[PATCH] um: fix time-travel syscall scheduling hack

[Johannes Berg]

From: Johannes Berg <johannes.berg at intel.com>

The schedule() call there really never did anything at
least since the introduction of the EEVDF scheduler,
but now I found a case where we permanently hang in a
loop of -ERESTARTNOINTR (due to locking.) Work around
it by making any syscalls with error return take time
(and then schedule after) so we cannot hang in such a
loop forever.

Signed-off-by: Johannes Berg <johannes.berg at intel.com>
---
 arch/um/kernel/skas/syscall.c | 34 ++++++++++++++++++++--------------
 1 file changed, 20 insertions(+), 14 deletions(-)

diff --git a/arch/um/kernel/skas/syscall.c b/arch/um/kernel/skas/syscall.c
index 9ee19e566da3..b09e85279d2b 100644
--- a/arch/um/kernel/skas/syscall.c
+++ b/arch/um/kernel/skas/syscall.c
@@ -12,23 +12,13 @@
 #include <sysdep/syscalls.h>
 #include <linux/time-internal.h>
 #include <asm/unistd.h>
+#include <asm/delay.h>
 
 void handle_syscall(struct uml_pt_regs *r)
 {
 	struct pt_regs *regs = container_of(r, struct pt_regs, regs);
 	int syscall;
 
-	/*
-	 * If we have infinite CPU resources, then make every syscall also a
-	 * preemption point, since we don't have any other preemption in this
-	 * case, and kernel threads would basically never run until userspace
-	 * went to sleep, even if said userspace interacts with the kernel in
-	 * various ways.
-	 */
-	if (time_travel_mode == TT_MODE_INFCPU ||
-	    time_travel_mode == TT_MODE_EXTERNAL)
-		schedule();
-
 	/* Initialize the syscall number and default return value. */
 	UPT_SYSCALL_NR(r) = PT_SYSCALL_NR(r->gp);
 	PT_REGS_SET_SYSCALL_RETURN(regs, -ENOSYS);
@@ -41,9 +31,25 @@ void handle_syscall(struct uml_pt_regs *r)
 		goto out;
 
 	syscall = UPT_SYSCALL_NR(r);
-	if (syscall >= 0 && syscall < __NR_syscalls)
-		PT_REGS_SET_SYSCALL_RETURN(regs,
-				EXECUTE_SYSCALL(syscall, regs));
+	if (syscall >= 0 && syscall < __NR_syscalls) {
+		unsigned long ret = EXECUTE_SYSCALL(syscall, regs);
+
+		PT_REGS_SET_SYSCALL_RETURN(regs, ret);
+
+		/*
+		 * An error value here can be some form of -ERESTARTSYS
+		 * and then we'd just loop. Make any error syscalls take
+		 * some time, so that it won't just loop if something is
+		 * not ready, and hopefully other things will make some
+		 * progress.
+		 */
+		if (IS_ERR_VALUE(ret) &&
+		    (time_travel_mode == TT_MODE_INFCPU ||
+		     time_travel_mode == TT_MODE_EXTERNAL)) {
+			um_udelay(1);
+			schedule();
+		}
+	}
 
 out:
 	syscall_trace_leave(regs);
-- 
2.46.0