Skip to contentSkip to conversation

libbpf: define bpf_tail_call_static when __clang__ is not defined

PATCH Superseded James Hilliard 3 participants 4 messages Sep 9, 2022 – Sep 12, 2022

Subsystems (2)

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 libbpf: define bpf_tail_call_static when __clang__ is not defined

The bpf_tail_call_static function is currently not defined unless using clang >= 8.

To support bpf_tail_call_static on GCC we can check if __clang__ is not defined to enable bpf_tail_call_static.

---
 tools/lib/bpf/bpf_helpers.h | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
tools/lib/bpf/bpf_helpers.h +5 -5
index 7349b16b8e2f..30fc95e7cd76 100644
@@ -131,7 +131,7 @@
 /*
  * Helper function to perform a tail call with a constant/immediate map slot.
  */
-#if __clang_major__ >= 8 && defined(__bpf__)
+#if (!defined(__clang__) || __clang_major__ >= 8) && defined(__bpf__)
 static __always_inline void
 bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
 {
@@ -139,8 +139,8 @@ bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
 		__bpf_unreachable();
 
 	/*
-	 * Provide a hard guarantee that LLVM won't optimize setting r2 (map
-	 * pointer) and r3 (constant map index) from _different paths_ ending
+	 * Provide a hard guarantee that the compiler won't optimize setting r2
+	 * (map pointer) and r3 (constant map index) from _different paths_ ending
 	 * up at the _same_ call insn as otherwise we won't be able to use the
 	 * jmpq/nopl retpoline-free patching by the x86-64 JIT in the kernel
 	 * given they mismatch. See also d2e4c1e6c294 ("bpf: Constant map key
@@ -148,8 +148,8 @@ bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
 	 *
 	 * Note on clobber list: we need to stay in-line with BPF calling
 	 * convention, so even if we don't end up using r0, r4, r5, we need
-	 * to mark them as clobber so that LLVM doesn't end up using them
-	 * before / after the call.
+	 * to mark them as clobber so that the compiler doesn't end up using
+	 * them before / after the call.
 	 */
 	asm volatile("r1 = %[ctx]\n\t"
 		     "r2 = %[map]\n\t"
Signature · 1 line
2.34.1
#1 lore raw
Quoted 38 lines · James Hilliard

The bpf_tail_call_static function is currently not defined unless using clang >= 8.

To support bpf_tail_call_static on GCC we can check if __clang__ is not defined to enable bpf_tail_call_static.

---
 tools/lib/bpf/bpf_helpers.h | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
tools/lib/bpf/bpf_helpers.h +3 -3
index 7349b16b8e2f..30fc95e7cd76 100644
@@ -131,7 +131,7 @@
 /*
  * Helper function to perform a tail call with a constant/immediate map slot.
  */
-#if __clang_major__ >= 8 && defined(__bpf__)
+#if (!defined(__clang__) || __clang_major__ >= 8) && defined(__bpf__)
 static __always_inline void
 bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
 {
@@ -139,8 +139,8 @@ bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
                __bpf_unreachable();
 
        /*
-        * Provide a hard guarantee that LLVM won't optimize setting r2 (map
-        * pointer) and r3 (constant map index) from _different paths_ ending
+        * Provide a hard guarantee that the compiler won't optimize setting r2
+        * (map pointer) and r3 (constant map index) from _different paths_ ending
         * up at the _same_ call insn as otherwise we won't be able to use the
         * jmpq/nopl retpoline-free patching by the x86-64 JIT in the kernel
         * given they mismatch. See also d2e4c1e6c294 ("bpf: Constant map key

tools/lib/bpf/bpf_helpers.h bpf_tail_call_static(void *ctx, const void *map, const __u32 slot) · line 155

Quoted 10 lines
tools/lib/bpf/bpf_helpers.h +2 -2
@@ -148,8 +148,8 @@ bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
         *
         * Note on clobber list: we need to stay in-line with BPF calling
         * convention, so even if we don't end up using r0, r4, r5, we need
-        * to mark them as clobber so that LLVM doesn't end up using them
-        * before / after the call.
+        * to mark them as clobber so that the compiler doesn't end up using
+        * them before / after the call.
         */
        asm volatile("r1 = %[ctx]\n\t"
                     "r2 = %[map]\n\t"

will this compile as is on GCC-BPF? I'm trying to understand what's the point. Once GCC supports this ASM syntax we can add similar check to __clang_major__, instead of allowing it for all GCC versions?

We must have done __clang_major__ check for a reason, old Clangs probably had some problems compiling this. Maybe Daniel remembers?

Quoted 3 lines

--
2.34.1

#2 lore raw
Quoted 54 lines · Andrii Nakryiko
Quoted 38 lines · James Hilliard

The bpf_tail_call_static function is currently not defined unless using clang >= 8.

To support bpf_tail_call_static on GCC we can check if __clang__ is not defined to enable bpf_tail_call_static.

---
  tools/lib/bpf/bpf_helpers.h | 10 +++++-----
  1 file changed, 5 insertions(+), 5 deletions(-)
tools/lib/bpf/bpf_helpers.h +3 -3
index 7349b16b8e2f..30fc95e7cd76 100644
@@ -131,7 +131,7 @@
  /*
   * Helper function to perform a tail call with a constant/immediate map slot.
   */
-#if __clang_major__ >= 8 && defined(__bpf__)
+#if (!defined(__clang__) || __clang_major__ >= 8) && defined(__bpf__)
  static __always_inline void
  bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
  {
@@ -139,8 +139,8 @@ bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
                 __bpf_unreachable();
 
         /*
-        * Provide a hard guarantee that LLVM won't optimize setting r2 (map
-        * pointer) and r3 (constant map index) from _different paths_ ending
+        * Provide a hard guarantee that the compiler won't optimize setting r2
+        * (map pointer) and r3 (constant map index) from _different paths_ ending
          * up at the _same_ call insn as otherwise we won't be able to use the
          * jmpq/nopl retpoline-free patching by the x86-64 JIT in the kernel
          * given they mismatch. See also d2e4c1e6c294 ("bpf: Constant map key

tools/lib/bpf/bpf_helpers.h bpf_tail_call_static(void *ctx, const void *map, const __u32 slot) · line 155

Quoted 10 lines
tools/lib/bpf/bpf_helpers.h +2 -2
@@ -148,8 +148,8 @@ bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
          *
          * Note on clobber list: we need to stay in-line with BPF calling
          * convention, so even if we don't end up using r0, r4, r5, we need
-        * to mark them as clobber so that LLVM doesn't end up using them
-        * before / after the call.
+        * to mark them as clobber so that the compiler doesn't end up using
+        * them before / after the call.
          */
         asm volatile("r1 = %[ctx]\n\t"
                      "r2 = %[map]\n\t"

will this compile as is on GCC-BPF? I'm trying to understand what's the point. Once GCC supports this ASM syntax we can add similar check to __clang_major__, instead of allowing it for all GCC versions?

Quoted 2 lines

We must have done __clang_major__ check for a reason, old Clangs probably had some problems compiling this. Maybe Daniel remembers?

Yes, clang >= 8 is needed to ensure 'slot' is propagated as 'const' instead of variables. clang 6 added support for inline asm but clang 8 is needed for above 'const' propagation into asm code.

Quoted 4 lines
Quoted 4 lines

--
2.34.1

#3 lore raw
In reply to Andrii Nakryiko
Quoted 46 lines · Andrii Nakryiko
Quoted 44 lines · James Hilliard

The bpf_tail_call_static function is currently not defined unless using clang >= 8.

To support bpf_tail_call_static on GCC we can check if __clang__ is not defined to enable bpf_tail_call_static.

---
 tools/lib/bpf/bpf_helpers.h | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
tools/lib/bpf/bpf_helpers.h +4 -5
index 7349b16b8e2f..30fc95e7cd76 100644
@@ -131,7 +131,7 @@
 /*
  * Helper function to perform a tail call with a constant/immediate map slot.
  */
-#if __clang_major__ >= 8 && defined(__bpf__)
+#if (!defined(__clang__) || __clang_major__ >= 8) && defined(__bpf__)
 static __always_inline void
 bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
 {
@@ -139,8 +139,8 @@ bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
                __bpf_unreachable();
 
        /*
-        * Provide a hard guarantee that LLVM won't optimize setting r2 (map
-        * pointer) and r3 (constant map index) from _different paths_ ending
+        * Provide a hard guarantee that the compiler won't optimize setting r2
+        * (map pointer) and r3 (constant map index) from _different paths_ ending
         * up at the _same_ call insn as otherwise we won't be able to use the
         * jmpq/nopl retpoline-free patching by the x86-64 JIT in the kernel
         * given they mismatch. See also d2e4c1e6c294 ("bpf: Constant map key
@@ -148,8 +148,8 @@ bpf_tail_call_static(void *ctx, const void *map, const __u32 slot)
         *
         * Note on clobber list: we need to stay in-line with BPF calling
         * convention, so even if we don't end up using r0, r4, r5, we need
-        * to mark them as clobber so that LLVM doesn't end up using them
-        * before / after the call.
+        * to mark them as clobber so that the compiler doesn't end up using
Quoted 8 lines

tools/lib/bpf/bpf_helpers.h bpf_tail_call_static(void *ctx, const void *map, const __u32 slot) · line 155

Quoted 4 lines
tools/lib/bpf/bpf_helpers.h +1 -0
+        * them before / after the call.
         */
        asm volatile("r1 = %[ctx]\n\t"
                     "r2 = %[map]\n\t"

will this compile as is on GCC-BPF? I'm trying to understand what's the point. Once GCC supports this ASM syntax we can add similar check to __clang_major__, instead of allowing it for all GCC versions?

So, it's a bit different since the ASM syntax issue is more of an issue with GAS,
we would need a check along these lines I think for a proper migration path:
https://lore.kernel.org/bpf/[email protected]/

We would set something like once GAS support for the llvm syntax is added:
.ifdef .gasversion. < 24000

I don't think we can do a compiler version check here like with llvm since GCC release cycles are out of sync with GAS(llvm assembler is kept in sync with the compiler AFAIU).

Quoted 7 lines

We must have done __clang_major__ check for a reason, old Clangs probably had some problems compiling this. Maybe Daniel remembers?

Quoted 3 lines

--
2.34.1