Skip to contentSkip to conversation

riscv: support for Svpbmt and D1 memory types

PATCH Superseded Heiko Stuebner 3 participants 32 messages Mar 24, 2022 – Apr 8, 2022

Subsystems (2)

v8 of 10 Superseded Latest is v10 →

lore
Share & tools
lore
#0 lore raw

v8 cover riscv: support for Svpbmt and D1 memory types

Svpbmt is an extension defining "Supervisor-mode: page-based memory types" for things like non-cacheable pages or I/O memory pages.

So this is my 2nd try at implementing Svpbmt (and the diverging D1 memory types) using the alternatives framework.

This includes a number of changes to the alternatives mechanism itself. The biggest one being the move to a more central location, as I expect in the future, nearly every chip needing some sort of patching, be it either for erratas or for optional features (svpbmt or others).

Detection of the svpbmt functionality is done via Atish's isa extension handling series [0] and thus does not need any dt-parsing of its own anymore.

The series also introduces support for the memory types of the D1 which are implemented differently to svpbmt. But when patching anyway it's pretty clean to add the D1 variant via ALTERNATIVE_2 to the same location.

The only slightly bigger difference is that the "normal" type is not 0 as with svpbmt, so kernel patches for this PMA type need to be applied even before the MMU is brought up, so the series introduces a separate stage for that.

In theory this series is 3 parts:
- sbi cache-flush / null-ptr
- alternatives improvements
- svpbmt+d1

So expecially patches from the first 2 areas could be applied when deemed ready, I just thought to keep it together to show-case where the end-goal is and not requiring jumping between different series.

I picked the recipient list from the previous versions, hopefully I didn't forget anybody.

changes in v8:
- rebase onto 5.17-final + isa extension series We're halfway through the merge-window, so this series should be merge after that
- adapt to fix limiting alternatives to non-xip-kernels
- add .norelax option for alternatives
- fix unused cpu_apply_errata in thead errata
- don't use static globals to store cpu-manufacturer data as it makes machines hang if done too early

changes in v7:
- fix typo in patch1 (Atish)
- moved to Atish's isa-extension framework
- and therefore move regular boot-alternatives directly behind fill_hwcaps
- change T-Head errata Kconfig text (Atish)

changes in v6:
- rebase onto 5.17-rc1
- handle sbi null-ptr differently
- improve commit messages
- use riscv,mmu as property name

changes in v5:
- move to use alternatives for runtime-patching
- add D1 variant

[0] https://lore.kernel.org/r/[email protected]

Heiko Stuebner (13):
riscv: prevent null-pointer dereference with sbi_remote_fence_i
riscv: integrate alternatives better into the main architecture
riscv: allow different stages with alternatives
riscv: implement module alternatives
riscv: implement ALTERNATIVE_2 macro
riscv: extend concatenated alternatives-lines to the same length
riscv: prevent compressed instructions in alternatives
riscv: move boot alternatives to after fill_hwcap
riscv: Fix accessing pfn bits in PTEs for non-32bit variants
riscv: add cpufeature handling via alternatives
riscv: remove FIXMAP_PAGE_IO and fall back to its default value
riscv: don't use global static vars to store alternative data
riscv: add memory-type errata for T-Head

Wei Fu (1):
riscv: add RISC-V Svpbmt extension support

 arch/riscv/Kconfig.erratas                  |  31 +++--
 arch/riscv/Kconfig.socs                     |   1 -
 arch/riscv/Makefile                         |   2 +-
 arch/riscv/errata/Makefile                  |   2 +-
 arch/riscv/errata/alternative.c             |  75 ------------
 arch/riscv/errata/sifive/errata.c           |  17 ++-
 arch/riscv/errata/thead/Makefile            |   1 +
 arch/riscv/errata/thead/errata.c            |  82 +++++++++++++
 arch/riscv/include/asm/alternative-macros.h | 121 ++++++++++++++------
 arch/riscv/include/asm/alternative.h        |  16 ++-
 arch/riscv/include/asm/errata_list.h        |  52 +++++++++
 arch/riscv/include/asm/fixmap.h             |   2 -
 arch/riscv/include/asm/hwcap.h              |   1 +
 arch/riscv/include/asm/pgtable-32.h         |  17 +++
 arch/riscv/include/asm/pgtable-64.h         |  79 ++++++++++++-
 arch/riscv/include/asm/pgtable-bits.h       |  10 --
 arch/riscv/include/asm/pgtable.h            |  53 +++++++--
 arch/riscv/include/asm/vendorid_list.h      |   1 +
 arch/riscv/kernel/Makefile                  |   1 +
 arch/riscv/kernel/alternative.c             | 104 +++++++++++++++++
 arch/riscv/kernel/cpu.c                     |   1 +
 arch/riscv/kernel/cpufeature.c              |  80 ++++++++++++-
 arch/riscv/kernel/module.c                  |  29 +++++
 arch/riscv/kernel/sbi.c                     |  10 +-
 arch/riscv/kernel/setup.c                   |   2 +
 arch/riscv/kernel/smpboot.c                 |   4 -
 arch/riscv/kernel/traps.c                   |   2 +-
 arch/riscv/mm/init.c                        |   1 +
 28 files changed, 629 insertions(+), 168 deletions(-)
 delete mode 100644 arch/riscv/errata/alternative.c
 create mode 100644 arch/riscv/errata/thead/Makefile
 create mode 100644 arch/riscv/errata/thead/errata.c
 create mode 100644 arch/riscv/kernel/alternative.c
Signature · 1 line
2.35.1
#1 lore raw

PATCH v8 1/14 riscv: prevent null-pointer dereference with sbi_remote_fence_i

The callback used inside sbi_remote_fence_i is set at sbi probe time to the needed variant. Before that it is a NULL pointer.

Some users like the flush_icache_*() functions suggest a generic functionality, that doesn't depend on a specific boot-stage but uses sbi_remote_fence_i as one option to flush other cpu cores.

So they definitely shouldn't run into null-pointer dereference issues when called "too early" during boot.

So introduce an empty function to be the standard for the __sbi_rfence function pointer until sbi_init has run.

Users of sbi_remote_fence_i will have separate code for the local cpu and sbi_init() is called before other cpus are brought up. So there are no other cpus present at the time when the issue might happen.

---
 arch/riscv/kernel/sbi.c | 10 +++++++++-
 1 file changed, 9 insertions(+), 1 deletion(-)
arch/riscv/kernel/sbi.c +9 -1
index 775d3322b422..5a60a458c0b7 100644
@@ -16,11 +16,19 @@
 unsigned long sbi_spec_version __ro_after_init = SBI_SPEC_VERSION_DEFAULT;
 EXPORT_SYMBOL(sbi_spec_version);
 
+static int __sbi_rfence_none(int fid, const struct cpumask *cpu_mask,
+			     unsigned long start, unsigned long size,
+			     unsigned long arg4, unsigned long arg5)
+{
+	return -EOPNOTSUPP;
+}
+
 static void (*__sbi_set_timer)(uint64_t stime) __ro_after_init;
 static int (*__sbi_send_ipi)(const struct cpumask *cpu_mask) __ro_after_init;
 static int (*__sbi_rfence)(int fid, const struct cpumask *cpu_mask,
 			   unsigned long start, unsigned long size,
-			   unsigned long arg4, unsigned long arg5) __ro_after_init;
+			   unsigned long arg4, unsigned long arg5)
+			   __ro_after_init = __sbi_rfence_none;
 
 struct sbiret sbi_ecall(int ext, int fid, unsigned long arg0,
 			unsigned long arg1, unsigned long arg2,
Signature · 1 line
2.35.1
Nacked-by on v8 1/14 NAK
#2 lore raw

Re: [PATCH v8 01/14] riscv: prevent null-pointer dereference with sbi_remote_fence_i

Quoted 10 lines · Heiko Stuebner

The callback used inside sbi_remote_fence_i is set at sbi probe time to the needed variant. Before that it is a NULL pointer.

Some users like the flush_icache_*() functions suggest a generic functionality, that doesn't depend on a specific boot-stage but uses sbi_remote_fence_i as one option to flush other cpu cores.

So they definitely shouldn't run into null-pointer dereference issues when called "too early" during boot.

Quoted 7 lines

So introduce an empty function to be the standard for the __sbi_rfence function pointer until sbi_init has run.

Users of sbi_remote_fence_i will have separate code for the local cpu and sbi_init() is called before other cpus are brought up. So there are no other cpus present at the time when the issue might happen.

I don't really understand this changelog. If flush_icache_* or other routines using SBI calls are called too early they won't do what they are asked to do, which implies a bug in the code.

So crashing absolutely is the right thing to do here as we don't really have any other error reporting method available.

So unless I'm totally misunderstanding what you are saying here:

#3 lore raw

Re: [PATCH v8 01/14] riscv: prevent null-pointer dereference with sbi_remote_fence_i

Hi,

Am Donnerstag, 31. März 2022, 11:51:55 CEST schrieb Christoph Hellwig:

Quoted 23 lines · Heiko Stuebner
Quoted 10 lines

The callback used inside sbi_remote_fence_i is set at sbi probe time to the needed variant. Before that it is a NULL pointer.

Some users like the flush_icache_*() functions suggest a generic functionality, that doesn't depend on a specific boot-stage but uses sbi_remote_fence_i as one option to flush other cpu cores.

So they definitely shouldn't run into null-pointer dereference issues when called "too early" during boot.

Quoted 7 lines

So introduce an empty function to be the standard for the __sbi_rfence function pointer until sbi_init has run.

Users of sbi_remote_fence_i will have separate code for the local cpu and sbi_init() is called before other cpus are brought up. So there are no other cpus present at the time when the issue might happen.

I don't really understand this changelog. If flush_icache_* or other routines using SBI calls are called too early they won't do what they are asked to do, which implies a bug in the code.

Quoted 6 lines

So crashing absolutely is the right thing to do here as we don't really have any other error reporting method available.

So unless I'm totally misunderstanding what you are saying here:

The function is defined as

void flush_icache_all(void)

{
	local_flush_icache_all();

	if (IS_ENABLED(CONFIG_RISCV_SBI))
		sbi_remote_fence_i(NULL);
	else
		on_each_cpu(ipi_remote_fence_i, NULL, 1);
}

so essentially flushes the _local_ icache first and then tries to flush caches on other cores, either via an ipi or via sbi.

The remote-fence callback is set correctly during sbi_init(). The other cores are only brought up after sbi-init is done.

So it's not really about error reporting but making sure that flush_icache_all() does something sane even when still running on the first core.
As I assume the "all" means on all available cores (which would be the core the system booted on).

Does this make it clearer what this tries to solve? Heiko

#4 lore raw

Re: [PATCH v8 01/14] riscv: prevent null-pointer dereference with sbi_remote_fence_i

Quoted 6 lines · Heiko Stübner

so essentially flushes the _local_ icache first and then tries to flush caches on other cores, either via an ipi or via sbi.

The remote-fence callback is set correctly during sbi_init(). The other cores are only brought up after sbi-init is done.

Quoted 6 lines

So it's not really about error reporting but making sure that flush_icache_all() does something sane even when still running on the first core.
As I assume the "all" means on all available cores (which would be the core the system booted on).

Does this make it clearer what this tries to solve?

A little. Whatever code calls this early still seems broken to me and just just do a local flush, though.

#5 lore raw

Re: [PATCH v8 01/14] riscv: prevent null-pointer dereference with sbi_remote_fence_i

Am Donnerstag, 31. März 2022, 14:33:22 CEST schrieb Christoph Hellwig:

Quoted 12 lines · Heiko Stübner
Quoted 11 lines

so essentially flushes the _local_ icache first and then tries to flush caches on other cores, either via an ipi or via sbi.

The remote-fence callback is set correctly during sbi_init(). The other cores are only brought up after sbi-init is done.

So it's not really about error reporting but making sure that flush_icache_all() does something sane even when still running on the first core.
As I assume the "all" means on all available cores (which would be the core the system booted on).

Quoted 4 lines
Quoted 1 line

Does this make it clearer what this tries to solve?

A little. Whatever code calls this early still seems broken to me and just just do a local flush, though.

Looking at this again made me realize that we don't need this patch anymore at all.

In a previous revision the "main" alternatives apply-point was earlier, triggering the issue when the generic patching-functions wanted to flush the cache, but with the alternatives getting applied where they are now, the issue is not triggered anymore.

And the "early" alternatives use a local flush anyway.

#6 lore raw

PATCH v8 2/14 riscv: integrate alternatives better into the main architecture

In reply to Heiko Stuebner

Right now the alternatives need to be explicitly enabled and erratas are limited to SiFive ones.

Over time with more SoCs and additional RiscV extensions, many more erratas or other patch-worthy features will emerge, so it doesn't really make sense to have the core alternatives able to get deactivated.

So make it part of the core RiscV kernel and drop the main RISCV_ERRATA_ALTERNATIVES config symbol.

This mimics how other architectures like for example arm64 handle their alternatives implementation.

---
 arch/riscv/Kconfig.erratas                  | 12 +-----------
 arch/riscv/Kconfig.socs                     |  1 -
 arch/riscv/Makefile                         |  2 +-
 arch/riscv/errata/Makefile                  |  1 -
 arch/riscv/include/asm/alternative-macros.h | 15 ++++++++-------
 arch/riscv/kernel/Makefile                  |  1 +
 arch/riscv/{errata => kernel}/alternative.c |  0
 arch/riscv/kernel/smpboot.c                 |  2 --
 arch/riscv/kernel/traps.c                   |  2 +-
 9 files changed, 12 insertions(+), 24 deletions(-)
 rename arch/riscv/{errata => kernel}/alternative.c (100%)
arch/riscv/Kconfig.erratas +1 -11
index 0aacd7052585..15cdeeb4151f 100644
@@ -1,18 +1,8 @@
 menu "CPU errata selection"
 
-config RISCV_ERRATA_ALTERNATIVE
-	bool "RISC-V alternative scheme"
-	depends on !XIP_KERNEL
-	default y
-	help
-	  This Kconfig allows the kernel to automatically patch the
-	  errata required by the execution platform at run time. The
-	  code patching is performed once in the boot stages. It means
-	  that the overhead from this mechanism is just taken once.
-
 config ERRATA_SIFIVE
 	bool "SiFive errata"
-	depends on RISCV_ERRATA_ALTERNATIVE
+	depends on !XIP_KERNEL
 	help
 	  All SiFive errata Kconfig depend on this Kconfig. Disabling
 	  this Kconfig will disable all SiFive errata. Please say "Y"
arch/riscv/Kconfig.socs +0 -1
index c112ab2a9052..236d7d46a6f1 100644
@@ -14,7 +14,6 @@ config SOC_SIFIVE
 	select CLK_SIFIVE
 	select CLK_SIFIVE_PRCI
 	select SIFIVE_PLIC
-	select RISCV_ERRATA_ALTERNATIVE if !XIP_KERNEL
 	select ERRATA_SIFIVE if !XIP_KERNEL
 	help
 	  This enables support for SiFive SoC platform hardware.
arch/riscv/Makefile +1 -1
index 7d81102cffd4..a7ed47ce9311 100644
@@ -103,7 +103,7 @@ endif
 
 head-y := arch/riscv/kernel/head.o
 
-core-$(CONFIG_RISCV_ERRATA_ALTERNATIVE) += arch/riscv/errata/
+core-y += arch/riscv/errata/
 core-$(CONFIG_KVM) += arch/riscv/kvm/
 
 libs-y += arch/riscv/lib/
arch/riscv/errata/Makefile +0 -1
index b8f8740a3e44..0ca1c5281a2d 100644
@@ -1,2 +1 @@
-obj-y	+= alternative.o
 obj-$(CONFIG_ERRATA_SIFIVE) += sifive/
arch/riscv/include/asm/alternative-macros.h +8 -7
index 67406c376389..e33d4bc54019 100644
@@ -2,7 +2,7 @@
 #ifndef __ASM_ALTERNATIVE_MACROS_H
 #define __ASM_ALTERNATIVE_MACROS_H
 
-#ifdef CONFIG_RISCV_ERRATA_ALTERNATIVE
+#ifndef CONFIG_XIP_KERNEL
 
 #ifdef __ASSEMBLY__
 
@@ -76,26 +76,27 @@
 
 #endif /* __ASSEMBLY__ */
 
-#else /* !CONFIG_RISCV_ERRATA_ALTERNATIVE*/
+#else /* CONFIG_XIP_KERNEL */
 #ifdef __ASSEMBLY__
 
 .macro __ALTERNATIVE_CFG old_c
-	\old_c
+       \old_c
 .endm
 
 #define _ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, CONFIG_k) \
-	__ALTERNATIVE_CFG old_c
+       __ALTERNATIVE_CFG old_c
 
 #else /* !__ASSEMBLY__ */
 
 #define __ALTERNATIVE_CFG(old_c)  \
-	old_c "\n"
+       old_c "\n"
 
 #define _ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, CONFIG_k) \
-	__ALTERNATIVE_CFG(old_c)
+       __ALTERNATIVE_CFG(old_c)
 
 #endif /* __ASSEMBLY__ */
-#endif /* CONFIG_RISCV_ERRATA_ALTERNATIVE */
+#endif /* CONFIG_XIP_KERNEL */
+
 /*
  * Usage:
  *   ALTERNATIVE(old_content, new_content, vendor_id, errata_id, CONFIG_k)
arch/riscv/kernel/Makefile +1 -0
index ffc87e76b1dd..b02142517f7f 100644
@@ -18,6 +18,7 @@ extra-y += head.o
 extra-y += vmlinux.lds
 
 obj-y	+= soc.o
+obj-y	+= alternative.o
 obj-y	+= cpu.o
 obj-y	+= cpufeature.o
 obj-y	+= entry.o
arch/riscv/kernel/alternative.c renamed from arch/riscv/errata/alternative.c +0 -0
similarity index 100%
rename from arch/riscv/errata/alternative.c
rename to arch/riscv/kernel/alternative.c
arch/riscv/kernel/smpboot.c +0 -2
index 622f226454d5..a6d13dca1403 100644
@@ -41,9 +41,7 @@ static DECLARE_COMPLETION(cpu_running);
 void __init smp_prepare_boot_cpu(void)
 {
 	init_cpu_topology();
-#ifdef CONFIG_RISCV_ERRATA_ALTERNATIVE
 	apply_boot_alternatives();
-#endif
 }
 
 void __init smp_prepare_cpus(unsigned int max_cpus)
arch/riscv/kernel/traps.c +1 -1
index fe92e119e6a3..9984c8622c3b 100644
@@ -86,7 +86,7 @@ static void do_trap_error(struct pt_regs *regs, int signo, int code,
 	}
 }
 
-#if defined (CONFIG_XIP_KERNEL) && defined (CONFIG_RISCV_ERRATA_ALTERNATIVE)
+#if defined (CONFIG_XIP_KERNEL)
 #define __trap_section		__section(".xip.traps")
 #else
 #define __trap_section
Signature · 7 lines
2.35.1


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
#7 lore raw

Re: [PATCH v8 02/14] riscv: integrate alternatives better into the main architecture

Quoted 8 lines · Heiko Stuebner

Right now the alternatives need to be explicitly enabled and erratas are limited to SiFive ones.

Over time with more SoCs and additional RiscV extensions, many more erratas or other patch-worthy features will emerge, so it doesn't really make sense to have the core alternatives able to get deactivated.

Quoted 5 lines

So make it part of the core RiscV kernel and drop the main RISCV_ERRATA_ALTERNATIVES config symbol.

This mimics how other architectures like for example arm64 handle their alternatives implementation.

For minimal kernels like the k210 it would be really good to be able to avoid any not strictly neeed code. So I'd much rather have the alternatives mechanism only built when it actually is needed, not (semi-)unconditional.

#8 lore raw

PATCH v8 3/14 riscv: allow different stages with alternatives

In reply to Heiko Stuebner

Future features may need to be applied at a different time during boot, so allow defining stages for alternatives and handling them differently depending on the stage.

Also make the alternatives-location more flexible so that future stages may provide their own location.

---
 arch/riscv/errata/sifive/errata.c    |  3 ++-
 arch/riscv/include/asm/alternative.h |  5 ++++-
 arch/riscv/kernel/alternative.c      | 26 +++++++++++++++++---------
 3 files changed, 23 insertions(+), 11 deletions(-)
arch/riscv/errata/sifive/errata.c +2 -1
index f5e5ae70e829..4fe03ac41fd7 100644
@@ -80,7 +80,8 @@ static void __init warn_miss_errata(u32 miss_errata)
 }
 
 void __init sifive_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
-				     unsigned long archid, unsigned long impid)
+				     unsigned long archid, unsigned long impid,
+				     unsigned int stage)
 {
 	struct alt_entry *alt;
 	u32 cpu_req_errata = sifive_errata_probe(archid, impid);
arch/riscv/include/asm/alternative.h +4 -1
index e625d3cafbed..811bdd8027db 100644
@@ -17,6 +17,8 @@
 #include <linux/stddef.h>
 #include <asm/hwcap.h>
 
+#define RISCV_ALTERNATIVES_BOOT		0 /* alternatives applied during regular boot */
+
 void __init apply_boot_alternatives(void);
 
 struct alt_entry {
@@ -33,7 +35,8 @@ struct errata_checkfunc_id {
 };
 
 void sifive_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
-			      unsigned long archid, unsigned long impid);
+			      unsigned long archid, unsigned long impid,
+			      unsigned int stage);
 
 #endif
 #endif
arch/riscv/kernel/alternative.c +17 -9
index e8b4a0fe488c..02db62f55bac 100644
@@ -22,8 +22,8 @@ static struct cpu_manufacturer_info_t {
 } cpu_mfr_info;
 
 static void (*vendor_patch_func)(struct alt_entry *begin, struct alt_entry *end,
-				 unsigned long archid,
-				 unsigned long impid) __initdata;
+				 unsigned long archid, unsigned long impid,
+				 unsigned int stage) __initdata;
 
 static inline void __init riscv_fill_cpu_mfr_info(void)
 {
@@ -58,6 +58,18 @@ static void __init init_alternative(void)
  * a feature detect on the boot CPU). No need to worry about other CPUs
  * here.
  */
+static void __init _apply_alternatives(struct alt_entry *begin,
+				       struct alt_entry *end,
+				       unsigned int stage)
+{
+	if (!vendor_patch_func)
+		return;
+
+	vendor_patch_func(begin, end,
+			  cpu_mfr_info.arch_id, cpu_mfr_info.imp_id,
+			  stage);
+}
+
 void __init apply_boot_alternatives(void)
 {
 	/* If called on non-boot cpu things could go wrong */
@@ -65,11 +77,7 @@ void __init apply_boot_alternatives(void)
 
 	init_alternative();
 
-	if (!vendor_patch_func)
-		return;
-
-	vendor_patch_func((struct alt_entry *)__alt_start,
-			  (struct alt_entry *)__alt_end,
-			  cpu_mfr_info.arch_id, cpu_mfr_info.imp_id);
+	_apply_alternatives((struct alt_entry *)__alt_start,
+			    (struct alt_entry *)__alt_end,
+			    RISCV_ALTERNATIVES_BOOT);
 }
-
Signature · 1 line
2.35.1
#9 lore raw

PATCH v8 4/14 riscv: implement module alternatives

In reply to Heiko Stuebner

This allows alternatives to also be applied when loading modules and follows the implementation of other architectures (e.g. arm64).

---
 arch/riscv/errata/sifive/errata.c    |  7 ++++---
 arch/riscv/include/asm/alternative.h |  3 ++-
 arch/riscv/kernel/alternative.c      | 18 +++++++++++++----
 arch/riscv/kernel/module.c           | 29 ++++++++++++++++++++++++++++
 4 files changed, 49 insertions(+), 8 deletions(-)
arch/riscv/errata/sifive/errata.c +4 -3
index 4fe03ac41fd7..42457efcaaa3 100644
@@ -4,6 +4,7 @@
  */
 
 #include <linux/kernel.h>
+#include <linux/module.h>
 #include <linux/string.h>
 #include <linux/bug.h>
 #include <asm/patch.h>
@@ -54,7 +55,7 @@ static struct errata_info_t errata_list[ERRATA_SIFIVE_NUMBER] = {
 	},
 };
 
-static u32 __init sifive_errata_probe(unsigned long archid, unsigned long impid)
+static u32 __init_or_module sifive_errata_probe(unsigned long archid, unsigned long impid)
 {
 	int idx;
 	u32 cpu_req_errata = 0;
@@ -66,7 +67,7 @@ static u32 __init sifive_errata_probe(unsigned long archid, unsigned long impid)
 	return cpu_req_errata;
 }
 
-static void __init warn_miss_errata(u32 miss_errata)
+static void __init_or_module warn_miss_errata(u32 miss_errata)
 {
 	int i;
 
@@ -79,7 +80,7 @@ static void __init warn_miss_errata(u32 miss_errata)
 	pr_warn("----------------------------------------------------------------\n");
 }
 
-void __init sifive_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
+void __init_or_module sifive_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
 				     unsigned long archid, unsigned long impid,
 				     unsigned int stage)
 {
arch/riscv/include/asm/alternative.h +2 -1
index 811bdd8027db..f0657b1b3174 100644
@@ -18,8 +18,10 @@
 #include <asm/hwcap.h>
 
 #define RISCV_ALTERNATIVES_BOOT		0 /* alternatives applied during regular boot */
+#define RISCV_ALTERNATIVES_MODULE	1 /* alternatives applied during module-init */
 
 void __init apply_boot_alternatives(void);
+void apply_module_alternatives(void *start, size_t length);
 
 struct alt_entry {
 	void *old_ptr;		 /* address of original instruciton or data  */
@@ -37,6 +39,5 @@ struct errata_checkfunc_id {
 void sifive_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
 			      unsigned long archid, unsigned long impid,
 			      unsigned int stage);
-
 #endif
 #endif
arch/riscv/kernel/alternative.c +14 -4
index 02db62f55bac..223770b3945c 100644
@@ -7,6 +7,7 @@
  */
 
 #include <linux/init.h>
+#include <linux/module.h>
 #include <linux/cpu.h>
 #include <linux/uaccess.h>
 #include <asm/alternative.h>
@@ -23,7 +24,7 @@ static struct cpu_manufacturer_info_t {
 
 static void (*vendor_patch_func)(struct alt_entry *begin, struct alt_entry *end,
 				 unsigned long archid, unsigned long impid,
-				 unsigned int stage) __initdata;
+				 unsigned int stage) __initdata_or_module;
 
 static inline void __init riscv_fill_cpu_mfr_info(void)
 {
@@ -58,9 +59,9 @@ static void __init init_alternative(void)
  * a feature detect on the boot CPU). No need to worry about other CPUs
  * here.
  */
-static void __init _apply_alternatives(struct alt_entry *begin,
-				       struct alt_entry *end,
-				       unsigned int stage)
+static void __init_or_module _apply_alternatives(struct alt_entry *begin,
+						 struct alt_entry *end,
+						 unsigned int stage)
 {
 	if (!vendor_patch_func)
 		return;
@@ -81,3 +82,12 @@ void __init apply_boot_alternatives(void)
 			    (struct alt_entry *)__alt_end,
 			    RISCV_ALTERNATIVES_BOOT);
 }
+
+#ifdef CONFIG_MODULES
+void apply_module_alternatives(void *start, size_t length)
+{
+	_apply_alternatives((struct alt_entry *)start,
+			    (struct alt_entry *)(start + length),
+			    RISCV_ALTERNATIVES_MODULE);
+}
+#endif
arch/riscv/kernel/module.c +29 -0
index 4a48287513c3..823b7ad47c9a 100644
@@ -11,6 +11,7 @@
 #include <linux/vmalloc.h>
 #include <linux/sizes.h>
 #include <linux/pgtable.h>
+#include <asm/alternative.h>
 #include <asm/sections.h>
 
 /*
@@ -427,3 +428,31 @@ void *module_alloc(unsigned long size)
 				    __builtin_return_address(0));
 }
 #endif
+
+static const Elf_Shdr *find_section(const Elf_Ehdr *hdr,
+				    const Elf_Shdr *sechdrs,
+				    const char *name)
+{
+	const Elf_Shdr *s, *se;
+	const char *secstrs = (void *)hdr + sechdrs[hdr->e_shstrndx].sh_offset;
+
+	for (s = sechdrs, se = sechdrs + hdr->e_shnum; s < se; s++) {
+		if (strcmp(name, secstrs + s->sh_name) == 0)
+			return s;
+	}
+
+	return NULL;
+}
+
+int module_finalize(const Elf_Ehdr *hdr,
+		    const Elf_Shdr *sechdrs,
+		    struct module *me)
+{
+	const Elf_Shdr *s;
+
+	s = find_section(hdr, sechdrs, ".alternative");
+	if (s)
+		apply_module_alternatives((void *)s->sh_addr, s->sh_size);
+
+	return 0;
+}
Signature · 7 lines
2.35.1


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
#10 lore raw

PATCH v8 5/14 riscv: implement ALTERNATIVE_2 macro

In reply to Heiko Stuebner

When the alternatives were added the commit already provided a template on how to implement 2 different alternatives for one piece of code.

Make this usable.

---
 arch/riscv/include/asm/alternative-macros.h | 60 ++++++++++++++-------
 1 file changed, 40 insertions(+), 20 deletions(-)
arch/riscv/include/asm/alternative-macros.h +40 -20
index e33d4bc54019..259d8246d612 100644
@@ -39,6 +39,20 @@
 #define _ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, CONFIG_k) \
 	__ALTERNATIVE_CFG old_c, new_c, vendor_id, errata_id, IS_ENABLED(CONFIG_k)
 
+.macro __ALTERNATIVE_CFG_2 old_c, new_c_1, vendor_id_1, errata_id_1, enable_1, \
+				  new_c_2, vendor_id_2, errata_id_2, enable_2
+886 :
+	\old_c
+887 :
+	ALT_NEW_CONTENT \vendor_id_1, \errata_id_1, \enable_1, \new_c_1
+	ALT_NEW_CONTENT \vendor_id_2, \errata_id_2, \enable_2, \new_c_2
+.endm
+
+#define _ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, CONFIG_k_1, \
+				  new_c_2, vendor_id_2, errata_id_2, CONFIG_k_2) \
+	__ALTERNATIVE_CFG_2 old_c, new_c_1, vendor_id_1, errata_id_1, IS_ENABLED(CONFIG_k_1), \
+				   new_c_2, vendor_id_2, errata_id_2, IS_ENABLED(CONFIG_k_2)
+
 #else /* !__ASSEMBLY__ */
 
 #include <asm/asm.h>
@@ -74,6 +88,19 @@
 #define _ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, CONFIG_k)	\
 	__ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, IS_ENABLED(CONFIG_k))
 
+#define __ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, enable_1, \
+				  new_c_2, vendor_id_2, errata_id_2, enable_2) \
+	"886 :\n"	\
+	old_c "\n"	\
+	"887 :\n"	\
+	ALT_NEW_CONTENT(vendor_id_1, errata_id_1, enable_1, new_c_1) \
+	ALT_NEW_CONTENT(vendor_id_2, errata_id_2, enable_2, new_c_2)
+
+#define _ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, CONFIG_k_1,		\
+				  new_c_2, vendor_id_2, errata_id_2, CONFIG_k_2)		\
+	__ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, IS_ENABLED(CONFIG_k_1),	\
+				   new_c_2, vendor_id_2, errata_id_2, IS_ENABLED(CONFIG_k_2))
+
 #endif /* __ASSEMBLY__ */
 
 #else /* CONFIG_XIP_KERNEL */
@@ -86,6 +113,10 @@
 #define _ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, CONFIG_k) \
        __ALTERNATIVE_CFG old_c
 
+#define _ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, CONFIG_k_1, \
+				  new_c_2, vendor_id_2, errata_id_2, CONFIG_k_2) \
+       __ALTERNATIVE_CFG old_c
+
 #else /* !__ASSEMBLY__ */
 
 #define __ALTERNATIVE_CFG(old_c)  \
@@ -94,6 +125,10 @@
 #define _ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, CONFIG_k) \
        __ALTERNATIVE_CFG(old_c)
 
+#define _ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, CONFIG_k_1, \
+				  new_c_2, vendor_id_2, errata_id_2, CONFIG_k_2) \
+       __ALTERNATIVE_CFG(old_c)
+
 #endif /* __ASSEMBLY__ */
 #endif /* CONFIG_XIP_KERNEL */
 
@@ -119,25 +154,10 @@
  * this case, this vendor can create a new macro ALTERNATIVE_2() based
  * on the following sample code and then replace ALTERNATIVE() with
  * ALTERNATIVE_2() to append its customized content.
- *
- * .macro __ALTERNATIVE_CFG_2 old_c, new_c_1, vendor_id_1, errata_id_1, enable_1, \
- *                                   new_c_2, vendor_id_2, errata_id_2, enable_2
- * 886 :
- *      \old_c
- * 887 :
- *      ALT_NEW_CONTENT \vendor_id_1, \errata_id_1, \enable_1, \new_c_1
- *      ALT_NEW_CONTENT \vendor_id_2, \errata_id_2, \enable_2, \new_c_2
- * .endm
- *
- * #define _ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, CONFIG_k_1, \
- *                                   new_c_2, vendor_id_2, errata_id_2, CONFIG_k_2) \
- *        __ALTERNATIVE_CFG_2 old_c, new_c_1, vendor_id_1, errata_id_1, IS_ENABLED(CONFIG_k_1), \
- *                                   new_c_2, vendor_id_2, errata_id_2, IS_ENABLED(CONFIG_k_2) \
- *
- * #define ALTERNATIVE_2(old_content, new_content_1, vendor_id_1, errata_id_1, CONFIG_k_1, \
- *                                    new_content_2, vendor_id_2, errata_id_2, CONFIG_k_2) \
- *         _ALTERNATIVE_CFG_2(old_content, new_content_1, vendor_id_1, errata_id_1, CONFIG_k_1, \
- *                                         new_content_2, vendor_id_2, errata_id_2, CONFIG_k_2)
- *
  */
+#define ALTERNATIVE_2(old_content, new_content_1, vendor_id_1, errata_id_1, CONFIG_k_1, \
+				   new_content_2, vendor_id_2, errata_id_2, CONFIG_k_2) \
+	_ALTERNATIVE_CFG_2(old_content, new_content_1, vendor_id_1, errata_id_1, CONFIG_k_1, \
+					new_content_2, vendor_id_2, errata_id_2, CONFIG_k_2)
+
 #endif
Signature · 1 line
2.35.1
#12 lore raw

PATCH v8 6/14 riscv: extend concatenated alternatives-lines to the same length

In reply to Heiko Stuebner

ALT_NEW_CONTENT already uses same-length assembler lines, so extend this to the other elements as well.

This makes it more readable when these elements need to be extended in the future.

---
 arch/riscv/include/asm/alternative-macros.h | 30 ++++++++++-----------
 1 file changed, 15 insertions(+), 15 deletions(-)
arch/riscv/include/asm/alternative-macros.h +15 -15
index 259d8246d612..25b4559d743b 100644
@@ -58,14 +58,14 @@
 #include <asm/asm.h>
 #include <linux/stringify.h>
 
-#define ALT_ENTRY(oldptr, newptr, vendor_id, errata_id, newlen) \
-	RISCV_PTR " " oldptr "\n" \
-	RISCV_PTR " " newptr "\n" \
-	REG_ASM " " vendor_id "\n" \
-	REG_ASM " " newlen "\n" \
+#define ALT_ENTRY(oldptr, newptr, vendor_id, errata_id, newlen)		\
+	RISCV_PTR " " oldptr "\n"					\
+	RISCV_PTR " " newptr "\n"					\
+	REG_ASM " " vendor_id "\n"					\
+	REG_ASM " " newlen "\n"						\
 	".word " errata_id "\n"
 
-#define ALT_NEW_CONTENT(vendor_id, errata_id, enable, new_c) \
+#define ALT_NEW_CONTENT(vendor_id, errata_id, enable, new_c)		\
 	".if " __stringify(enable) " == 1\n"				\
 	".pushsection .alternative, \"a\"\n"				\
 	ALT_ENTRY("886b", "888f", __stringify(vendor_id), __stringify(errata_id), "889f - 888f") \
@@ -79,21 +79,21 @@
 	".org	. - (889b - 888b) + (887b - 886b)\n"			\
 	".endif\n"
 
-#define __ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, enable) \
-	"886 :\n"	\
-	old_c "\n"	\
-	"887 :\n"	\
+#define __ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, enable)	\
+	"886 :\n"							\
+	old_c "\n"							\
+	"887 :\n"							\
 	ALT_NEW_CONTENT(vendor_id, errata_id, enable, new_c)
 
-#define _ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, CONFIG_k)	\
+#define _ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, CONFIG_k) \
 	__ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, IS_ENABLED(CONFIG_k))
 
 #define __ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, enable_1, \
 				  new_c_2, vendor_id_2, errata_id_2, enable_2) \
-	"886 :\n"	\
-	old_c "\n"	\
-	"887 :\n"	\
-	ALT_NEW_CONTENT(vendor_id_1, errata_id_1, enable_1, new_c_1) \
+	"886 :\n"							\
+	old_c "\n"							\
+	"887 :\n"							\
+	ALT_NEW_CONTENT(vendor_id_1, errata_id_1, enable_1, new_c_1)	\
 	ALT_NEW_CONTENT(vendor_id_2, errata_id_2, enable_2, new_c_2)
 
 #define _ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, CONFIG_k_1,		\
Signature · 1 line
2.35.1
#13 lore raw

PATCH v8 7/14 riscv: prevent compressed instructions in alternatives

In reply to Heiko Stuebner

Instructions are opportunistically compressed by the RISC-V assembler when possible, but in alternatives-blocks both the old and new content need to be the same size, so having the toolchain do somewhat random optimizations will cause strange side-effects like
"attempt to move .org backwards" compile-time errors.

Already a simple "and" used in alternatives assembly will cause these mismatched code sizes.

So prevent compressed instructions to be generated in alternatives- code and use option-push and -pop to only limit this to the relevant code blocks

---
 arch/riscv/include/asm/alternative-macros.h | 24 +++++++++++++++++++++
 1 file changed, 24 insertions(+)
arch/riscv/include/asm/alternative-macros.h +24 -0
index 25b4559d743b..3c12396e247c 100644
@@ -21,7 +21,11 @@
 	.popsection
 	.subsection 1
 888 :
+	.option push
+	.option norvc
+	.option norelax
 	\new_c
+	.option pop
 889 :
 	.previous
 	.org    . - (889b - 888b) + (887b - 886b)
@@ -31,7 +35,11 @@
 
 .macro __ALTERNATIVE_CFG old_c, new_c, vendor_id, errata_id, enable
 886 :
+	.option push
+	.option norvc
+	.option norelax
 	\old_c
+	.option pop
 887 :
 	ALT_NEW_CONTENT \vendor_id, \errata_id, \enable, \new_c
 .endm
@@ -42,7 +50,11 @@
 .macro __ALTERNATIVE_CFG_2 old_c, new_c_1, vendor_id_1, errata_id_1, enable_1, \
 				  new_c_2, vendor_id_2, errata_id_2, enable_2
 886 :
+	.option push
+	.option norvc
+	.option norelax
 	\old_c
+	.option pop
 887 :
 	ALT_NEW_CONTENT \vendor_id_1, \errata_id_1, \enable_1, \new_c_1
 	ALT_NEW_CONTENT \vendor_id_2, \errata_id_2, \enable_2, \new_c_2
@@ -72,7 +84,11 @@
 	".popsection\n"							\
 	".subsection 1\n"						\
 	"888 :\n"							\
+	".option push\n"						\
+	".option norvc\n"						\
+	".option norelax\n"						\
 	new_c "\n"							\
+	".option pop\n"							\
 	"889 :\n"							\
 	".previous\n"							\
 	".org	. - (887b - 886b) + (889b - 888b)\n"			\
@@ -81,7 +97,11 @@
 
 #define __ALTERNATIVE_CFG(old_c, new_c, vendor_id, errata_id, enable)	\
 	"886 :\n"							\
+	".option push\n"						\
+	".option norvc\n"						\
+	".option norelax\n"						\
 	old_c "\n"							\
+	".option pop\n"							\
 	"887 :\n"							\
 	ALT_NEW_CONTENT(vendor_id, errata_id, enable, new_c)
 
@@ -91,7 +111,11 @@
 #define __ALTERNATIVE_CFG_2(old_c, new_c_1, vendor_id_1, errata_id_1, enable_1, \
 				  new_c_2, vendor_id_2, errata_id_2, enable_2) \
 	"886 :\n"							\
+	".option push\n"						\
+	".option norvc\n"						\
+	".option norelax\n"						\
 	old_c "\n"							\
+	".option pop\n"							\
 	"887 :\n"							\
 	ALT_NEW_CONTENT(vendor_id_1, errata_id_1, enable_1, new_c_1)	\
 	ALT_NEW_CONTENT(vendor_id_2, errata_id_2, enable_2, new_c_2)
Signature · 7 lines
2.35.1


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Reviewed-by on v8 7/14
#14 lore raw

Re: [PATCH v8 07/14] riscv: prevent compressed instructions in alternatives

Quoted 6 lines · Heiko Stuebner

Instructions are opportunistically compressed by the RISC-V assembler when possible, but in alternatives-blocks both the old and new content need to be the same size, so having the toolchain do somewhat random optimizations will cause strange side-effects like
"attempt to move .org backwards" compile-time errors.

Quoted 6 lines

Already a simple "and" used in alternatives assembly will cause these mismatched code sizes.

So prevent compressed instructions to be generated in alternatives- code and use option-push and -pop to only limit this to the relevant code blocks

Yes, that makes sense:

#15 lore raw

PATCH v8 8/14 riscv: move boot alternatives to after fill_hwcap

In reply to Heiko Stuebner

Move the application of boot alternatives to after the hw-capabilities are populated. This allows to check for available extensions when determining which alternatives to apply and also makes it actually work if CONFIG_SMP is disabled for whatever reason.

---
 arch/riscv/kernel/setup.c   | 2 ++
 arch/riscv/kernel/smpboot.c | 2 --
 2 files changed, 2 insertions(+), 2 deletions(-)
arch/riscv/kernel/setup.c +2 -0
index 339ceb595b38..b4879c942b42 100644
@@ -21,6 +21,7 @@
 #include <linux/efi.h>
 #include <linux/crash_dump.h>
 
+#include <asm/alternative.h>
 #include <asm/cpu_ops.h>
 #include <asm/early_ioremap.h>
 #include <asm/pgtable.h>
@@ -295,6 +296,7 @@ void __init setup_arch(char **cmdline_p)
 #endif
 
 	riscv_fill_hwcap();
+	apply_boot_alternatives();
 }
 
 static int __init topology_init(void)
arch/riscv/kernel/smpboot.c +0 -2
index a6d13dca1403..f1e4948a4b52 100644
@@ -32,7 +32,6 @@
 #include <asm/sections.h>
 #include <asm/sbi.h>
 #include <asm/smp.h>
-#include <asm/alternative.h>
 
 #include "head.h"
 
@@ -41,7 +40,6 @@ static DECLARE_COMPLETION(cpu_running);
 void __init smp_prepare_boot_cpu(void)
 {
 	init_cpu_topology();
-	apply_boot_alternatives();
 }
 
 void __init smp_prepare_cpus(unsigned int max_cpus)
Signature · 7 lines
2.35.1


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Reviewed-by on v8 8/14
#16 lore raw

Re: [PATCH v8 08/14] riscv: move boot alternatives to after fill_hwcap

Quoted 4 lines · Heiko Stuebner

Move the application of boot alternatives to after the hw-capabilities are populated. This allows to check for available extensions when determining which alternatives to apply and also makes it actually work if CONFIG_SMP is disabled for whatever reason.

Looks good:

#17 lore raw

PATCH v8 9/14 riscv: Fix accessing pfn bits in PTEs for non-32bit variants

In reply to Heiko Stuebner

On rv32 the PFN part of PTEs is defined to use bits [xlen-1:10] while on rv64 it is defined to use bits [53:10], leaving [63:54] as reserved.

With upcoming optional extensions like svpbmt these previously reserved bits will get used so simply right-shifting the PTE to get the PFN won't be enough.

So introduce a _PAGE_PFN_MASK constant to mask the correct bits for both rv32 and rv64 before shifting.

---
 arch/riscv/include/asm/pgtable-32.h   |  8 ++++++++
 arch/riscv/include/asm/pgtable-64.h   | 14 +++++++++++---
 arch/riscv/include/asm/pgtable-bits.h |  6 ------
 arch/riscv/include/asm/pgtable.h      |  6 +++---
 4 files changed, 22 insertions(+), 12 deletions(-)
arch/riscv/include/asm/pgtable-32.h +8 -0
index 5b2e79e5bfa5..e266a4fe7f43 100644
@@ -7,6 +7,7 @@
 #define _ASM_RISCV_PGTABLE_32_H
 
 #include <asm-generic/pgtable-nopmd.h>
+#include <linux/bits.h>
 #include <linux/const.h>
 
 /* Size of region mapped by a page global directory */
@@ -16,4 +17,11 @@
 
 #define MAX_POSSIBLE_PHYSMEM_BITS 34
 
+/*
+ * rv32 PTE format:
+ * | XLEN-1  10 | 9             8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0
+ *       PFN      reserved for SW   D   A   G   U   X   W   R   V
+ */
+#define _PAGE_PFN_MASK  GENMASK(31, 10)
+
 #endif /* _ASM_RISCV_PGTABLE_32_H */
arch/riscv/include/asm/pgtable-64.h +11 -3
index bbbdd66e5e2f..9412c6157c88 100644
@@ -6,6 +6,7 @@
 #ifndef _ASM_RISCV_PGTABLE_64_H
 #define _ASM_RISCV_PGTABLE_64_H
 
+#include <linux/bits.h>
 #include <linux/const.h>
 
 extern bool pgtable_l4_enabled;
@@ -48,6 +49,13 @@ typedef struct {
 
 #define PTRS_PER_PMD    (PAGE_SIZE / sizeof(pmd_t))
 
+/*
+ * rv64 PTE format:
+ * | 63 | 62 61 | 60 54 | 53  10 | 9             8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0
+ *   N      MT     RSV    PFN      reserved for SW   D   A   G   U   X   W   R   V
+ */
+#define _PAGE_PFN_MASK  GENMASK(53, 10)
+
 static inline int pud_present(pud_t pud)
 {
 	return (pud_val(pud) & _PAGE_PRESENT);
@@ -91,12 +99,12 @@ static inline unsigned long _pud_pfn(pud_t pud)
 
 static inline pmd_t *pud_pgtable(pud_t pud)
 {
-	return (pmd_t *)pfn_to_virt(pud_val(pud) >> _PAGE_PFN_SHIFT);
+	return (pmd_t *)pfn_to_virt((pud_val(pud) & _PAGE_PFN_MASK) >> _PAGE_PFN_SHIFT);
 }
 
 static inline struct page *pud_page(pud_t pud)
 {
-	return pfn_to_page(pud_val(pud) >> _PAGE_PFN_SHIFT);
+	return pfn_to_page((pud_val(pud) & _PAGE_PFN_MASK) >> _PAGE_PFN_SHIFT);
 }
 
 #define mm_pud_folded  mm_pud_folded
@@ -117,7 +125,7 @@ static inline pmd_t pfn_pmd(unsigned long pfn, pgprot_t prot)
 
 static inline unsigned long _pmd_pfn(pmd_t pmd)
 {
-	return pmd_val(pmd) >> _PAGE_PFN_SHIFT;
+	return (pmd_val(pmd) & _PAGE_PFN_MASK) >> _PAGE_PFN_SHIFT;
 }
 
 #define mk_pmd(page, prot)    pfn_pmd(page_to_pfn(page), prot)
arch/riscv/include/asm/pgtable-bits.h +0 -6
index a6b0c89824c2..e571fa954afc 100644
@@ -6,12 +6,6 @@
 #ifndef _ASM_RISCV_PGTABLE_BITS_H
 #define _ASM_RISCV_PGTABLE_BITS_H
 
-/*
- * PTE format:
- * | XLEN-1  10 | 9             8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0
- *       PFN      reserved for SW   D   A   G   U   X   W   R   V
- */
-
 #define _PAGE_ACCESSED_OFFSET 6
 
 #define _PAGE_PRESENT   (1 << 0)
arch/riscv/include/asm/pgtable.h +3 -3
index e3549e50de95..6d31489818cd 100644
@@ -259,12 +259,12 @@ static inline unsigned long _pgd_pfn(pgd_t pgd)
 
 static inline struct page *pmd_page(pmd_t pmd)
 {
-	return pfn_to_page(pmd_val(pmd) >> _PAGE_PFN_SHIFT);
+	return pfn_to_page((pmd_val(pmd) & _PAGE_PFN_MASK) >> _PAGE_PFN_SHIFT);
 }
 
 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 {
-	return (unsigned long)pfn_to_virt(pmd_val(pmd) >> _PAGE_PFN_SHIFT);
+	return (unsigned long)pfn_to_virt((pmd_val(pmd) & _PAGE_PFN_MASK) >> _PAGE_PFN_SHIFT);
 }
 
 static inline pte_t pmd_pte(pmd_t pmd)
@@ -280,7 +280,7 @@ static inline pte_t pud_pte(pud_t pud)
 /* Yields the page frame number (PFN) of a page table entry */
 static inline unsigned long pte_pfn(pte_t pte)
 {
-	return (pte_val(pte) >> _PAGE_PFN_SHIFT);
+	return ((pte_val(pte) & _PAGE_PFN_MASK) >> _PAGE_PFN_SHIFT);
 }
 
 #define pte_page(x)     pfn_to_page(pte_pfn(x))
Signature · 1 line
2.35.1
#18 lore raw

Re: [PATCH v8 09/14] riscv: Fix accessing pfn bits in PTEs for non-32bit variants

Quoted 32 lines · Heiko Stuebner

On rv32 the PFN part of PTEs is defined to use bits [xlen-1:10] while on rv64 it is defined to use bits [53:10], leaving [63:54] as reserved.

With upcoming optional extensions like svpbmt these previously reserved bits will get used so simply right-shifting the PTE to get the PFN won't be enough.

So introduce a _PAGE_PFN_MASK constant to mask the correct bits for both rv32 and rv64 before shifting.

---
 arch/riscv/include/asm/pgtable-32.h   |  8 ++++++++
 arch/riscv/include/asm/pgtable-64.h   | 14 +++++++++++---
 arch/riscv/include/asm/pgtable-bits.h |  6 ------
 arch/riscv/include/asm/pgtable.h      |  6 +++---
 4 files changed, 22 insertions(+), 12 deletions(-)
arch/riscv/include/asm/pgtable-32.h +1 -0
index 5b2e79e5bfa5..e266a4fe7f43 100644
@@ -7,6 +7,7 @@
 #define _ASM_RISCV_PGTABLE_32_H
 
 #include <asm-generic/pgtable-nopmd.h>
+#include <linux/bits.h>
 #include <linux/const.h>
 
 /* Size of region mapped by a page global directory */

arch/riscv/include/asm/pgtable-32.h line 25

Quoted 9 lines
arch/riscv/include/asm/pgtable-32.h +6 -0
@@ -16,4 +17,11 @@
 
 #define MAX_POSSIBLE_PHYSMEM_BITS 34
 
+/*
+ * rv32 PTE format:
+ * | XLEN-1  10 | 9             8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0
+ *       PFN      reserved for SW   D   A   G   U   X   W   R   V
+ */
+#define _PAGE_PFN_MASK  GENMASK(31, 10)

I have to say I really hate this obsfucating GENMASK macro, but it does have a few other uses in the riscv code.

Quoted 12 lines
arch/riscv/include/asm/pgtable-32.h +7 -0
+/*
+ * rv64 PTE format:
+ * | 63 | 62 61 | 60 54 | 53  10 | 9             8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0
+ *   N      MT     RSV    PFN      reserved for SW   D   A   G   U   X   W   R   V
+ */
 
+#define _PAGE_PFN_MASK  GENMASK(53, 10)
+
 static inline int pud_present(pud_t pud)
 {
 	return (pud_val(pud) & _PAGE_PRESENT);

arch/riscv/include/asm/pgtable-32.h static inline unsigned long _pud_pfn(pud_t pud) · line 102

Quoted 5 lines
arch/riscv/include/asm/pgtable-32.h +1 -1
@@ -91,12 +99,12 @@ static inline unsigned long _pud_pfn(pud_t pud)
 
 static inline pmd_t *pud_pgtable(pud_t pud)
 {
-	return (pmd_t *)pfn_to_virt(pud_val(pud) >> _PAGE_PFN_SHIFT);
+	return (pmd_t *)pfn_to_virt((pud_val(pud) & _PAGE_PFN_MASK) >> _PAGE_PFN_SHIFT);

Lots of overly long lins making this pretty unreadable.

But in general the (pfn & _PAGE_PFN_MASK) >> _PAGE_PFN_SHIFT logic really should have a helper anyway.

#19 lore raw

PATCH v8 10/14 riscv: add cpufeature handling via alternatives

In reply to Heiko Stuebner

Some cpufeatures should be handled via the alternatives mechanism to not incur penalties on unsupporting variants.

So add a mechanism to handle these similar to cpu erratas.

---
 arch/riscv/include/asm/alternative.h |  3 ++
 arch/riscv/include/asm/errata_list.h |  2 +
 arch/riscv/kernel/alternative.c      |  2 +
 arch/riscv/kernel/cpufeature.c       | 57 +++++++++++++++++++++++++++-
 4 files changed, 63 insertions(+), 1 deletion(-)
arch/riscv/include/asm/alternative.h +3 -0
index f0657b1b3174..cf3b22173834 100644
@@ -39,5 +39,8 @@ struct errata_checkfunc_id {
 void sifive_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
 			      unsigned long archid, unsigned long impid,
 			      unsigned int stage);
+
+void riscv_cpufeature_patch_func(struct alt_entry *begin, struct alt_entry *end,
+				 unsigned int stage);
 #endif
 #endif
arch/riscv/include/asm/errata_list.h +2 -0
index 5f1046e82d9f..6b95bd9aee82 100644
@@ -14,6 +14,8 @@
 #define	ERRATA_SIFIVE_NUMBER 2
 #endif
 
+#define	CPUFEATURE_NUMBER 0
+
 #ifdef __ASSEMBLY__
 
 #define ALT_INSN_FAULT(x)						\
arch/riscv/kernel/alternative.c +2 -0
index 223770b3945c..e6c9de9f9ba6 100644
@@ -63,6 +63,8 @@ static void __init_or_module _apply_alternatives(struct alt_entry *begin,
 						 struct alt_entry *end,
 						 unsigned int stage)
 {
+	riscv_cpufeature_patch_func(begin, end, stage);
+
 	if (!vendor_patch_func)
 		return;
 
arch/riscv/kernel/cpufeature.c +56 -1

68 lines not on this page · raw message

Signature · 1 line
2.35.1
#20 lore raw

Re: [PATCH v8 10/14] riscv: add cpufeature handling via alternatives

Can you please fold the somewhat confusingly named vendor patching into this instead of keeping that magic standalone function pointer mess?

arch/riscv/kernel/cpufeature.c void __init riscv_fill_hwcap(void) · line 250

Quoted 2 lines
arch/riscv/kernel/cpufeature.c +2 -0
+static const struct cpufeature_info __initdata_or_module cpufeature_list[CPUFEATURE_NUMBER] = {
+};

Overly lone line.

#21 lore raw

Re: [PATCH v8 10/14] riscv: add cpufeature handling via alternatives

Hi Christoph,

Am Donnerstag, 31. März 2022, 12:01:39 CEST schrieb Christoph Hellwig:

Quoted 3 lines

Can you please fold the somewhat confusingly named vendor patching into this instead of keeping that magic standalone function pointer mess?

I'm not sure it's what you meant, but I did merge patches 10+11 (cpufeatures + svpbmt), as having them separate created build warnings / errors, when a bisected build happens (aka up to patch10 but without patch11)

The vendor-patch-func got an overhaul in patch13 "riscv: don't use global static vars to store alternative data"

Quoted 4 lines

arch/riscv/kernel/cpufeature.c void __init riscv_fill_hwcap(void) · line 250

Quoted 2 lines
arch/riscv/kernel/cpufeature.c +2 -0
+static const struct cpufeature_info __initdata_or_module cpufeature_list[CPUFEATURE_NUMBER] = {
+};

Overly lone line.

ok, I'll wrap that line

Heiko

_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv

#22 lore raw

PATCH v8 11/14 riscv: add RISC-V Svpbmt extension support

In reply to Heiko Stuebner

From: Wei Fu <[email protected]>

Svpbmt (the S should be capitalized) is the "Supervisor-mode: page-based memory types" extension that specifies attributes for cacheability, idempotency and ordering.

The relevant settings are done in special bits in PTEs:

Here is the svpbmt PTE format:

| 63 | 62-61 | 60-8 | 7 | 6 | 5 | 4 | 3 | 2 | 1 | 0
  N     MT     RSW    D   A   G   U   X   W   R   V
        ^

Of the Reserved bits [63:54] in a leaf PTE, the high bit is already allocated (as the N bit), so bits [62:61] are used as the MT (aka MemType) field. This field specifies one of three memory types that are close equivalents (or equivalent in effect) to the three main x86 and ARMv8 memory types - as shown in the following table.

RISC-V
Encoding &

MemType     RISC-V Description
----------  ------------------------------------------------
00 - PMA    Normal Cacheable, No change to implied PMA memory type
01 - NC     Non-cacheable, idempotent, weakly-ordered Main Memory
10 - IO     Non-cacheable, non-idempotent, strongly-ordered I/O memory
11 - Rsvd   Reserved for future standard use

As the extension will not be present on all implementations use alternatives to check its existence at runtime and if needed adapt the code to use memory-types.

[moved to use the alternatives mechanism]
Cc Christoph Hellwig <[email protected]>
Cc Arnd Bergmann <[email protected]>
Cc Drew Fustini <[email protected]>
Cc Wei Fu <[email protected]>
Cc Wei Wu <[email protected]>
Cc Chen-Yu Tsai <[email protected]>
Cc Maxime Ripard <[email protected]>
Cc Daniel Lustig <[email protected]>
Cc Greg Favor <[email protected]>
Cc Andrea Mondelli <[email protected]>
Cc Jonathan Behrens <[email protected]>
Cc Xinhaoqu (Freddie) <[email protected]>
Cc Bill Huffman <[email protected]>
Cc Nick Kossifidis <[email protected]>
Cc Allen Baum <[email protected]>
Cc Josh Scheid <[email protected]>
Cc Richard Trauben <[email protected]>
---
 arch/riscv/include/asm/errata_list.h  | 13 +++++++-
 arch/riscv/include/asm/hwcap.h        |  1 +
 arch/riscv/include/asm/pgtable-32.h   |  9 +++++
 arch/riscv/include/asm/pgtable-64.h   | 47 +++++++++++++++++++++++++++
 arch/riscv/include/asm/pgtable-bits.h |  4 ---
 arch/riscv/include/asm/pgtable.h      | 29 ++++++++++++++---
 arch/riscv/kernel/cpu.c               |  1 +
 arch/riscv/kernel/cpufeature.c        | 18 ++++++++++
 8 files changed, 112 insertions(+), 10 deletions(-)
arch/riscv/include/asm/errata_list.h +12 -1
index 6b95bd9aee82..a4a9b0842922 100644
@@ -14,7 +14,8 @@
 #define	ERRATA_SIFIVE_NUMBER 2
 #endif
 
-#define	CPUFEATURE_NUMBER 0
+#define	CPUFEATURE_SVPBMT 0
+#define	CPUFEATURE_NUMBER 1
 
 #ifdef __ASSEMBLY__
 
@@ -36,6 +37,16 @@ asm(ALTERNATIVE("sfence.vma %0", "sfence.vma", SIFIVE_VENDOR_ID,	\
 		ERRATA_SIFIVE_CIP_1200, CONFIG_ERRATA_SIFIVE_CIP_1200)	\
 		: : "r" (addr) : "memory")
 
+/*
+ * _val is marked as "will be overwritten", so need to set it to 0
+ * in the default case.
+ */
+#define ALT_SVPBMT_SHIFT 61
+#define ALT_SVPBMT(_val, prot)						\
+asm(ALTERNATIVE("li %0, 0\t\nnop", "li %0, %1\t\nslli %0,%0,%2", 0,	\
+		CPUFEATURE_SVPBMT, CONFIG_64BIT)			\
+		: "=r"(_val) : "I"(prot##_SVPBMT >> ALT_SVPBMT_SHIFT), "I"(ALT_SVPBMT_SHIFT))
+
 #endif /* __ASSEMBLY__ */
 
 #endif
arch/riscv/include/asm/hwcap.h +1 -0
index 691fc9c8099b..656cd626eb1a 100644
@@ -51,6 +51,7 @@ extern unsigned long elf_hwcap;
  * available logical extension id.
  */
 enum riscv_isa_ext_id {
+	RISCV_ISA_EXT_SVPBMT = RISCV_ISA_EXT_BASE,
 	RISCV_ISA_EXT_ID_MAX = RISCV_ISA_EXT_MAX,
 };
 
arch/riscv/include/asm/pgtable-32.h +9 -0
index e266a4fe7f43..59ba1fbaf784 100644
@@ -24,4 +24,13 @@
  */
 #define _PAGE_PFN_MASK  GENMASK(31, 10)
 
+#define _PAGE_NOCACHE		0
+#define _PAGE_IO		0
+#define _PAGE_MTMASK		0
+
+/* Set of bits to preserve across pte_modify() */
+#define _PAGE_CHG_MASK  (~(unsigned long)(_PAGE_PRESENT | _PAGE_READ |	\
+					  _PAGE_WRITE | _PAGE_EXEC |	\
+					  _PAGE_USER | _PAGE_GLOBAL))
+
 #endif /* _ASM_RISCV_PGTABLE_32_H */
arch/riscv/include/asm/pgtable-64.h +47 -0

59 lines not on this page · raw message

arch/riscv/include/asm/pgtable-bits.h +0 -4
index e571fa954afc..b9e13a8fe2b7 100644
@@ -29,10 +29,6 @@
 
 #define _PAGE_PFN_SHIFT 10
 
-/* Set of bits to preserve across pte_modify() */
-#define _PAGE_CHG_MASK  (~(unsigned long)(_PAGE_PRESENT | _PAGE_READ |	\
-					  _PAGE_WRITE | _PAGE_EXEC |	\
-					  _PAGE_USER | _PAGE_GLOBAL))
 /*
  * when all of R/W/X are zero, the PTE is a pointer to the next level
  * of the page table; otherwise, it is a leaf PTE.
arch/riscv/include/asm/pgtable.h +24 -5
index 6d31489818cd..009a17f94b86 100644
@@ -175,11 +175,8 @@ extern struct pt_alloc_ops pt_ops __initdata;
 
 #define PAGE_TABLE		__pgprot(_PAGE_TABLE)
 
-/*
- * The RISC-V ISA doesn't yet specify how to query or modify PMAs, so we can't
- * change the properties of memory regions.
- */
-#define _PAGE_IOREMAP _PAGE_KERNEL
+#define _PAGE_IOREMAP	((_PAGE_KERNEL & ~_PAGE_MTMASK) | _PAGE_IO)
+#define PAGE_KERNEL_IO		__pgprot(_PAGE_IOREMAP)
 
 extern pgd_t swapper_pg_dir[];
 
@@ -519,6 +516,28 @@ static inline int ptep_clear_flush_young(struct vm_area_struct *vma,
 	return ptep_test_and_clear_young(vma, address, ptep);
 }
 
+#define pgprot_noncached pgprot_noncached
+static inline pgprot_t pgprot_noncached(pgprot_t _prot)
+{
+	unsigned long prot = pgprot_val(_prot);
+
+	prot &= ~_PAGE_MTMASK;
+	prot |= _PAGE_IO;
+
+	return __pgprot(prot);
+}
+
+#define pgprot_writecombine pgprot_writecombine
+static inline pgprot_t pgprot_writecombine(pgprot_t _prot)
+{
+	unsigned long prot = pgprot_val(_prot);
+
+	prot &= ~_PAGE_MTMASK;
+	prot |= _PAGE_NOCACHE;
+
+	return __pgprot(prot);
+}
+
 /*
  * THP functions
  */
arch/riscv/kernel/cpu.c +1 -0
index fc115e307ef5..01a4799e7e1c 100644
@@ -87,6 +87,7 @@ int riscv_of_parent_hartid(struct device_node *node)
  *    extensions by an underscore.
  */
 static struct riscv_isa_ext_data isa_ext_arr[] = {
+	__RISCV_ISA_EXT_DATA("svpbmt", RISCV_ISA_EXT_SVPBMT),
 	__RISCV_ISA_EXT_DATA("", RISCV_ISA_EXT_MAX),
 };
 
arch/riscv/kernel/cpufeature.c +18 -0
index c0ffc26c3ef3..fab9c5367949 100644
@@ -8,6 +8,7 @@
 
 #include <linux/bitmap.h>
 #include <linux/ctype.h>
+#include <linux/libfdt.h>
 #include <linux/module.h>
 #include <linux/of.h>
 #include <asm/alternative.h>
@@ -195,6 +196,8 @@ void __init riscv_fill_hwcap(void)
 			if (!ext_long) {
 				this_hwcap |= isa2hwcap[(unsigned char)(*ext)];
 				set_bit(*ext - 'a', this_isa);
+			} else {
+				SET_ISA_EXT_MAP("svpbmt", RISCV_ISA_EXT_SVPBMT);
 			}
 #undef SET_ISA_EXT_MAP
 		}
@@ -246,7 +249,22 @@ struct cpufeature_info {
 	bool (*check_func)(unsigned int stage);
 };
 
+static bool __init_or_module cpufeature_svpbmt_check_func(unsigned int stage)
+{
+	bool ret = false;
+
+#if defined(CONFIG_MMU) && defined(CONFIG_64BIT)
+	return riscv_isa_extension_available(NULL, SVPBMT);
+#endif
+
+	return ret;
+}
+
 static const struct cpufeature_info __initdata_or_module cpufeature_list[CPUFEATURE_NUMBER] = {
+	{
+		.name = "svpbmt",
+		.check_func = cpufeature_svpbmt_check_func
+	},
 };
 
 static u32 __init_or_module cpufeature_probe(unsigned int stage)
Signature · 1 line
2.35.1
#24 lore raw

Re: [PATCH v8 11/14] riscv: add RISC-V Svpbmt extension support

Hi Christoph,

Am Donnerstag, 31. März 2022, 12:03:37 CEST schrieb Christoph Hellwig:

Quoted 1 line

looks like this mail somehow lost its content.

At first I assumed some issue on my end, but lore [0] agrees that this mail somehow was empty.

Heiko

[0] https://lore.kernel.org/all/[email protected]/

#25 lore raw

Re: [PATCH v8 11/14] riscv: add RISC-V Svpbmt extension support

Quoted 9 lines · Heiko Stübner

Hi Christoph,

Am Donnerstag, 31. März 2022, 12:03:37 CEST schrieb Christoph Hellwig:

Quoted 1 line

looks like this mail somehow lost its content.

At first I assumed some issue on my end, but lore [0] agrees that this mail somehow was empty.

It also is empty in my sent mail folder, and I have no idea why.

Reviewed-by on v8 11/14
#26 lore raw

Re: [PATCH v8 11/14] riscv: add RISC-V Svpbmt extension support

In reply to Heiko Stuebner

arch/riscv/include/asm/errata_list.h asm(ALTERNATIVE("sfence.vma %0", "sfence.vma", SIFIVE_VENDOR_ID, \ · line 48

Quoted 3 lines · Heiko Stuebner
arch/riscv/include/asm/errata_list.h +3 -0
+asm(ALTERNATIVE("li %0, 0\t\nnop", "li %0, %1\t\nslli %0,%0,%2", 0,	\
+		CPUFEATURE_SVPBMT, CONFIG_64BIT)			\
+		: "=r"(_val) : "I"(prot##_SVPBMT >> ALT_SVPBMT_SHIFT), "I"(ALT_SVPBMT_SHIFT))

This would be much more readable as;

asm(ALTERNATIVE("li %0, 0\t\nnop", "li %0, %1\t\nslli %0,%0,%2", 0,	\
		CPUFEATURE_SVPBMT, CONFIG_64BIT)			\
		: "=r"(_val)						\
		: "I"(prot##_SVPBMT >> ALT_SVPBMT_SHIFT),		\
		"I"(ALT_SVPBMT_SHIFT))

Otherwise this looks good to me:

#27 lore raw

PATCH v8 12/14 riscv: remove FIXMAP_PAGE_IO and fall back to its default value

In reply to Heiko Stuebner

If not defined in the arch, FIXMAP_PAGE_IO defaults to PAGE_KERNEL_IO, which we defined when adding the svpbmt implementation.

So drop the FIXMAP_PAGE_IO riscv define.

---
 arch/riscv/include/asm/fixmap.h | 2 --
 1 file changed, 2 deletions(-)
arch/riscv/include/asm/fixmap.h +0 -2
index 58a718573ad6..692e25e56d8e 100644
@@ -44,8 +44,6 @@ enum fixed_addresses {
 	__end_of_fixed_addresses
 };
 
-#define FIXMAP_PAGE_IO		PAGE_KERNEL
-
 #define __early_set_fixmap	__set_fixmap
 
 #define __late_set_fixmap	__set_fixmap
Signature · 1 line
2.35.1
Reviewed-by on v8 12/14
#28 lore raw

Re: [PATCH v8 12/14] riscv: remove FIXMAP_PAGE_IO and fall back to its default value

Quoted 4 lines · Heiko Stuebner

If not defined in the arch, FIXMAP_PAGE_IO defaults to PAGE_KERNEL_IO, which we defined when adding the svpbmt implementation.

So drop the FIXMAP_PAGE_IO riscv define.

Looks good:

#29 lore raw

PATCH v8 13/14 riscv: don't use global static vars to store alternative data

In reply to Heiko Stuebner

Right now the code uses a global struct to store vendor-ids and another global variable to store the vendor-patch-function.

There exist specific cases where we'll need to patch the kernel at an even earlier stage, where trying to write to a static variable might actually result in hangs.

Also collecting the vendor-information consists of 3 sbi-ecalls (or csr-reads) which is pretty negligible in the context of booting a kernel.

So rework the code to not rely on static variables and instead collect the vendor-information when a round of alternatives is to be applied.

---
 arch/riscv/kernel/alternative.c | 51 ++++++++++++++++-----------------
 1 file changed, 24 insertions(+), 27 deletions(-)
arch/riscv/kernel/alternative.c +24 -27
index e6c9de9f9ba6..27f722ae452b 100644
@@ -16,41 +16,35 @@
 #include <asm/sbi.h>
 #include <asm/csr.h>
 
-static struct cpu_manufacturer_info_t {
+struct cpu_manufacturer_info_t {
 	unsigned long vendor_id;
 	unsigned long arch_id;
 	unsigned long imp_id;
-} cpu_mfr_info;
+	void (*vendor_patch_func)(struct alt_entry *begin, struct alt_entry *end,
+				  unsigned long archid, unsigned long impid,
+				  unsigned int stage);
+};
 
-static void (*vendor_patch_func)(struct alt_entry *begin, struct alt_entry *end,
-				 unsigned long archid, unsigned long impid,
-				 unsigned int stage) __initdata_or_module;
-
-static inline void __init riscv_fill_cpu_mfr_info(void)
+static void __init_or_module riscv_fill_cpu_mfr_info(struct cpu_manufacturer_info_t *cpu_mfr_info)
 {
 #ifdef CONFIG_RISCV_M_MODE
-	cpu_mfr_info.vendor_id = csr_read(CSR_MVENDORID);
-	cpu_mfr_info.arch_id = csr_read(CSR_MARCHID);
-	cpu_mfr_info.imp_id = csr_read(CSR_MIMPID);
+	cpu_mfr_info->vendor_id = csr_read(CSR_MVENDORID);
+	cpu_mfr_info->arch_id = csr_read(CSR_MARCHID);
+	cpu_mfr_info->imp_id = csr_read(CSR_MIMPID);
 #else
-	cpu_mfr_info.vendor_id = sbi_get_mvendorid();
-	cpu_mfr_info.arch_id = sbi_get_marchid();
-	cpu_mfr_info.imp_id = sbi_get_mimpid();
+	cpu_mfr_info->vendor_id = sbi_get_mvendorid();
+	cpu_mfr_info->arch_id = sbi_get_marchid();
+	cpu_mfr_info->imp_id = sbi_get_mimpid();
 #endif
-}
-
-static void __init init_alternative(void)
-{
-	riscv_fill_cpu_mfr_info();
 
-	switch (cpu_mfr_info.vendor_id) {
+	switch (cpu_mfr_info->vendor_id) {
 #ifdef CONFIG_ERRATA_SIFIVE
 	case SIFIVE_VENDOR_ID:
-		vendor_patch_func = sifive_errata_patch_func;
+		cpu_mfr_info->vendor_patch_func = sifive_errata_patch_func;
 		break;
 #endif
 	default:
-		vendor_patch_func = NULL;
+		cpu_mfr_info->vendor_patch_func = NULL;
 	}
 }
 
@@ -63,14 +57,19 @@ static void __init_or_module _apply_alternatives(struct alt_entry *begin,
 						 struct alt_entry *end,
 						 unsigned int stage)
 {
+	struct cpu_manufacturer_info_t cpu_mfr_info;
+
+	riscv_fill_cpu_mfr_info(&cpu_mfr_info);
+
 	riscv_cpufeature_patch_func(begin, end, stage);
 
-	if (!vendor_patch_func)
+	if (!cpu_mfr_info.vendor_patch_func)
 		return;
 
-	vendor_patch_func(begin, end,
-			  cpu_mfr_info.arch_id, cpu_mfr_info.imp_id,
-			  stage);
+	cpu_mfr_info.vendor_patch_func(begin, end,
+				   cpu_mfr_info.arch_id,
+				   cpu_mfr_info.imp_id,
+				   stage);
 }
 
 void __init apply_boot_alternatives(void)
@@ -78,8 +77,6 @@ void __init apply_boot_alternatives(void)
 	/* If called on non-boot cpu things could go wrong */
 	WARN_ON(smp_processor_id() != 0);
 
-	init_alternative();
-
 	_apply_alternatives((struct alt_entry *)__alt_start,
 			    (struct alt_entry *)__alt_end,
 			    RISCV_ALTERNATIVES_BOOT);
Signature · 1 line
2.35.1
#30 lore raw

PATCH v8 14/14 riscv: add memory-type errata for T-Head

In reply to Heiko Stuebner

Some current cpus based on T-Head cores implement memory-types way different than described in the svpbmt spec even going so far as using PTE bits marked as reserved.

Add the T-Head vendor-id and necessary errata code to replace the affected instructions.

---
 arch/riscv/Kconfig.erratas             | 19 ++++++
 arch/riscv/errata/Makefile             |  1 +
 arch/riscv/errata/sifive/errata.c      |  7 ++-
 arch/riscv/errata/thead/Makefile       |  1 +
 arch/riscv/errata/thead/errata.c       | 82 ++++++++++++++++++++++++++
 arch/riscv/include/asm/alternative.h   |  5 ++
 arch/riscv/include/asm/errata_list.h   | 47 +++++++++++++--
 arch/riscv/include/asm/pgtable-64.h    | 18 +++++-
 arch/riscv/include/asm/pgtable.h       | 18 +++++-
 arch/riscv/include/asm/vendorid_list.h |  1 +
 arch/riscv/kernel/alternative.c        | 12 ++++
 arch/riscv/kernel/cpufeature.c         |  7 ++-
 arch/riscv/mm/init.c                   |  1 +
 13 files changed, 209 insertions(+), 10 deletions(-)
 create mode 100644 arch/riscv/errata/thead/Makefile
 create mode 100644 arch/riscv/errata/thead/errata.c
arch/riscv/Kconfig.erratas +19 -0
index 15cdeeb4151f..8221c5df8da8 100644
@@ -32,4 +32,23 @@ config ERRATA_SIFIVE_CIP_1200
 
 	  If you don't know what to do here, say "Y".
 
+config ERRATA_THEAD
+	bool "T-HEAD errata"
+	help
+	  All T-HEAD errata Kconfig depend on this Kconfig. Disabling
+	  this Kconfig will disable all T-HEAD errata. Please say "Y"
+	  here if your platform uses T-HEAD CPU cores.
+
+	  Otherwise, please say "N" here to avoid unnecessary overhead.
+
+config ERRATA_THEAD_PBMT
+	bool "Apply T-Head memory type errata"
+	depends on ERRATA_THEAD && 64BIT
+	default y
+	help
+	  This will apply the memory type errata to handle the non-standard
+	  memory type bits in page-table-entries on T-Head SoCs.
+
+	  If you don't know what to do here, say "Y".
+
 endmenu
arch/riscv/errata/Makefile +1 -0
index 0ca1c5281a2d..a1055965fbee 100644
@@ -1 +1,2 @@
 obj-$(CONFIG_ERRATA_SIFIVE) += sifive/
+obj-$(CONFIG_ERRATA_THEAD) += thead/
arch/riscv/errata/sifive/errata.c +6 -1
index 42457efcaaa3..0b14d5475573 100644
@@ -85,10 +85,15 @@ void __init_or_module sifive_errata_patch_func(struct alt_entry *begin, struct a
 				     unsigned int stage)
 {
 	struct alt_entry *alt;
-	u32 cpu_req_errata = sifive_errata_probe(archid, impid);
+	u32 cpu_req_errata;
 	u32 cpu_apply_errata = 0;
 	u32 tmp;
 
+	if (stage == RISCV_ALTERNATIVES_EARLY_BOOT)
+		return;
+
+	cpu_req_errata = sifive_errata_probe(archid, impid);
+
 	for (alt = begin; alt < end; alt++) {
 		if (alt->vendor_id != SIFIVE_VENDOR_ID)
 			continue;
arch/riscv/errata/thead/Makefile new file +1 -0
new file mode 100644
index 000000000000..2d644e19caef
@@ -0,0 +1 @@
+obj-y += errata.o
arch/riscv/errata/thead/errata.c new file +82 -0

82 lines not on this page · raw message

arch/riscv/include/asm/alternative.h +5 -0
index cf3b22173834..d1154c91ab03 100644
@@ -19,8 +19,10 @@
 
 #define RISCV_ALTERNATIVES_BOOT		0 /* alternatives applied during regular boot */
 #define RISCV_ALTERNATIVES_MODULE	1 /* alternatives applied during module-init */
+#define RISCV_ALTERNATIVES_EARLY_BOOT	2 /* alternatives applied before mmu start */
 
 void __init apply_boot_alternatives(void);
+void __init apply_early_boot_alternatives(void);
 void apply_module_alternatives(void *start, size_t length);
 
 struct alt_entry {
@@ -39,6 +41,9 @@ struct errata_checkfunc_id {
 void sifive_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
 			      unsigned long archid, unsigned long impid,
 			      unsigned int stage);
+void thead_errata_patch_func(struct alt_entry *begin, struct alt_entry *end,
+			     unsigned long archid, unsigned long impid,
+			     unsigned int stage);
 
 void riscv_cpufeature_patch_func(struct alt_entry *begin, struct alt_entry *end,
 				 unsigned int stage);
arch/riscv/include/asm/errata_list.h +43 -4

59 lines not on this page · raw message

arch/riscv/include/asm/pgtable-64.h +17 -1
index 07ba3416cb19..6d59e4695200 100644
@@ -69,6 +69,18 @@ typedef struct {
 #define _PAGE_IO_SVPBMT		(1UL << 62)
 #define _PAGE_MTMASK_SVPBMT	(_PAGE_NOCACHE_SVPBMT | _PAGE_IO_SVPBMT)
 
+/*
+ * [63:59] T-Head Memory Type definitions:
+ *
+ * 00000 - NC   Weakly-ordered, Non-cacheable, Non-bufferable, Non-shareable, Non-trustable
+ * 01110 - PMA  Weakly-ordered, Cacheable, Bufferable, Shareable, Non-trustable
+ * 10000 - IO   Strongly-ordered, Non-cacheable, Non-bufferable, Non-shareable, Non-trustable
+ */
+#define _PAGE_PMA_THEAD		((1UL << 62) | (1UL << 61) | (1UL << 60))
+#define _PAGE_NOCACHE_THEAD	0UL
+#define _PAGE_IO_THEAD		(1UL << 63)
+#define _PAGE_MTMASK_THEAD	(_PAGE_PMA_THEAD | _PAGE_IO_THEAD | (1UL << 59))
+
 static inline u64 riscv_page_mtmask(void)
 {
 	u64 val;
@@ -167,7 +179,11 @@ static inline bool mm_pud_folded(struct mm_struct *mm)
 
 static inline pmd_t pfn_pmd(unsigned long pfn, pgprot_t prot)
 {
-	return __pmd((pfn << _PAGE_PFN_SHIFT) | pgprot_val(prot));
+	unsigned long prot_val = pgprot_val(prot);
+
+	ALT_THEAD_PMA(prot_val);
+
+	return __pmd((pfn << _PAGE_PFN_SHIFT) | prot_val);
 }
 
 static inline unsigned long _pmd_pfn(pmd_t pmd)
arch/riscv/include/asm/pgtable.h +15 -3
index 009a17f94b86..e2658a25f06d 100644
@@ -246,7 +246,11 @@ static inline void pmd_clear(pmd_t *pmdp)
 
 static inline pgd_t pfn_pgd(unsigned long pfn, pgprot_t prot)
 {
-	return __pgd((pfn << _PAGE_PFN_SHIFT) | pgprot_val(prot));
+	unsigned long prot_val = pgprot_val(prot);
+
+	ALT_THEAD_PMA(prot_val);
+
+	return __pgd((pfn << _PAGE_PFN_SHIFT) | prot_val);
 }
 
 static inline unsigned long _pgd_pfn(pgd_t pgd)
@@ -285,7 +289,11 @@ static inline unsigned long pte_pfn(pte_t pte)
 /* Constructs a page table entry */
 static inline pte_t pfn_pte(unsigned long pfn, pgprot_t prot)
 {
-	return __pte((pfn << _PAGE_PFN_SHIFT) | pgprot_val(prot));
+	unsigned long prot_val = pgprot_val(prot);
+
+	ALT_THEAD_PMA(prot_val);
+
+	return __pte((pfn << _PAGE_PFN_SHIFT) | prot_val);
 }
 
 #define mk_pte(page, prot)       pfn_pte(page_to_pfn(page), prot)
@@ -394,7 +402,11 @@ static inline int pmd_protnone(pmd_t pmd)
 /* Modify page protection bits */
 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 {
-	return __pte((pte_val(pte) & _PAGE_CHG_MASK) | pgprot_val(newprot));
+	unsigned long newprot_val = pgprot_val(newprot);
+
+	ALT_THEAD_PMA(newprot_val);
+
+	return __pte((pte_val(pte) & _PAGE_CHG_MASK) | newprot_val);
 }
 
 #define pgd_ERROR(e) \
arch/riscv/include/asm/vendorid_list.h +1 -0
index 9d934215b3c8..cb89af3f0704 100644
@@ -6,5 +6,6 @@
 #define ASM_VENDOR_LIST_H
 
 #define SIFIVE_VENDOR_ID	0x489
+#define THEAD_VENDOR_ID		0x5b7
 
 #endif
arch/riscv/kernel/alternative.c +12 -0
index 27f722ae452b..e7980bf17e8b 100644
@@ -42,6 +42,11 @@ static void __init_or_module riscv_fill_cpu_mfr_info(struct cpu_manufacturer_inf
 	case SIFIVE_VENDOR_ID:
 		cpu_mfr_info->vendor_patch_func = sifive_errata_patch_func;
 		break;
+#endif
+#ifdef CONFIG_ERRATA_THEAD
+	case THEAD_VENDOR_ID:
+		cpu_mfr_info->vendor_patch_func = thead_errata_patch_func;
+		break;
 #endif
 	default:
 		cpu_mfr_info->vendor_patch_func = NULL;
@@ -82,6 +87,13 @@ void __init apply_boot_alternatives(void)
 			    RISCV_ALTERNATIVES_BOOT);
 }
 
+void __init apply_early_boot_alternatives(void)
+{
+	_apply_alternatives((struct alt_entry *)__alt_start,
+			    (struct alt_entry *)__alt_end,
+			    RISCV_ALTERNATIVES_EARLY_BOOT);
+}
+
 #ifdef CONFIG_MODULES
 void apply_module_alternatives(void *start, size_t length)
 {
arch/riscv/kernel/cpufeature.c +6 -1
index fab9c5367949..564c1917c8aa 100644
@@ -254,7 +254,12 @@ static bool __init_or_module cpufeature_svpbmt_check_func(unsigned int stage)
 	bool ret = false;
 
 #if defined(CONFIG_MMU) && defined(CONFIG_64BIT)
-	return riscv_isa_extension_available(NULL, SVPBMT);
+	switch (stage) {
+	case RISCV_ALTERNATIVES_EARLY_BOOT:
+		return false;
+	default:
+		return riscv_isa_extension_available(NULL, SVPBMT);
+	}
 #endif
 
 	return ret;
arch/riscv/mm/init.c +1 -0
index 0d588032d6e6..245d54a3616c 100644
@@ -821,6 +821,7 @@ asmlinkage void __init setup_vm(uintptr_t dtb_pa)
 	BUG_ON((kernel_map.virt_addr + kernel_map.size) > ADDRESS_SPACE_END - SZ_4K);
 #endif
 
+	apply_early_boot_alternatives();
 	pt_ops_set_early();
 
 	/* Setup early PGD for fixmap */
Signature · 1 line
2.35.1
maintainer
#31 lore raw
In reply to Heiko Stuebner
Quoted 40 lines · On Wed, 23 Mar 2022 17:06:56 PDT (-0700), [email protected] wrote:

Svpbmt is an extension defining "Supervisor-mode: page-based memory types" for things like non-cacheable pages or I/O memory pages.

So this is my 2nd try at implementing Svpbmt (and the diverging D1 memory types) using the alternatives framework.

This includes a number of changes to the alternatives mechanism itself. The biggest one being the move to a more central location, as I expect in the future, nearly every chip needing some sort of patching, be it either for erratas or for optional features (svpbmt or others).

Detection of the svpbmt functionality is done via Atish's isa extension handling series [0] and thus does not need any dt-parsing of its own anymore.

The series also introduces support for the memory types of the D1 which are implemented differently to svpbmt. But when patching anyway it's pretty clean to add the D1 variant via ALTERNATIVE_2 to the same location.

The only slightly bigger difference is that the "normal" type is not 0 as with svpbmt, so kernel patches for this PMA type need to be applied even before the MMU is brought up, so the series introduces a separate stage for that.

In theory this series is 3 parts:
- sbi cache-flush / null-ptr
- alternatives improvements
- svpbmt+d1

So expecially patches from the first 2 areas could be applied when deemed ready, I just thought to keep it together to show-case where the end-goal is and not requiring jumping between different series.

I picked the recipient list from the previous versions, hopefully I didn't forget anybody.

Quoted 9 lines

changes in v8:
- rebase onto 5.17-final + isa extension series We're halfway through the merge-window, so this series should be merge after that
- adapt to fix limiting alternatives to non-xip-kernels
- add .norelax option for alternatives
- fix unused cpu_apply_errata in thead errata
- don't use static globals to store cpu-manufacturer data as it makes machines hang if done too early

I'm still seeing a very similar failure mode. There were a few merge conflicts when I tried to put this on top of my merged first PR, I've put what I'm testing at riscv-d1. Not sure if I'm missing something.

Quoted 72 lines

changes in v7:
- fix typo in patch1 (Atish)
- moved to Atish's isa-extension framework
- and therefore move regular boot-alternatives directly behind fill_hwcaps
- change T-Head errata Kconfig text (Atish)

changes in v6:
- rebase onto 5.17-rc1
- handle sbi null-ptr differently
- improve commit messages
- use riscv,mmu as property name

changes in v5:
- move to use alternatives for runtime-patching
- add D1 variant

[0] https://lore.kernel.org/r/[email protected]

Heiko Stuebner (13):
riscv: prevent null-pointer dereference with sbi_remote_fence_i
riscv: integrate alternatives better into the main architecture
riscv: allow different stages with alternatives
riscv: implement module alternatives
riscv: implement ALTERNATIVE_2 macro
riscv: extend concatenated alternatives-lines to the same length
riscv: prevent compressed instructions in alternatives
riscv: move boot alternatives to after fill_hwcap
riscv: Fix accessing pfn bits in PTEs for non-32bit variants
riscv: add cpufeature handling via alternatives
riscv: remove FIXMAP_PAGE_IO and fall back to its default value
riscv: don't use global static vars to store alternative data
riscv: add memory-type errata for T-Head

Wei Fu (1):
riscv: add RISC-V Svpbmt extension support

 arch/riscv/Kconfig.erratas                  |  31 +++--
 arch/riscv/Kconfig.socs                     |   1 -
 arch/riscv/Makefile                         |   2 +-
 arch/riscv/errata/Makefile                  |   2 +-
 arch/riscv/errata/alternative.c             |  75 ------------
 arch/riscv/errata/sifive/errata.c           |  17 ++-
 arch/riscv/errata/thead/Makefile            |   1 +
 arch/riscv/errata/thead/errata.c            |  82 +++++++++++++
 arch/riscv/include/asm/alternative-macros.h | 121 ++++++++++++++------
 arch/riscv/include/asm/alternative.h        |  16 ++-
 arch/riscv/include/asm/errata_list.h        |  52 +++++++++
 arch/riscv/include/asm/fixmap.h             |   2 -
 arch/riscv/include/asm/hwcap.h              |   1 +
 arch/riscv/include/asm/pgtable-32.h         |  17 +++
 arch/riscv/include/asm/pgtable-64.h         |  79 ++++++++++++-
 arch/riscv/include/asm/pgtable-bits.h       |  10 --
 arch/riscv/include/asm/pgtable.h            |  53 +++++++--
 arch/riscv/include/asm/vendorid_list.h      |   1 +
 arch/riscv/kernel/Makefile                  |   1 +
 arch/riscv/kernel/alternative.c             | 104 +++++++++++++++++
 arch/riscv/kernel/cpu.c                     |   1 +
 arch/riscv/kernel/cpufeature.c              |  80 ++++++++++++-
 arch/riscv/kernel/module.c                  |  29 +++++
 arch/riscv/kernel/sbi.c                     |  10 +-
 arch/riscv/kernel/setup.c                   |   2 +
 arch/riscv/kernel/smpboot.c                 |   4 -
 arch/riscv/kernel/traps.c                   |   2 +-
 arch/riscv/mm/init.c                        |   1 +
 28 files changed, 629 insertions(+), 168 deletions(-)
 delete mode 100644 arch/riscv/errata/alternative.c
 create mode 100644 arch/riscv/errata/thead/Makefile
 create mode 100644 arch/riscv/errata/thead/errata.c
 create mode 100644 arch/riscv/kernel/alternative.c