kvm.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Nico Boehr <nrb@linux.ibm.com>
To: kvm@vger.kernel.org
Cc: frankja@linux.ibm.com, imbrenda@linux.ibm.com, thuth@redhat.com
Subject: [kvm-unit-tests PATCH v4 3/3] s390x: smp: add tests for calls in wait state
Date: Wed, 10 Aug 2022 09:46:16 +0200	[thread overview]
Message-ID: <20220810074616.1223561-4-nrb@linux.ibm.com> (raw)
In-Reply-To: <20220810074616.1223561-1-nrb@linux.ibm.com>

When the SIGP interpretation facility is in use a SIGP external call to
a waiting CPU will result in an exit of the calling cpu. For non-pv
guests it's a code 56 (partial execution) exit otherwise its a code 108
(secure instruction notification) exit. Those exits are handled
differently from a normal SIGP instruction intercept that happens
without interpretation and hence need to be tested.

Signed-off-by: Nico Boehr <nrb@linux.ibm.com>
---
 s390x/smp.c | 97 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 97 insertions(+)

diff --git a/s390x/smp.c b/s390x/smp.c
index 5a269087581f..91f3e3bcc12a 100644
--- a/s390x/smp.c
+++ b/s390x/smp.c
@@ -356,6 +356,102 @@ static void test_calls(void)
 	}
 }
 
+static void call_in_wait_ext_int_fixup(struct stack_frame_int *stack)
+{
+	/* Clear wait bit so we don't immediately wait again after the fixup */
+	lowcore.ext_old_psw.mask &= ~PSW_MASK_WAIT;
+}
+
+static void call_in_wait_setup(void)
+{
+	expect_ext_int();
+	ctl_set_bit(0, current_sigp_call_case->cr0_bit);
+	register_ext_cleanup_func(call_in_wait_ext_int_fixup);
+
+	set_flag(1);
+}
+
+static void call_in_wait_received(void)
+{
+	report(lowcore.ext_int_code == current_sigp_call_case->ext_int_expected_type, "received");
+
+	set_flag(1);
+}
+
+static void call_in_wait_cleanup(void)
+{
+	ctl_clear_bit(0, current_sigp_call_case->cr0_bit);
+	register_ext_cleanup_func(NULL);
+
+	set_flag(1);
+}
+
+static void test_calls_in_wait(void)
+{
+	int i;
+	struct psw psw;
+
+	report_prefix_push("psw wait");
+	for (i = 0; i < ARRAY_SIZE(cases_sigp_call); i++) {
+		current_sigp_call_case = &cases_sigp_call[i];
+
+		report_prefix_push(current_sigp_call_case->name);
+		if (!current_sigp_call_case->supports_pv && uv_os_is_guest()) {
+			report_skip("Not supported under PV");
+			report_prefix_pop();
+			continue;
+		}
+
+		/* Let the secondary CPU setup the external mask and the external interrupt cleanup function */
+		set_flag(0);
+		psw.mask = extract_psw_mask();
+		psw.addr = (unsigned long)call_in_wait_setup;
+		smp_cpu_start(1, psw);
+
+		/* Wait until the receiver has finished setup */
+		wait_for_flag();
+		set_flag(0);
+
+		/*
+		 * To avoid races, we need to know that the secondary CPU has entered wait,
+		 * but the architecture provides no way to check whether the secondary CPU
+		 * is in wait.
+		 *
+		 * But since a waiting CPU is considered operating, simply stop the CPU, set
+		 * up the restart new PSW mask in wait, send the restart interrupt and then
+		 * wait until the CPU becomes operating (done by smp_cpu_start).
+		 */
+		smp_cpu_stop(1);
+		psw.mask = extract_psw_mask() | PSW_MASK_EXT | PSW_MASK_WAIT;
+		psw.addr = (unsigned long)call_in_wait_received;
+		smp_cpu_start(1, psw);
+
+		smp_sigp(1, current_sigp_call_case->call, 0, NULL);
+
+		/* Wait until the receiver has handled the call */
+		wait_for_flag();
+		smp_cpu_stop(1);
+		set_flag(0);
+
+		/*
+		 * Now clean up the mess we have left behind. If the cleanup
+		 * were part of call_in_wait_received we would not get a chance
+		 * to catch an interrupt that is presented twice since we would
+		 * disable the external call on the first interrupt.
+		 */
+		psw.mask = extract_psw_mask();
+		psw.addr = (unsigned long)call_in_wait_cleanup;
+		smp_cpu_start(1, psw);
+
+		/* Wait until the cleanup has been completed */
+		wait_for_flag();
+		smp_cpu_stop(1);
+
+		report_prefix_pop();
+	}
+	report_prefix_pop();
+}
+
 static void test_sense_running(void)
 {
 	report_prefix_push("sense_running");
@@ -474,6 +570,7 @@ int main(void)
 	test_store_status();
 	test_set_prefix();
 	test_calls();
+	test_calls_in_wait();
 	test_sense_running();
 	test_reset();
 	test_reset_initial();
-- 
2.36.1


  parent reply	other threads:[~2022-08-10  7:46 UTC|newest]

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-08-10  7:46 [kvm-unit-tests PATCH v4 0/3] s390x: add tests for SIGP call orders in enabled wait Nico Boehr
2022-08-10  7:46 ` [kvm-unit-tests PATCH v4 1/3] s390x: smp: move sigp calls with invalid cpu address to array Nico Boehr
2022-08-16 12:12   ` Claudio Imbrenda
2022-08-10  7:46 ` [kvm-unit-tests PATCH v4 2/3] s390x: smp: use an array for sigp calls Nico Boehr
2022-08-16 15:09   ` Claudio Imbrenda
2022-08-10  7:46 ` Nico Boehr [this message]
2022-08-16 15:37   ` [kvm-unit-tests PATCH v4 3/3] s390x: smp: add tests for calls in wait state Claudio Imbrenda
2022-08-17  7:34   ` Janosch Frank
2022-08-17 14:50     ` Nico Boehr

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220810074616.1223561-4-nrb@linux.ibm.com \
    --to=nrb@linux.ibm.com \
    --cc=frankja@linux.ibm.com \
    --cc=imbrenda@linux.ibm.com \
    --cc=kvm@vger.kernel.org \
    --cc=thuth@redhat.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).