linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: tip-bot for Andrew Banman <tipbot@zytor.com>
To: linux-tip-commits@vger.kernel.org
Cc: colin.king@canonical.com, hpa@zytor.com, mike.travis@hpe.com,
	sivanich@hpe.com, linux-kernel@vger.kernel.org, mingo@kernel.org,
	abanman@hpe.com, tglx@linutronix.de
Subject: [tip:x86/urgent] x86/platform/uv/BAU: Gracefully disable BAU during panic
Date: Mon, 20 Aug 2018 09:12:59 -0700	[thread overview]
Message-ID: <tip-99f3965878759d36baac944df004b4dafcc272b4@git.kernel.org> (raw)
In-Reply-To: <1534168477-311448-1-git-send-email-abanman@hpe.com>

Commit-ID:  99f3965878759d36baac944df004b4dafcc272b4
Gitweb:     https://git.kernel.org/tip/99f3965878759d36baac944df004b4dafcc272b4
Author:     Andrew Banman <abanman@hpe.com>
AuthorDate: Mon, 13 Aug 2018 08:54:37 -0500
Committer:  Thomas Gleixner <tglx@linutronix.de>
CommitDate: Mon, 20 Aug 2018 18:04:43 +0200

x86/platform/uv/BAU: Gracefully disable BAU during panic

When a panic happens while the BAU is active, outstanding broadcasts may go
un-acknowledged by the kernel. These broadcasts cause timeouts in the UV
ASIC that tie up BAU resources, which may cause a fatal error that brings
down the system, thereby crashing the kdump kexec.

Add uv_bau_crash_shutdown() to bring BAU to quiescence ahead of the crash
shutdown routine saved in smp machine_ops. Assign uv_bau_crash_shutdown
to machine_ops during init.

Signed-off-by: Andrew Banman <abanman@hpe.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Acked-by: Mike Travis <mike.travis@hpe.com>
Cc: "H. Peter Anvin" <hpa@zytor.com>
Cc: Colin Ian King <colin.king@canonical.com>
Cc: Dimitri Sivanich <sivanich@hpe.com>
Link: https://lkml.kernel.org/r/1534168477-311448-1-git-send-email-abanman@hpe.com

---
 arch/x86/platform/uv/tlb_uv.c | 49 +++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 49 insertions(+)

diff --git a/arch/x86/platform/uv/tlb_uv.c b/arch/x86/platform/uv/tlb_uv.c
index a4130b84d1ff..4c1e1197b82d 100644
--- a/arch/x86/platform/uv/tlb_uv.c
+++ b/arch/x86/platform/uv/tlb_uv.c
@@ -22,8 +22,13 @@
 #include <asm/tsc.h>
 #include <asm/irq_vectors.h>
 #include <asm/timer.h>
+#include <asm/reboot.h>
 
 static struct bau_operations ops __ro_after_init;
+#ifdef CONFIG_KEXEC_CORE
+static void (*crash_shutdown)(struct pt_regs *regs) __ro_after_init;
+static bool crash_in_progress;
+#endif
 
 /* timeouts in nanoseconds (indexed by UVH_AGING_PRESCALE_SEL urgency7 30:28) */
 static const int timeout_base_ns[] = {
@@ -2196,6 +2201,44 @@ static const struct bau_operations uv4_bau_ops __initconst = {
 	.wait_completion         = uv4_wait_completion,
 };
 
+#ifdef CONFIG_KEXEC_CORE
+/*
+ * Bring BAU to quiesence by abandoning current broadcasts and freeing up
+ * resources needed by firmware-initiated BAU broadcasts.
+ */
+void uv_bau_crash_shutdown(struct pt_regs *regs)
+{
+	int pnode = 0;
+	int uvhub = 0;
+
+	if (crash_in_progress)
+		return;
+	crash_in_progress = true;
+
+	/*
+	 * Don't bother turning off BAU in the per-cpu structs. We free up
+	 * enough INTD resources to accommodate any new broadcasts that
+	 * may (however unlikley) start up before we complete the panic,
+	 * without tying up FW-initiated General Broadcasts that must not
+	 * time out.
+	 */
+
+	for_each_possible_blade(uvhub) {
+		if (!uv_blade_nr_possible_cpus(uvhub))
+			continue;
+		pnode = uv_blade_to_pnode(uvhub);
+		/* Set STATUS registers to idle to free source cpus */
+		write_gmmr(pnode, UVH_LB_BAU_SB_ACTIVATION_STATUS_0, 0x0);
+		write_gmmr(pnode, UVH_LB_BAU_SB_ACTIVATION_STATUS_1, 0x0);
+		write_gmmr(pnode, UVH_LB_BAU_SB_ACTIVATION_STATUS_2, 0x0);
+		/* Clear TIMEOUT and PENDING bits to free up BAU resources */
+		ops.write_g_sw_ack(pnode, ops.read_g_sw_ack(pnode) & 0xFFFF);
+	}
+
+	crash_shutdown(regs);
+}
+#endif
+
 /*
  * Initialization of BAU-related structures
  */
@@ -2268,6 +2311,12 @@ static int __init uv_bau_init(void)
 		}
 	}
 
+#ifdef CONFIG_KEXEC_CORE
+	crash_shutdown = machine_ops.crash_shutdown;
+	machine_ops.crash_shutdown = uv_bau_crash_shutdown;
+	crash_in_progress = false;
+#endif
+
 	return 0;
 
 err_bau_disable:

      reply	other threads:[~2018-08-20 16:13 UTC|newest]

Thread overview: 2+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-08-13 13:54 [PATCH] x86/platform/uv/BAU: gracefully disable BAU during panic Andrew Banman
2018-08-20 16:12 ` tip-bot for Andrew Banman [this message]

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=tip-99f3965878759d36baac944df004b4dafcc272b4@git.kernel.org \
    --to=tipbot@zytor.com \
    --cc=abanman@hpe.com \
    --cc=colin.king@canonical.com \
    --cc=hpa@zytor.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-tip-commits@vger.kernel.org \
    --cc=mike.travis@hpe.com \
    --cc=mingo@kernel.org \
    --cc=sivanich@hpe.com \
    --cc=tglx@linutronix.de \
    /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).