Skip to contentSkip to conversation

RISC-V: Ensure Zicbom has a valid block size

PATCH Superseded Andrew Jones 3 participants 13 messages Oct 21, 2022 – Oct 24, 2022 RISC-V

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

cover RISC-V: Ensure Zicbom has a valid block size

When a DT puts zicbom in the isa string, but does not provide a block size, ALT_CMO_OP() will attempt to do cache operations on address zero since the start address will be ANDed with zero. We can't simply BUG() in riscv_init_cbom_blocksize() when we fail to find a block size because the failure will happen before logging works, leaving users to scratch their heads as to why the boot hung. Instead, ensure Zicbom is disabled and output an error which will hopefully alert people that the DT needs to be fixed. While at it, add a check that the block size is a power-of-2 too.

The first patch of the series is a cleanup of code that crossed the path of this work. The second patch prepares for isa ext. checking and the third finally does what this cover letter says.

Thanks,
drew

Andrew Jones (3):
RISC-V: Improve use of isa2hwcap[]
RISC-V: Introduce riscv_isa_extension_check
RISC-V: Ensure Zicbom has a valid block size

 arch/riscv/kernel/cpufeature.c | 45 ++++++++++++++++++++++++++--------
 1 file changed, 35 insertions(+), 10 deletions(-)
Signature · 7 lines
2.37.3


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

PATCH 1/3 RISC-V: Improve use of isa2hwcap[]

Improve isa2hwcap[] by removing it from static storage, as riscv_fill_hwcap() is only called once, and by reducing its size from 256 bytes to 26. The latter improvement is possible because isa2hwcap[] will never be indexed with capital letters and we can precompute the offsets from 'a'.

No functional change intended.

---
 arch/riscv/kernel/cpufeature.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)
arch/riscv/kernel/cpufeature.c +11 -9
index 694267d1fe81..4677320d7e31 100644
@@ -74,15 +74,15 @@ void __init riscv_fill_hwcap(void)
 	const char *isa;
 	char print_str[NUM_ALPHA_EXTS + 1];
 	int i, j, rc;
-	static unsigned long isa2hwcap[256] = {0};
+	unsigned long isa2hwcap[26] = {0};
 	unsigned long hartid;
 
-	isa2hwcap['i'] = isa2hwcap['I'] = COMPAT_HWCAP_ISA_I;
-	isa2hwcap['m'] = isa2hwcap['M'] = COMPAT_HWCAP_ISA_M;
-	isa2hwcap['a'] = isa2hwcap['A'] = COMPAT_HWCAP_ISA_A;
-	isa2hwcap['f'] = isa2hwcap['F'] = COMPAT_HWCAP_ISA_F;
-	isa2hwcap['d'] = isa2hwcap['D'] = COMPAT_HWCAP_ISA_D;
-	isa2hwcap['c'] = isa2hwcap['C'] = COMPAT_HWCAP_ISA_C;
+	isa2hwcap['i' - 'a'] = COMPAT_HWCAP_ISA_I;
+	isa2hwcap['m' - 'a'] = COMPAT_HWCAP_ISA_M;
+	isa2hwcap['a' - 'a'] = COMPAT_HWCAP_ISA_A;
+	isa2hwcap['f' - 'a'] = COMPAT_HWCAP_ISA_F;
+	isa2hwcap['d' - 'a'] = COMPAT_HWCAP_ISA_D;
+	isa2hwcap['c' - 'a'] = COMPAT_HWCAP_ISA_C;
 
 	elf_hwcap = 0;
 
@@ -196,8 +196,10 @@ void __init riscv_fill_hwcap(void)
 			if (unlikely(ext_err))
 				continue;
 			if (!ext_long) {
-				this_hwcap |= isa2hwcap[(unsigned char)(*ext)];
-				set_bit(*ext - 'a', this_isa);
+				int nr = *ext - 'a';
+
+				this_hwcap |= isa2hwcap[nr];
+				set_bit(nr, this_isa);
 			} else {
 				SET_ISA_EXT_MAP("sscofpmf", RISCV_ISA_EXT_SSCOFPMF);
 				SET_ISA_EXT_MAP("svpbmt", RISCV_ISA_EXT_SVPBMT);
Signature · 7 lines
2.37.3


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

Re: [PATCH 1/3] RISC-V: Improve use of isa2hwcap[]

Quoted 5 lines · Andrew Jones

Improve isa2hwcap[] by removing it from static storage, as riscv_fill_hwcap() is only called once, and by reducing its size from 256 bytes to 26. The latter improvement is possible because isa2hwcap[] will never be indexed with capital letters and we can precompute the offsets from 'a'.

Hey Drew, couple questions for you - mostly due to naivety I think..

How do we know that isa2hwcap will never interact with capital letters? It pulls the isa string from dt and the no-capitals enforcement comes from there since one with capitals is invalid? I didn't dig particularly deeply into the code, but is there a risk that we regress some user that has a dt with capitals in the isa string? Or is that a "your dt was wrong and you're out-of-tree so that's your problem" situation?

Secondly, in the UAPI header, the COMPAT_HWCAP_ISA_FOO defines are computed as I - A rather than i - a. Should those be changed too for the sake of consistently using the lowercase everywhere, or do you think that doesn't really matter?

Thanks,
Conor.

Quoted 56 lines

No functional change intended.

---
 arch/riscv/kernel/cpufeature.c | 20 +++++++++++---------
 1 file changed, 11 insertions(+), 9 deletions(-)
arch/riscv/kernel/cpufeature.c +11 -9
index 694267d1fe81..4677320d7e31 100644
@@ -74,15 +74,15 @@ void __init riscv_fill_hwcap(void)
 	const char *isa;
 	char print_str[NUM_ALPHA_EXTS + 1];
 	int i, j, rc;
-	static unsigned long isa2hwcap[256] = {0};
+	unsigned long isa2hwcap[26] = {0};
 	unsigned long hartid;
 
-	isa2hwcap['i'] = isa2hwcap['I'] = COMPAT_HWCAP_ISA_I;
-	isa2hwcap['m'] = isa2hwcap['M'] = COMPAT_HWCAP_ISA_M;
-	isa2hwcap['a'] = isa2hwcap['A'] = COMPAT_HWCAP_ISA_A;
-	isa2hwcap['f'] = isa2hwcap['F'] = COMPAT_HWCAP_ISA_F;
-	isa2hwcap['d'] = isa2hwcap['D'] = COMPAT_HWCAP_ISA_D;
-	isa2hwcap['c'] = isa2hwcap['C'] = COMPAT_HWCAP_ISA_C;
+	isa2hwcap['i' - 'a'] = COMPAT_HWCAP_ISA_I;
+	isa2hwcap['m' - 'a'] = COMPAT_HWCAP_ISA_M;
+	isa2hwcap['a' - 'a'] = COMPAT_HWCAP_ISA_A;
+	isa2hwcap['f' - 'a'] = COMPAT_HWCAP_ISA_F;
+	isa2hwcap['d' - 'a'] = COMPAT_HWCAP_ISA_D;
+	isa2hwcap['c' - 'a'] = COMPAT_HWCAP_ISA_C;
 
 	elf_hwcap = 0;
 
@@ -196,8 +196,10 @@ void __init riscv_fill_hwcap(void)
 			if (unlikely(ext_err))
 				continue;
 			if (!ext_long) {
-				this_hwcap |= isa2hwcap[(unsigned char)(*ext)];
-				set_bit(*ext - 'a', this_isa);
+				int nr = *ext - 'a';
+
+				this_hwcap |= isa2hwcap[nr];
+				set_bit(nr, this_isa);
 			} else {
 				SET_ISA_EXT_MAP("sscofpmf", RISCV_ISA_EXT_SSCOFPMF);
 				SET_ISA_EXT_MAP("svpbmt", RISCV_ISA_EXT_SVPBMT);
Signature · 7 lines
2.37.3


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

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

#3 lore raw

Re: [PATCH 1/3] RISC-V: Improve use of isa2hwcap[]

Quoted 7 lines · Conor Dooley
Quoted 5 lines · Andrew Jones

Improve isa2hwcap[] by removing it from static storage, as riscv_fill_hwcap() is only called once, and by reducing its size from 256 bytes to 26. The latter improvement is possible because isa2hwcap[] will never be indexed with capital letters and we can precompute the offsets from 'a'.

Quoted 8 lines

Hey Drew, couple questions for you - mostly due to naivety I think..

How do we know that isa2hwcap will never interact with capital letters? It pulls the isa string from dt and the no-capitals enforcement comes from there since one with capitals is invalid? I didn't dig particularly deeply into the code, but is there a risk that we regress some user that has a dt with capitals in the isa string? Or is that a "your dt was wrong and you're out-of-tree so that's your problem" situation?

Our ISA string parser here in riscv_fill_hwcap() ignores non-capital letters, see arch/riscv/kernel/cpufeature.c lines 141, 165 and 196. (Maybe we should be noisier about that in case there are DTs out there with capitals which aren't realizing they're being ignored.)

Quoted 5 lines

Secondly, in the UAPI header, the COMPAT_HWCAP_ISA_FOO defines are computed as I - A rather than i - a. Should those be changed too for the sake of consistently using the lowercase everywhere, or do you think that doesn't really matter?

I think I'd leave it since it doesn't change the math and I'm reluctant to churn UAPI.

Thanks,
drew

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

Reviewed-by
#4 lore raw

Re: [PATCH 1/3] RISC-V: Improve use of isa2hwcap[]

Quoted 17 lines · Andrew Jones
Quoted 7 lines · Conor Dooley
Quoted 5 lines · Andrew Jones

Improve isa2hwcap[] by removing it from static storage, as riscv_fill_hwcap() is only called once, and by reducing its size from 256 bytes to 26. The latter improvement is possible because isa2hwcap[] will never be indexed with capital letters and we can precompute the offsets from 'a'.

Quoted 8 lines

Hey Drew, couple questions for you - mostly due to naivety I think..

How do we know that isa2hwcap will never interact with capital letters? It pulls the isa string from dt and the no-capitals enforcement comes from there since one with capitals is invalid? I didn't dig particularly deeply into the code, but is there a risk that we regress some user that has a dt with capitals in the isa string? Or is that a "your dt was wrong and you're out-of-tree so that's your problem" situation?

Quoted 2 lines

Our ISA string parser here in riscv_fill_hwcap() ignores non-capital letters, see arch/riscv/kernel/cpufeature.c lines 141, 165 and 196.

Ah, I missed the one on L165, that's a bit silly... Apologies!

Quoted 2 lines

(Maybe we should be noisier about that in case there are DTs out there with capitals which aren't realizing they're being ignored.)

Possibly, but it's not this patch that has introduced the behaviour. Checking very briefly though, looks like it was 2a31c54be097 ("RISC-V:
Minimal parser for "riscv, isa" strings") that introduced the checks and that was not really all that long ago..

Quoted 9 lines
Quoted 6 lines

Secondly, in the UAPI header, the COMPAT_HWCAP_ISA_FOO defines are computed as I - A rather than i - a. Should those be changed too for the sake of consistently using the lowercase everywhere, or do you think that doesn't really matter?

I think I'd leave it since it doesn't change the math and I'm reluctant to churn UAPI.

Yeah, figured you'd be of that opinion. Anyways, only "issue" I had was my own confusion so

Thanks...

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

#5 lore raw

PATCH 2/3 RISC-V: Introduce riscv_isa_extension_check

In reply to Andrew Jones

Currently any isa extension found in the isa string is set in the isa bitmap. An isa extension set in the bitmap indicates that the extension is present and may be used (a.k.a is enabled). However, when an extension cannot be used due to missing dependencies or errata it should not be added to the bitmap. Introduce a function where additional checks may be placed in order to determine if an extension should be enabled or not.

Note, the checks may simply indicate an issue with the DT, but, since extensions may be used in early boot, it's not always possible to simply produce an error at the point the issue is determined. It's best to keep the extension disabled and produce an error.

No functional change intended, as the function is only introduced and always returns true. A later patch will provide checks for an isa extension.

---
 arch/riscv/kernel/cpufeature.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)
arch/riscv/kernel/cpufeature.c +11 -3
index 4677320d7e31..220be7222129 100644
@@ -68,6 +68,11 @@ bool __riscv_isa_extension_available(const unsigned long *isa_bitmap, int bit)
 }
 EXPORT_SYMBOL_GPL(__riscv_isa_extension_available);
 
+static bool riscv_isa_extension_check(int id)
+{
+	return true;
+}
+
 void __init riscv_fill_hwcap(void)
 {
 	struct device_node *node;
@@ -189,7 +194,8 @@ void __init riscv_fill_hwcap(void)
 #define SET_ISA_EXT_MAP(name, bit)						\
 			do {							\
 				if ((ext_end - ext == sizeof(name) - 1) &&	\
-				     !memcmp(ext, name, sizeof(name) - 1))	\
+				     !memcmp(ext, name, sizeof(name) - 1) &&	\
+				     riscv_isa_extension_check(bit))		\
 					set_bit(bit, this_isa);			\
 			} while (false)						\
 
@@ -198,8 +204,10 @@ void __init riscv_fill_hwcap(void)
 			if (!ext_long) {
 				int nr = *ext - 'a';
 
-				this_hwcap |= isa2hwcap[nr];
-				set_bit(nr, this_isa);
+				if (riscv_isa_extension_check(nr)) {
+					this_hwcap |= isa2hwcap[nr];
+					set_bit(nr, this_isa);
+				}
 			} else {
 				SET_ISA_EXT_MAP("sscofpmf", RISCV_ISA_EXT_SSCOFPMF);
 				SET_ISA_EXT_MAP("svpbmt", RISCV_ISA_EXT_SVPBMT);
Signature · 7 lines
2.37.3


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
Reviewed-by on v1 2/3
#6 lore raw

Re: [PATCH 2/3] RISC-V: Introduce riscv_isa_extension_check

Quoted 8 lines · Andrew Jones

Currently any isa extension found in the isa string is set in the isa bitmap. An isa extension set in the bitmap indicates that the extension is present and may be used (a.k.a is enabled). However, when an extension cannot be used due to missing dependencies or errata it should not be added to the bitmap. Introduce a function where additional checks may be placed in order to determine if an extension should be enabled or not.

Quoted 4 lines

Note, the checks may simply indicate an issue with the DT, but, since extensions may be used in early boot, it's not always possible to simply produce an error at the point the issue is determined. It's best to keep the extension disabled and produce an error.

Aye, agreed. Could be pretty nasty chasing down something like this...

Quoted 57 lines

No functional change intended, as the function is only introduced and always returns true. A later patch will provide checks for an isa extension.

---
 arch/riscv/kernel/cpufeature.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)
arch/riscv/kernel/cpufeature.c +11 -3
index 4677320d7e31..220be7222129 100644
@@ -68,6 +68,11 @@ bool __riscv_isa_extension_available(const unsigned long *isa_bitmap, int bit)
 }
 EXPORT_SYMBOL_GPL(__riscv_isa_extension_available);
 
+static bool riscv_isa_extension_check(int id)
+{
+	return true;
+}
+
 void __init riscv_fill_hwcap(void)
 {
 	struct device_node *node;
@@ -189,7 +194,8 @@ void __init riscv_fill_hwcap(void)
 #define SET_ISA_EXT_MAP(name, bit)						\
 			do {							\
 				if ((ext_end - ext == sizeof(name) - 1) &&	\
-				     !memcmp(ext, name, sizeof(name) - 1))	\
+				     !memcmp(ext, name, sizeof(name) - 1) &&	\
+				     riscv_isa_extension_check(bit))		\
 					set_bit(bit, this_isa);			\
 			} while (false)						\
 
@@ -198,8 +204,10 @@ void __init riscv_fill_hwcap(void)
 			if (!ext_long) {
 				int nr = *ext - 'a';
 
-				this_hwcap |= isa2hwcap[nr];
-				set_bit(nr, this_isa);
+				if (riscv_isa_extension_check(nr)) {
+					this_hwcap |= isa2hwcap[nr];
+					set_bit(nr, this_isa);
+				}
 			} else {
 				SET_ISA_EXT_MAP("sscofpmf", RISCV_ISA_EXT_SSCOFPMF);
 				SET_ISA_EXT_MAP("svpbmt", RISCV_ISA_EXT_SVPBMT);
Signature · 7 lines
2.37.3


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

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

#7 lore raw

PATCH 3/3 RISC-V: Ensure Zicbom has a valid block size

In reply to Andrew Jones

When a DT puts zicbom in the isa string, but does not provide a block size, ALT_CMO_OP() will attempt to do cache operations on address zero since the start address will be ANDed with zero. We can't simply BUG() in riscv_init_cbom_blocksize() when we fail to find a block size because the failure will happen before logging works, leaving users to scratch their heads as to why the boot hung. Instead, ensure Zicbom is disabled and output an error which will hopefully alert people that the DT needs to be fixed. While at it, add a check that the block size is a power-of-2 too.

---
 arch/riscv/kernel/cpufeature.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
arch/riscv/kernel/cpufeature.c +15 -0
index 220be7222129..a4430a77df53 100644
@@ -9,6 +9,7 @@
 #include <linux/bitmap.h>
 #include <linux/ctype.h>
 #include <linux/libfdt.h>
+#include <linux/log2.h>
 #include <linux/module.h>
 #include <linux/of.h>
 #include <asm/alternative.h>
@@ -70,6 +71,20 @@ EXPORT_SYMBOL_GPL(__riscv_isa_extension_available);
 
 static bool riscv_isa_extension_check(int id)
 {
+	switch (id) {
+	case RISCV_ISA_EXT_ZICBOM:
+		if (!riscv_cbom_block_size) {
+			if (IS_ENABLED(CONFIG_RISCV_ISA_ZICBOM))
+				pr_err("cbom-block-size not found, cannot use Zicbom\n");
+			return false;
+		} else if (!is_power_of_2(riscv_cbom_block_size)) {
+			if (IS_ENABLED(CONFIG_RISCV_ISA_ZICBOM))
+				pr_err("cbom-block-size is not a power-of-2, cannot use Zicbom\n");
+			return false;
+		}
+		return true;
+	}
+
 	return true;
 }
 
Signature · 7 lines
2.37.3


_______________________________________________
linux-riscv mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-riscv
#8 lore raw
Quoted 28 lines · Andrew Jones

When a DT puts zicbom in the isa string, but does not provide a block size, ALT_CMO_OP() will attempt to do cache operations on address zero since the start address will be ANDed with zero. We can't simply BUG() in riscv_init_cbom_blocksize() when we fail to find a block size because the failure will happen before logging works, leaving users to scratch their heads as to why the boot hung. Instead, ensure Zicbom is disabled and output an error which will hopefully alert people that the DT needs to be fixed. While at it, add a check that the block size is a power-of-2 too.

---
 arch/riscv/kernel/cpufeature.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
arch/riscv/kernel/cpufeature.c +1 -0
index 220be7222129..a4430a77df53 100644
@@ -9,6 +9,7 @@
 #include <linux/bitmap.h>
 #include <linux/ctype.h>
 #include <linux/libfdt.h>
+#include <linux/log2.h>
 #include <linux/module.h>
 #include <linux/of.h>
 #include <asm/alternative.h>

arch/riscv/kernel/cpufeature.c EXPORT_SYMBOL_GPL(__riscv_isa_extension_available); · line 77

Quoted 7 lines
arch/riscv/kernel/cpufeature.c +4 -0
@@ -70,6 +71,20 @@ EXPORT_SYMBOL_GPL(__riscv_isa_extension_available);
 
 static bool riscv_isa_extension_check(int id)
 {
+	switch (id) {
+	case RISCV_ISA_EXT_ZICBOM:
+		if (!riscv_cbom_block_size) {
+			if (IS_ENABLED(CONFIG_RISCV_ISA_ZICBOM))

Maybe I've missed something... Why not check if !IS_ENABLED() & return false immediately rather than doing it inside the attribute checks? I'll be compiled out of zicbom is enabled, so not like the non-error patch will be penalised with an extra check.

Quoted 21 lines
arch/riscv/kernel/cpufeature.c +10 -0
+				pr_err("cbom-block-size not found, cannot use Zicbom\n");
+			return false;
+		} else if (!is_power_of_2(riscv_cbom_block_size)) {
+			if (IS_ENABLED(CONFIG_RISCV_ISA_ZICBOM))
+				pr_err("cbom-block-size is not a power-of-2, cannot use Zicbom\n");
+			return false;
+		}
+		return true;
+	}
+
 	return true;
 }
 
Signature · 7 lines
2.37.3


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

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

#9 lore raw
Quoted 30 lines · Conor Dooley
Quoted 29 lines · Andrew Jones

When a DT puts zicbom in the isa string, but does not provide a block size, ALT_CMO_OP() will attempt to do cache operations on address zero since the start address will be ANDed with zero. We can't simply BUG() in riscv_init_cbom_blocksize() when we fail to find a block size because the failure will happen before logging works, leaving users to scratch their heads as to why the boot hung. Instead, ensure Zicbom is disabled and output an error which will hopefully alert people that the DT needs to be fixed. While at it, add a check that the block size is a power-of-2 too.

---
 arch/riscv/kernel/cpufeature.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
arch/riscv/kernel/cpufeature.c +1 -0
index 220be7222129..a4430a77df53 100644
@@ -9,6 +9,7 @@
 #include <linux/bitmap.h>
 #include <linux/ctype.h>
 #include <linux/libfdt.h>
+#include <linux/log2.h>
 #include <linux/module.h>
 #include <linux/of.h>
 #include <asm/alternative.h>
@@ -70,6 +71,20 @@ EXPORT_SYMBOL_GPL(__riscv_isa_extension_available);
 
Quoted 11 lines

arch/riscv/kernel/cpufeature.c EXPORT_SYMBOL_GPL(__riscv_isa_extension_available); · line 77

Quoted 6 lines
arch/riscv/kernel/cpufeature.c +4 -0
 static bool riscv_isa_extension_check(int id)
 {
+	switch (id) {
+	case RISCV_ISA_EXT_ZICBOM:
+		if (!riscv_cbom_block_size) {
+			if (IS_ENABLED(CONFIG_RISCV_ISA_ZICBOM))

Maybe I've missed something... Why not check if !IS_ENABLED() & return false immediately rather than doing it inside the attribute checks? I'll be compiled out of zicbom is enabled, so not like the non-error patch will be penalised with an extra check.

Ths IS_ENABLED checks are only here to decide if we want to complain or not, not to decide if we want to add the extension to the isa bitmap or not. As we've decided for the KVM compilation fix for Zicbom patch we don't mind the kernel knowing the extension 'foo' is present, even when it's compiled without CONFIG_RISCV_ISA_FOO. Indeed, for Zicbom, KVM can still offer the extension to guests, whether the host wants to use it or not. So, for this code, we can't factor out the IS_ENABLED checks since they're tied to the pr_err()s they're guarding, which are tied to the sanity checks they're inside.

Thinking about it some more, though, maybe we should unconditionally complain, but with a different message, one that doesn't imply the kernel wanted to use the extension. Something like

  if (!riscv_cbom_block_size) {
      pr_err("Zicbom detected in ISA string, but no cbom-block-size found\n");
      return false;
  } else if (!is_power_of_2(riscv_cbom_block_size)) {
      pr_err("riscv_cbom_block_size present, but it is not a power-of-2\n");
      return false;
  }

I'm still not sure if those should be errors or warnings when CONFIG_RISCV_ISA_ZICBOM=no, though.

Thanks,
drew

Quoted 14 lines
Quoted 14 lines
arch/riscv/kernel/cpufeature.c +10 -0
+				pr_err("cbom-block-size not found, cannot use Zicbom\n");
+			return false;
+		} else if (!is_power_of_2(riscv_cbom_block_size)) {
+			if (IS_ENABLED(CONFIG_RISCV_ISA_ZICBOM))
+				pr_err("cbom-block-size is not a power-of-2, cannot use Zicbom\n");
+			return false;
+		}
+		return true;
+	}
+
 	return true;
 }
 
Quoted 8 lines
Quoted 8 lines
Signature · 7 lines
2.37.3


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

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

#10 lore raw
Quoted 43 lines · Andrew Jones
Quoted 30 lines · Conor Dooley
Quoted 29 lines · Andrew Jones

When a DT puts zicbom in the isa string, but does not provide a block size, ALT_CMO_OP() will attempt to do cache operations on address zero since the start address will be ANDed with zero. We can't simply BUG() in riscv_init_cbom_blocksize() when we fail to find a block size because the failure will happen before logging works, leaving users to scratch their heads as to why the boot hung. Instead, ensure Zicbom is disabled and output an error which will hopefully alert people that the DT needs to be fixed. While at it, add a check that the block size is a power-of-2 too.

---
 arch/riscv/kernel/cpufeature.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
arch/riscv/kernel/cpufeature.c +1 -0
index 220be7222129..a4430a77df53 100644
@@ -9,6 +9,7 @@
 #include <linux/bitmap.h>
 #include <linux/ctype.h>
 #include <linux/libfdt.h>
+#include <linux/log2.h>
 #include <linux/module.h>
 #include <linux/of.h>
 #include <asm/alternative.h>
@@ -70,6 +71,20 @@ EXPORT_SYMBOL_GPL(__riscv_isa_extension_available);
 
Quoted 11 lines

arch/riscv/kernel/cpufeature.c EXPORT_SYMBOL_GPL(__riscv_isa_extension_available); · line 77

Quoted 6 lines
arch/riscv/kernel/cpufeature.c +4 -0
 static bool riscv_isa_extension_check(int id)
 {
+	switch (id) {
+	case RISCV_ISA_EXT_ZICBOM:
+		if (!riscv_cbom_block_size) {
+			if (IS_ENABLED(CONFIG_RISCV_ISA_ZICBOM))

Maybe I've missed something... Why not check if !IS_ENABLED() & return false immediately rather than doing it inside the attribute checks? I'll be compiled out of zicbom is enabled, so not like the non-error patch will be penalised with an extra check.

Quoted 5 lines

Ths IS_ENABLED checks are only here to decide if we want to complain or not, not to decide if we want to add the extension to the isa bitmap or not. As we've decided for the KVM compilation fix for Zicbom patch we don't mind the kernel knowing the extension 'foo' is present, even when it's compiled without CONFIG_RISCV_ISA_FOO. Indeed, for Zicbom, KVM can

Of course, how did I forget last weeks breakage already...

Quoted 4 lines

still offer the extension to guests, whether the host wants to use it or not. So, for this code, we can't factor out the IS_ENABLED checks since they're tied to the pr_err()s they're guarding, which are tied to the sanity checks they're inside.

So yeah, this all makes sense.

Quoted 13 lines

Thinking about it some more, though, maybe we should unconditionally complain, but with a different message, one that doesn't imply the kernel wanted to use the extension. Something like

  if (!riscv_cbom_block_size) {
      pr_err("Zicbom detected in ISA string, but no cbom-block-size found\n");
      return false;
  } else if (!is_power_of_2(riscv_cbom_block_size)) {
      pr_err("riscv_cbom_block_size present, but it is not a power-of-2\n");
      return false;
  }
Quoted 2 lines

I'm still not sure if those should be errors or warnings when CONFIG_RISCV_ISA_ZICBOM=no, though.

My initial thought here was along the lines of "why care about the issues when it's not enabled", but I went to get some caffeine before replying and had a change of heart.
IMO, it'd be good to tell people that their DT doesn't describe their hardware properly when it's gone wrong in a way that we cannot really spot via dtbs_check etc. I'll go away and have a think about whether the DT checks can force cbom-block-size if zicbom is in the isa string, but I'm not sure if that sort of conditional is checkable. If either of these error cases are hit (regardless of whether you set CONFIG_RISCV_ISA_ZICBOM) then it'd cause issues for KVM right? That would imply printing unconditionally to me. The level at which to actually emit that is beyond the scope of what I care about though ;)

Thanks,
Conor.

Quoted 17 lines
Quoted 17 lines
Quoted 17 lines
arch/riscv/kernel/cpufeature.c +10 -0
+				pr_err("cbom-block-size not found, cannot use Zicbom\n");
+			return false;
+		} else if (!is_power_of_2(riscv_cbom_block_size)) {
+			if (IS_ENABLED(CONFIG_RISCV_ISA_ZICBOM))
+				pr_err("cbom-block-size is not a power-of-2, cannot use Zicbom\n");
+			return false;
+		}
+		return true;
+	}
+
 	return true;
 }
 
Signature · 1 line
2.37.3
Quoted 9 lines
Quoted 4 lines
Quoted 4 lines

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

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

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

#11 lore raw
Quoted 75 lines · Conor Dooley
Quoted 43 lines · Andrew Jones
Quoted 30 lines · Conor Dooley
Quoted 29 lines · Andrew Jones

When a DT puts zicbom in the isa string, but does not provide a block size, ALT_CMO_OP() will attempt to do cache operations on address zero since the start address will be ANDed with zero. We can't simply BUG() in riscv_init_cbom_blocksize() when we fail to find a block size because the failure will happen before logging works, leaving users to scratch their heads as to why the boot hung. Instead, ensure Zicbom is disabled and output an error which will hopefully alert people that the DT needs to be fixed. While at it, add a check that the block size is a power-of-2 too.

---
 arch/riscv/kernel/cpufeature.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)
arch/riscv/kernel/cpufeature.c +1 -0
index 220be7222129..a4430a77df53 100644
@@ -9,6 +9,7 @@
 #include <linux/bitmap.h>
 #include <linux/ctype.h>
 #include <linux/libfdt.h>
+#include <linux/log2.h>
 #include <linux/module.h>
 #include <linux/of.h>
 #include <asm/alternative.h>
@@ -70,6 +71,20 @@ EXPORT_SYMBOL_GPL(__riscv_isa_extension_available);
 
Quoted 11 lines

arch/riscv/kernel/cpufeature.c EXPORT_SYMBOL_GPL(__riscv_isa_extension_available); · line 77

Quoted 6 lines
arch/riscv/kernel/cpufeature.c +4 -0
 static bool riscv_isa_extension_check(int id)
 {
+	switch (id) {
+	case RISCV_ISA_EXT_ZICBOM:
+		if (!riscv_cbom_block_size) {
+			if (IS_ENABLED(CONFIG_RISCV_ISA_ZICBOM))

Maybe I've missed something... Why not check if !IS_ENABLED() & return false immediately rather than doing it inside the attribute checks? I'll be compiled out of zicbom is enabled, so not like the non-error patch will be penalised with an extra check.

Quoted 5 lines

Ths IS_ENABLED checks are only here to decide if we want to complain or not, not to decide if we want to add the extension to the isa bitmap or not. As we've decided for the KVM compilation fix for Zicbom patch we don't mind the kernel knowing the extension 'foo' is present, even when it's compiled without CONFIG_RISCV_ISA_FOO. Indeed, for Zicbom, KVM can

Of course, how did I forget last weeks breakage already...

Quoted 4 lines

still offer the extension to guests, whether the host wants to use it or not. So, for this code, we can't factor out the IS_ENABLED checks since they're tied to the pr_err()s they're guarding, which are tied to the sanity checks they're inside.

So yeah, this all makes sense.

Quoted 13 lines

Thinking about it some more, though, maybe we should unconditionally complain, but with a different message, one that doesn't imply the kernel wanted to use the extension. Something like

  if (!riscv_cbom_block_size) {
      pr_err("Zicbom detected in ISA string, but no cbom-block-size found\n");
      return false;
  } else if (!is_power_of_2(riscv_cbom_block_size)) {
      pr_err("riscv_cbom_block_size present, but it is not a power-of-2\n");
      return false;
  }
Quoted 2 lines

I'm still not sure if those should be errors or warnings when CONFIG_RISCV_ISA_ZICBOM=no, though.

Quoted 11 lines

My initial thought here was along the lines of "why care about the issues when it's not enabled", but I went to get some caffeine before replying and had a change of heart.
IMO, it'd be good to tell people that their DT doesn't describe their hardware properly when it's gone wrong in a way that we cannot really spot via dtbs_check etc. I'll go away and have a think about whether the DT checks can force cbom-block-size if zicbom is in the isa string, but I'm not sure if that sort of conditional is checkable. If either of these error cases are hit (regardless of whether you set CONFIG_RISCV_ISA_ZICBOM) then it'd cause issues for KVM right? That would imply printing unconditionally to me.

Yes and no. KVM won't have any issues as it won't see Zicbom in the ISA bitmap and therefore won't consider offering it to guests. However, it could have offered it to guests, if only the DT was fixed, so I agree, it's best to unconditionally complain when we see something wrong.

Quoted 2 lines

The level at which to
actually emit that is beyond the scope of what I care about though ;)

This stackoverflow.com thread indicates I should use 'error'.
https://stackoverflow.com/questions/2031163/when-to-use-the-different-log-levels

I'll send a v2 with IS_ENABLED dropped and the wordage changed to not imply that the kernel wanted to use the extension, in case it didn't.

Thanks,
drew

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

#12 lore raw
Quoted 13 lines · Andrew Jones
Quoted 11 lines · Conor Dooley

My initial thought here was along the lines of "why care about the issues when it's not enabled", but I went to get some caffeine before replying and had a change of heart.
IMO, it'd be good to tell people that their DT doesn't describe their hardware properly when it's gone wrong in a way that we cannot really spot via dtbs_check etc. I'll go away and have a think about whether the DT checks can force cbom-block-size if zicbom is in the isa string, but I'm not sure if that sort of conditional is checkable. If either of these error cases are hit (regardless of whether you set CONFIG_RISCV_ISA_ZICBOM) then it'd cause issues for KVM right? That would imply printing unconditionally to me.

Quoted 4 lines

Yes and no. KVM won't have any issues as it won't see Zicbom in the ISA bitmap and therefore won't consider offering it to guests. However, it could have offered it to guests, if only the DT was fixed, so I agree, it's best to unconditionally complain when we see something wrong.

Yeah, that's pretty much what I was getting at. By "KVM having issues" I meant that without disabling Zicbom, as your patch does, KVM would carry on & offer Zicbom with the "bad" cbom block size to guests. Following that line of thought, we'd then want to disable Zicbom & complain no matter the state of the kconfig symbol so that there'd be info there for the folks running a non-Zicbom kernel but trying to use Zicbom in their guest.

Quoted 6 lines
Quoted 3 lines

The level at which to
actually emit that is beyond the scope of what I care about though ;)

This stackoverflow.com thread indicates I should use 'error'.
https://stackoverflow.com/questions/2031163/when-to-use-the-different-log-levels

Yeah, on the basis of that thread error does seem appropriate.

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