Description of problem:
Attempting to access context variables from a kernel function .return probe gives a compiler error on Fedora 36 / 5.17.5-300.fc36.x86_64:

  # stap -e 'probe kernel.function("dm_lock_for_deletion").return { printf("%s", kernel_string($md->name)) }'
  WARNING: confusing usage, value is captured as @entry($md->name) in .return probe [man stapprobes] RETURN PROBES: identifier '$md' at <input>:1:83
   source: probe kernel.function("dm_lock_for_deletion").return { printf("%s", kernel_string($md->name)) }
                                                                                             ^
  warning: the compiler differs from the one used to build the kernel
    The kernel was built by: gcc (GCC) 12.0.1 20220413 (Red Hat 12.0.1-0)
    You are using:           gcc (GCC) 12.1.1 20220507 (Red Hat 12.1.1-1)
  /tmp/stap3oyVT2/stap_f6ccae714cb9f51153235f45abadb532_2993_src.c: In function ‘_kretprobe_data’:
  /tmp/stap3oyVT2/stap_f6ccae714cb9f51153235f45abadb532_2993_src.c:34:46: error: ‘struct kretprobe_instance’ has no member named ‘rp’; did you mean ‘rph’?
     34 |         if (end > offset && pi && end <= pi->rp->data_size)
        |                                              ^~
        |                                              rph
  make[1]: *** [scripts/Makefile.build:288: /tmp/stap3oyVT2/stap_f6ccae714cb9f51153235f45abadb532_2993_src.o] Error 1
  make: *** [Makefile:1841: /tmp/stap3oyVT2] Error 2
  WARNING: kbuild exited with status: 2
  Pass 4: compilation failed.  [man error::pass4]

This is due to upstream commit d741bf4 (v5.11-rc1~176^2~6):

  commit d741bf41d7c7db4898bacfcb020353cddc032fd8
  Author: Peter Zijlstra <peterz>
  Date:   Sat Aug 29 22:03:24 2020 +0900

    kprobes: Remove kretprobe hash
    
    The kretprobe hash is mostly superfluous, replace it with a per-task
    variable.
    
    This gets rid of the task hash and it's related locking.
    
    Note that this may change the kprobes module-exported API for kretprobe
    handlers. If any out-of-tree kretprobe user uses ri->rp, use
    get_kretprobe(ri) instead.
    
    Signed-off-by: Peter Zijlstra (Intel) <peterz>
    Signed-off-by: Masami Hiramatsu <mhiramat>
    Signed-off-by: Ingo Molnar <mingo>
    Link: https://lore.kernel.org/r/159870620431.1229682.16325792502413731312.stgit@devnote2

This removes the "kretprobe *rp" member from struct kretprobe_instance and inserts a new "struct kretprobe_holder *rph" in its place: accessing the kretprobe now goes via ri->rhp->rh rather than directly ri->rh.

Applying the following diff to tapset/linux/kretprobe.stp fixes the problem for me on F36 and allowed me to get at the variables of interest:

--- /usr/share/systemtap/tapset/linux/kretprobe.stp.orig	2022-05-06 15:13:40.000000000 +0100
+++ /usr/share/systemtap/tapset/linux/kretprobe.stp	2022-06-08 17:28:59.324207123 +0100
@@ -20,7 +20,7 @@ _kretprobe_data(struct kretprobe_instanc
 {
 #if LINUX_VERSION_CODE >= KERNEL_VERSION(2,6,25)
 	size_t end = offset + length;
-	if (end > offset && pi && end <= pi->rp->data_size)
+	if (end > offset && pi && end <= pi->rph->rp->data_size)
 		return &pi->data[offset];
 #endif
 	return NULL;

Version-Release number of selected component (if applicable):
systemtap-runtime-4.7-1.fc36.x86_64
systemtap-client-4.7-1.fc36.x86_64
systemtap-4.7-1.fc36.x86_64
systemtap-devel-4.7-1.fc36.x86_64


How reproducible:
100% on affected kernels

Steps to Reproduce:
1. uname -r >= 5.11
2. stap -e 'probe kernel.function("dm_lock_for_deletion").return { printf("%s", kernel_string($md->name)) }' (any kernel.function().return probe should work, as long as it has a context variable you can access in the probe body - without that I do not see the compile error as the _kretprobe_data() bits don't get compiled).

Actual results:
  /tmp/stap3oyVT2/stap_f6ccae714cb9f51153235f45abadb532_2993_src.c: In function ‘_kretprobe_data’:
  /tmp/stap3oyVT2/stap_f6ccae714cb9f51153235f45abadb532_2993_src.c:34:46: error: ‘struct kretprobe_instance’ has no member named ‘rp’; did you mean ‘rph’?
     34 |         if (end > offset && pi && end <= pi->rp->data_size)
        |                                              ^~
        |                                              rph
  make[1]: *** [scripts/Makefile.build:288: /tmp/stap3oyVT2/stap_f6ccae714cb9f51153235f45abadb532_2993_src.o] Error 1

Expected results:
No error. Variable values available in probe.

Additional info:

I'm happy to fix up the diff to do a proper check for the different version ranges and submit the patch upstream if this seems correct (or switch it to using get_kretprobe() for those versions as suggested in the commit).