Error 1
> make: *** [Makefile:326: __build_one_by_one] Error 2
>
> Could you take a look at this?
I think it is because the sections are too far for bqeu to jump and
the config I used just small enough for it to jump so I didn't see
this bug. Sorry about that.
I tried this fix to boot in Unleashed board.
#ifdef CONFIG_SMP
li t0, CONFIG_NR_CPUS
- bgeu a0, t0, .Lsecondary_park
+ blt a0, t0, .Lgood_cores
+ tail .Lsecondary_park
+.Lgood_cores:
#endif
Hi Greentime,
On Tue, 14 Jan 2020, Greentime Hu wrote:
> I think it is because the sections are too far for bqeu to jump and
> the config I used just small enough for it to jump so I didn't see
> this bug. Sorry about that.
No problem.
> I tried this fix to boot in Unleashed board.
>
> #ifdef CONFIG_SMP
> li t0, CONFIG_NR_CPUS
> - bgeu a0, t0, .Lsecondary_park
> + blt a0, t0, .Lgood_cores
> + tail .Lsecondary_park
> +.Lgood_cores:
> #endif
Looks reasonable to me. Care to update and repost the patch?
- Paul
On Wed, Jan 15, 2020 at 6:55 AM Paul Walmsley <[email protected]> wrote:
>
> Hi Greentime,
>
> On Tue, 14 Jan 2020, Greentime Hu wrote:
>
> > I think it is because the sections are too far for bqeu to jump and
> > the config I used just small enough for it to jump so I didn't see
> > this bug. Sorry about that.
>
> No problem.
>
> > I tried this fix to boot in Unleashed board.
> >
> > #ifdef CONFIG_SMP
> > li t0, CONFIG_NR_CPUS
> > - bgeu a0, t0, .Lsecondary_park
> > + blt a0, t0, .Lgood_cores
> > + tail .Lsecondary_park
> > +.Lgood_cores:
> > #endif
>
> Looks reasonable to me. Care to update and repost the patch?
Yes, Paul. I have sent the v4 patch.
It is tested in Unleashed board.