Skip to contentSkip to conversation

Request for 4.19.x backport (with conflicts)

PATCH Applied Laura Abbott 4 participants 6 messages Jun 21, 2019 – Jun 23, 2019

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

Hi,

We're attempting to build stable kernels with gcc9. 4.19.x fails to build with gcc9 as 146448524bdd ("s390/jump_label: Use "jdd" constraint on gcc9") is missing. This doesn't apply cleanly to 4.19.x as it needs changes from 13ddb52c165b ("s390/jump_label: Switch to relative references")

Which is better, taking both 13ddb52c165b and 146448524bdd or doing a backport of 146448524bdd?

Thanks,
Laura

#1 lore raw

Hello Laura,

Quoted 7 lines

We're attempting to build stable kernels with gcc9. 4.19.x fails to build with gcc9 as 146448524bdd ("s390/jump_label: Use "jdd" constraint on gcc9") is missing. This doesn't apply cleanly to 4.19.x as it needs changes from 13ddb52c165b ("s390/jump_label: Switch to relative references")

Which is better, taking both 13ddb52c165b and 146448524bdd or doing a backport of 146448524bdd?

I don't know which kernel version you are referring to exactly, however 4.19.53 from linux-stable does not contain the common code infrastructure for relative jump labels. The infrastructure was merged with 4.20: commit 50ff18ab497aa ("jump_label: Implement generic support for relative references").

Therefore a backport of only 146448524bdd seems to be the way to go.

#2 lore raw
Quoted 10 lines · Heiko Carstens

Hello Laura,

Quoted 7 lines

We're attempting to build stable kernels with gcc9. 4.19.x fails to build with gcc9 as 146448524bdd ("s390/jump_label: Use "jdd" constraint on gcc9") is missing. This doesn't apply cleanly to 4.19.x as it needs changes from 13ddb52c165b ("s390/jump_label: Switch to relative references")

Which is better, taking both 13ddb52c165b and 146448524bdd or doing a backport of 146448524bdd?

Quoted 8 lines

I don't know which kernel version you are referring to exactly, however 4.19.53 from linux-stable does not contain the common code infrastructure for relative jump labels. The infrastructure was merged with 4.20: commit 50ff18ab497aa ("jump_label: Implement generic support for relative references").

Therefore a backport of only 146448524bdd seems to be the way to go.

Ah okay, I didn't realize there was more needed, I was just looking at the clean cherry-pick. I'm not sure how to do the backport, if you give me the patch I can verify.

Thanks,
Laura

#3 lore raw

PATCH 1/1 s390/jump_label: Use "jdd" constraint on gcc9

Quoted 3 lines

Ah okay, I didn't realize there was more needed, I was just looking at the clean cherry-pick. I'm not sure how to do the backport, if you give me the patch I can verify.

Please find the cherry-picked 146448524bdd below.

I also had to cherry-pick 159491f3b509 to fix an unrelated compilation error and make the build fully work.

Best regards,
Ilya

----

[[email protected]]:
-----
Laura Abbott reported that the kernel doesn't build anymore with gcc 9, due to the "X" constraint. Ilya provided the gcc 9 patch "S/390:
Introduce jdd constraint" which introduces the new "jdd" constraint which fixes this.
-----

The support for section anchors on S/390 introduced in gcc9 has changed the behavior of "X" constraint, which can now produce register references. Since existing constraints, in particular, "i", do not fit the intended use case on S/390, the new machine-specific "jdd" constraint was introduced. This patch makes jump labels use "jdd" constraint when building with gcc9.

---
 arch/s390/include/asm/jump_label.h | 14 ++++++++++----
 1 file changed, 10 insertions(+), 4 deletions(-)
arch/s390/include/asm/jump_label.h +10 -4
index 40f651292aa7..9c7dc970e966 100644
@@ -10,6 +10,12 @@
 #define JUMP_LABEL_NOP_SIZE 6
 #define JUMP_LABEL_NOP_OFFSET 2
 
+#if __GNUC__ < 9
+#define JUMP_LABEL_STATIC_KEY_CONSTRAINT "X"
+#else
+#define JUMP_LABEL_STATIC_KEY_CONSTRAINT "jdd"
+#endif
+
 /*
  * We use a brcl 0,2 instruction for jump labels at compile time so it
  * can be easily distinguished from a hotpatch generated instruction.
@@ -19,9 +25,9 @@ static __always_inline bool arch_static_branch(struct static_key *key, bool bran
 	asm_volatile_goto("0:	brcl 0,"__stringify(JUMP_LABEL_NOP_OFFSET)"\n"
 		".pushsection __jump_table, \"aw\"\n"
 		".balign 8\n"
-		".quad 0b, %l[label], %0\n"
+		".quad 0b, %l[label], %0+%1\n"
 		".popsection\n"
-		: : "X" (&((char *)key)[branch]) : : label);
+		: : JUMP_LABEL_STATIC_KEY_CONSTRAINT (key), "i" (branch) : : label);
 
 	return false;
 label:
@@ -33,9 +39,9 @@ static __always_inline bool arch_static_branch_jump(struct static_key *key, bool
 	asm_volatile_goto("0:	brcl 15, %l[label]\n"
 		".pushsection __jump_table, \"aw\"\n"
 		".balign 8\n"
-		".quad 0b, %l[label], %0\n"
+		".quad 0b, %l[label], %0+%1\n"
 		".popsection\n"
-		: : "X" (&((char *)key)[branch]) : : label);
+		: : JUMP_LABEL_STATIC_KEY_CONSTRAINT (key), "i" (branch) : : label);
 
 	return false;
 label:
Signature · 1 line
2.21.0
#4 lore raw

Re: [PATCH] s390/jump_label: Use "jdd" constraint on gcc9

Quoted 9 lines · Ilya Leoshkevich
Quoted 3 lines

Ah okay, I didn't realize there was more needed, I was just looking at the clean cherry-pick. I'm not sure how to do the backport, if you give me the patch I can verify.

Please find the cherry-picked 146448524bdd below.

I also had to cherry-pick 159491f3b509 to fix an unrelated compilation error and make the build fully work.

Yes, this worked for me (plus 159491f3b509). Thanks!

Quoted 70 lines

Best regards,
Ilya

----

[[email protected]]:
-----
Laura Abbott reported that the kernel doesn't build anymore with gcc 9, due to the "X" constraint. Ilya provided the gcc 9 patch "S/390:
Introduce jdd constraint" which introduces the new "jdd" constraint which fixes this.
-----

The support for section anchors on S/390 introduced in gcc9 has changed the behavior of "X" constraint, which can now produce register references. Since existing constraints, in particular, "i", do not fit the intended use case on S/390, the new machine-specific "jdd" constraint was introduced. This patch makes jump labels use "jdd" constraint when building with gcc9.

---
  arch/s390/include/asm/jump_label.h | 14 ++++++++++----
  1 file changed, 10 insertions(+), 4 deletions(-)
arch/s390/include/asm/jump_label.h +10 -4
index 40f651292aa7..9c7dc970e966 100644
@@ -10,6 +10,12 @@
  #define JUMP_LABEL_NOP_SIZE 6
  #define JUMP_LABEL_NOP_OFFSET 2
 
+#if __GNUC__ < 9
+#define JUMP_LABEL_STATIC_KEY_CONSTRAINT "X"
+#else
+#define JUMP_LABEL_STATIC_KEY_CONSTRAINT "jdd"
+#endif
+
  /*
   * We use a brcl 0,2 instruction for jump labels at compile time so it
   * can be easily distinguished from a hotpatch generated instruction.
@@ -19,9 +25,9 @@ static __always_inline bool arch_static_branch(struct static_key *key, bool bran
  	asm_volatile_goto("0:	brcl 0,"__stringify(JUMP_LABEL_NOP_OFFSET)"\n"
  		".pushsection __jump_table, \"aw\"\n"
  		".balign 8\n"
-		".quad 0b, %l[label], %0\n"
+		".quad 0b, %l[label], %0+%1\n"
  		".popsection\n"
-		: : "X" (&((char *)key)[branch]) : : label);
+		: : JUMP_LABEL_STATIC_KEY_CONSTRAINT (key), "i" (branch) : : label);
 
  	return false;
  label:
@@ -33,9 +39,9 @@ static __always_inline bool arch_static_branch_jump(struct static_key *key, bool
  	asm_volatile_goto("0:	brcl 15, %l[label]\n"
  		".pushsection __jump_table, \"aw\"\n"
  		".balign 8\n"
-		".quad 0b, %l[label], %0\n"
+		".quad 0b, %l[label], %0+%1\n"
  		".popsection\n"
-		: : "X" (&((char *)key)[branch]) : : label);
+		: : JUMP_LABEL_STATIC_KEY_CONSTRAINT (key), "i" (branch) : : label);
 
  	return false;
  label:
maintainer Applied
#5 lore raw

Re: [PATCH] s390/jump_label: Use "jdd" constraint on gcc9

Quoted 5 lines · Laura Abbott
Quoted 4 lines · Ilya Leoshkevich
Quoted 3 lines

Ah okay, I didn't realize there was more needed, I was just looking at the clean cherry-pick. I'm not sure how to do the backport, if you give me the patch I can verify.

Quoted 7 lines
Quoted 5 lines

Please find the cherry-picked 146448524bdd below.

I also had to cherry-pick 159491f3b509 to fix an unrelated compilation error and make the build fully work.

Yes, this worked for me (plus 159491f3b509). Thanks!

I've queued the backport of 146448524bdd and 159491f3b509 to 4.19, thank you.

--
Thanks,
Sasha