gsp_sema_service: add content rescue for unmapped push-buffer pool VAs

- Add expected_sem_payload tracking to svc_channel for CE channels
- Implement svc_rescue_push_page() to scan guest RAM for push with
  expected semaphore release when VA translation fails
- Add svc_init_expected_payload() to initialize expected payload from pool
- Add rate-limited logging of distinct stalled VAs (svc_log_stalled_va)
- Advance expected_sem_payload on successful push decode
- Pushes SEMA_MAP_MAX to 65536, SVC_PIN_SLOTS to 32768 for scale

1 files changed, 67 insertions, 0 deletions

diff --git a/tools/guest-diag/peek.c b/tools/guest-diag/peek.c
new file mode 100644
index 0000000..857afd0
--- /dev/null
+++ b/tools/guest-diag/peek.c
@@ -0,0 +1,67 @@
+/* peek.c - print RIP/RBX/RSP/RDI of every thread of a process (root).
+   usage: peek PID [lo_off hi_off]   (offsets from the libcuda r-xp base)
+   Flags threads whose RIP is inside [base+lo, base+hi) with SPIN. */
+#define _GNU_SOURCE
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <dirent.h>
+#include <sys/ptrace.h>
+#include <sys/wait.h>
+#include <sys/user.h>
+#include <sys/uio.h>
+#include <unistd.h>
+
+int main(int argc, char **argv)
+{
+    if (argc < 2) { fprintf(stderr, "usage: peek PID [lo_off hi_off]\n"); return 1; }
+    int pid = atoi(argv[1]);
+    unsigned long lo = 0x23f700, hi = 0x23fd10;
+    if (argc >= 4) { lo = strtoul(argv[2], 0, 16); hi = strtoul(argv[3], 0, 16); }
+
+    char path[256];
+    snprintf(path, sizeof path, "/proc/%d/maps", pid);
+    FILE *f = fopen(path, "r");
+    if (!f) { perror("maps"); return 1; }
+    unsigned long base = 0;
+    char line[512];
+    while (fgets(line, sizeof line, f)) {
+        if (strstr(line, "libcuda") && strstr(line, "r-xp")) {
+            unsigned long a, b, off; char perms[8], dev[16], name[256]; unsigned long ino;
+            if (sscanf(line, "%lx-%lx %7s %lx %15s %lu %255s",
+                       &a, &b, perms, &off, dev, &ino, name) >= 6) {
+                base = a - off;
+                break;
+            }
+        }
+    }
+    fclose(f);
+    printf("libcuda base=0x%lx\n", base);
+
+    snprintf(path, sizeof path, "/proc/%d/task", pid);
+    DIR *d = opendir(path);
+    if (!d) { perror("task"); return 1; }
+    struct dirent *e;
+    while ((e = readdir(d))) {
+        if (e->d_name[0] == '.') continue;
+        int tid = atoi(e->d_name);
+        if (ptrace(PTRACE_ATTACH, tid, 0, 0)) { printf("tid %d attach fail\n", tid); continue; }
+        int st;
+        waitpid(tid, &st, 0);
+        struct user_regs_struct r;
+        struct iovec io = { &r, sizeof r };
+        if (ptrace(PTRACE_GETREGSET, tid, (void *)1, &io)) {
+            printf("tid %d getregs fail\n", tid);
+            ptrace(PTRACE_DETACH, tid, 0, 0);
+            continue;
+        }
+        printf("tid %d rip=0x%llx rsp=0x%llx rbx=0x%llx rdi=0x%llx%s\n",
+               tid,
+               (unsigned long long)r.rip, (unsigned long long)r.rsp,
+               (unsigned long long)r.rbx, (unsigned long long)r.rdi,
+               (r.rip >= base + lo && r.rip < base + hi) ? " SPIN" : "");
+        ptrace(PTRACE_DETACH, tid, 0, 0);
+    }
+    closedir(d);
+    return 0;
+}
\ No newline at end of file