All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH 1/5] linux-user/alpha: write the floating-point registers to a core dump
@ 2026-08-05 17:15 Matt Turner
  2026-08-05 17:15 ` [PATCH 2/5] linux-user/mips: " Matt Turner
                   ` (4 more replies)
  0 siblings, 5 replies; 6+ messages in thread
From: Matt Turner @ 2026-08-05 17:15 UTC (permalink / raw)
  To: qemu-devel; +Cc: Matt Turner, Laurent Vivier, Helge Deller, Pierrick Bouvier

A guest core carried only the general-purpose registers, so a debugger
opening one reported every floating-point register as unavailable --
including the arguments of the function that crashed.

Implement HAVE_ELF_CORE_FPREGS for Alpha: define target_elf_fpregset_t
to match the kernel's layout ($f0-$f30 plus the control register in the
slot $f31 would occupy) and fill it from elf_core_copy_fpregs().

Checked with lldb on a core from a program that faults with live values
in $f16 and $f17: both read back correctly, as does the control register.

Signed-off-by: Matt Turner <mattst88@gmail.com>
---
 linux-user/alpha/elfload.c    | 10 ++++++++++
 linux-user/alpha/target_elf.h | 11 +++++++++++
 2 files changed, 21 insertions(+)

diff --git ./linux-user/alpha/elfload.c ./linux-user/alpha/elfload.c
index 7be9e466b6..c2517516ad 100644
--- ./linux-user/alpha/elfload.c
+++ ./linux-user/alpha/elfload.c
@@ -32,3 +32,13 @@ const char *get_elf_cpu_model(uint32_t eflags)
 {
     return "ev67";
 }
+
+void elf_core_copy_fpregs(target_elf_fpregset_t *r, const CPUAlphaState *env)
+{
+    int i;
+
+    for (i = 0; i < 31; i++) {
+        r->fpr[i] = tswap64(env->fir[i]);
+    }
+    r->fpcr = tswap64(cpu_alpha_load_fpcr((CPUAlphaState *)env));
+}
diff --git ./linux-user/alpha/target_elf.h ./linux-user/alpha/target_elf.h
index dd90c6f783..5efe7f3174 100644
--- ./linux-user/alpha/target_elf.h
+++ ./linux-user/alpha/target_elf.h
@@ -19,6 +19,17 @@
  *   r0-r30 at indices 0-30, pc at 31, ps at 32.
  * r31 (hardwired zero) is not stored; pc occupies index 31.
  */
+/*
+ * The floating-point note holds $f0 through $f30 and then the control
+ * register in the slot $f31 would occupy; $f31 reads as zero.
+ */
+#define HAVE_ELF_CORE_FPREGS    1
+
+typedef struct target_elf_fpregset_t {
+    uint64_t fpr[31];    /* $f0-$f30 */
+    uint64_t fpcr;       /* the slot for $f31 */
+} target_elf_fpregset_t;
+
 typedef struct target_elf_gregset_t {
     abi_ulong regs[31];  /* integer registers r0-r30  [0..30] */
     abi_ulong pc;        /* program counter           [31]    */
-- 
2.54.0



^ permalink raw reply related	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2026-08-07 16:13 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-08-05 17:15 [PATCH 1/5] linux-user/alpha: write the floating-point registers to a core dump Matt Turner
2026-08-05 17:15 ` [PATCH 2/5] linux-user/mips: " Matt Turner
2026-08-05 17:15 ` [PATCH 3/5] linux-user/hppa: " Matt Turner
2026-08-05 17:15 ` [PATCH 4/5] linux-user/riscv: " Matt Turner
2026-08-05 17:15 ` [PATCH 5/5] linux-user/sh4: " Matt Turner
2026-08-07 16:12 ` [PATCH 1/5] linux-user/alpha: " Helge Deller

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.