From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from mail-ot1-f47.google.com (mail-ot1-f47.google.com [209.85.210.47]) (using TLSv1.2 with cipher ECDHE-RSA-AES128-GCM-SHA256 (128/128 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id DF89F346E7A for ; Thu, 7 May 2026 16:54:55 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=209.85.210.47 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1778172899; cv=none; b=OXF+5oKCcRYFnppzyUFccWKaWFNSVuFkDIGloiQsQEQr6r2FmTbjtem42RUtLof8v14HhZfB8g5znCtXaPzSErKoMRlLcU7TymNGXHlLmS4yg6G+OMqDcihY1h8ZdVZISMk1KOysqfQx4aWlwRxBHrCkhwSVQsD3M1p28skQyp0= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1778172899; c=relaxed/simple; bh=dde3GX5hj/UJ5hROJEoexQzYZMjo82yVdn2mONdrBNQ=; h=Date:From:To:Cc:Subject:Message-ID:References:MIME-Version: Content-Type:Content-Disposition:In-Reply-To; b=Pqs8FVWS9PcvCIZyqSpKePwhIEpGUqBv+fS2byxFWImBWPFE3t7nrEmntCTnrb1XZOgTF2kWPMokFAA2LM8iIav/+JtELG35ZDuQHpcH6hcUtKXACqsN7rOSMrwyNtRUJ3kEbyGSIGYWZBGTgHNL6eXYZgA5GN/C0x5o6XeceEk= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=minyard.net; spf=pass smtp.mailfrom=minyard.net; dkim=pass (2048-bit key) header.d=minyard.net header.i=@minyard.net header.b=I866x1Oq; arc=none smtp.client-ip=209.85.210.47 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=minyard.net Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=minyard.net Authentication-Results: smtp.subspace.kernel.org; dkim=pass (2048-bit key) header.d=minyard.net header.i=@minyard.net header.b="I866x1Oq" Received: by mail-ot1-f47.google.com with SMTP id 46e09a7af769-7dcd89701acso1022581a34.1 for ; Thu, 07 May 2026 09:54:55 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=minyard.net; s=google; t=1778172895; x=1778777695; darn=vger.kernel.org; h=in-reply-to:content-disposition:mime-version:references:reply-to :message-id:subject:cc:to:from:date:from:to:cc:subject:date :message-id:reply-to; bh=MVa/SANg2C2YOjW+mWwzuGeV40cb1Yh8WbrkHXk9mnA=; b=I866x1OqR6FGIH38c5ZAqHZB048ptWn1SCBWB8vj374pOURVSaigXDPfZKI3Wpk8iP lHtLja3vg5eFbO26BTR2NYWja06tRJR9a/6NedqdtBfpks5hLM7gV/Q/tpKXSTNpHD0+ +OEFLbyORnutl82OasbPEeQy0EPJpGZb8183s9TC8FWMQFLpDN13fPccHM9SGs8ApyzV 41EuFIiNKx+1KRlCcCGBMhlVQ+c6iOGS9ixWX7Xm4TNz4TvMZRfxiGQcsLwv3RKCdGBr ehd8rJVHmeqpHzM70jAdb4d5UpOsNId0HvFcgOITVY0U+2wxGdwivo2LEHKfiqjNvtcK nfsQ== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20251104; t=1778172895; x=1778777695; h=in-reply-to:content-disposition:mime-version:references:reply-to :message-id:subject:cc:to:from:date:x-gm-gg:x-gm-message-state:from :to:cc:subject:date:message-id:reply-to; bh=MVa/SANg2C2YOjW+mWwzuGeV40cb1Yh8WbrkHXk9mnA=; b=kQvBzhGLWR/wTinmBQRFCdjCYivXzuDnKRSqKMRK9GR/9wVCO9R3gss2xoHHedqvPs eQuDEXltu9zVvDjW+lS5y4FiG4uN7EBg44aueWkNHigFrfWuRROkIxisaq1dIgqFHY8I AH9KHsid+RgEGBMWd0loU697/5H06bRIYUvM7LL0wTwDP7MT7/h+f2JvC2QglcxE2vZl i9erlIZrfvWBa1BiAJoG5XmsU2tLKfD37cyQt1dB3wP3p4vWr62SafHFvPJBZyiJkKNX ZgQXi4LPq2CJE4ZdFeCP5E/q6z8d2KhEHfu4Qzv5Z2UnXtxQKdORjtSqTPdwjqNBKEVg mARQ== X-Forwarded-Encrypted: i=1; AFNElJ+lb/77fllCDuWeHooSjaJ746fUYCSzVhJaVC2qZlGx7P5flJjqwmfcs8C7v+bkUxoz0Z0e0WXqQOc=@vger.kernel.org X-Gm-Message-State: AOJu0YwxEus/n5LhiR6iqRa8xSMozx746zUQyctvk/ulbWC1i2z3LaIR 1q/e1q7e9csucfzpRzGLYxg92X/7KQnz+mFg2EXuM/7f/R2i0nOuzNFNEpXtCb3oDGM= X-Gm-Gg: AeBDiesKemy5dYTxMDD7xUMvs70BsrmZyPHJzCgU3v6OJQRBXxVvPywT5OxGSAiOa+i oqMPLm66fFxvoACifqZF+k4VnTgeEAuH/mQUZmiRenIKaudN5eCpRlr+OKVHybUgcRPs6hr60MI KXJTAyAeu731hz7bkKilkEo7rUfttsRNSJwBp2cG8qukZnDJD7gNm2ksRiP2kqIO5Ij3eNu9nSX APBQWJqeCQovjOY4CVzAO0PiCSBnsHczSx8+L73yj6fPXJR7su5mL9qry9zvfCSGrIKKs3soTSz /fdivvcci2PPx8merUaLmxaiEr5mJ0fXbsBZSFBD0fR2sl3sOiREle5s8nQt4ffjOB/ciRYuJGn AZqtZT79Zms9Qxjy1aDtwc830LhaIdKjd4i4+wyECW7GDGqZKuTXLTp5scb+rwqnut0eDRMwc1l eENsNBzC+PHkLBpvmH9+ddBsX7lOQoiUf0wgzl28RkmM4/uklr8LC3MbOmOThcg2ctKZM5CKPCI ELn754baN1ZChqgDWIN3R0X X-Received: by 2002:a4a:e845:0:b0:696:637e:4831 with SMTP id 006d021491bc7-69998ccc23cmr5040401eaf.25.1778172894572; Thu, 07 May 2026 09:54:54 -0700 (PDT) Received: from mail.minyard.net ([2001:470:b8f6:1b:6f35:6142:1491:cdf]) by smtp.gmail.com with ESMTPSA id 006d021491bc7-6968945dc14sm12051955eaf.6.2026.05.07.09.54.53 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Thu, 07 May 2026 09:54:53 -0700 (PDT) Date: Thu, 7 May 2026 11:54:50 -0500 From: Corey Minyard To: Tony Hutter Cc: Lukas Wunner , Bjorn Helgaas , alok.a.tiwari@oracle.com, mariusz.tkaczyk@linux.intel.com, minyard@acm.org, linux-pci@vger.kernel.org, openipmi-developer@lists.sourceforge.net, Linux Kernel Mailing List Subject: Re: [PATCH v8 RESEND] Introduce Cray ClusterStor E1000 NVMe slot LED driver Message-ID: Reply-To: corey@minyard.net References: <768409a2-0593-49bd-9065-e8b93c60d4ce@llnl.gov> <02a02a9d-6d8d-4c91-bdad-0a14d0211952@llnl.gov> Precedence: bulk X-Mailing-List: linux-pci@vger.kernel.org List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <02a02a9d-6d8d-4c91-bdad-0a14d0211952@llnl.gov> Well dang it, I sent the previous reply using the wrong email client and it stuck in some HTML. On Thu, May 07, 2026 at 09:42:53AM -0700, Tony Hutter wrote: > > Have you tested removing and adding the IPMI interface while this is up? > > You can do that with the hotmod interface on IPMI. > > Thanks for the tip Corey. I just tried reloading the device via hotmod, and the craye1k driver worked as expected. Here's the hotmod removal + add dmesg lines: All looks good. I expected it to work, but it's always good to test. -corey > > craye1k: Got unexpected smi_gone, iface=0 > ipmi_si hotmod-ipmi-si.3: ipmi_platform: probing via hotmod > ipmi_platform: ipmi_si: hotmod: io 0xca2 regsize 1 spacing 1 irq 0 > ipmi_si: Adding hotmod-specified kcs state machine > ipmi_si: Trying hotmod-specified kcs state machine at i/o address 0xca2, slave address 0x20, irq 0 > ipmi_si hotmod-ipmi-si.3: IPMI message handler: Found new BMC (man_id: 0x002415, prod_id: 0x0101, dev_id: 0x20) > ipmi_si hotmod-ipmi-si.3: Cray ClusterStor E1000 slot LEDs registered > ipmi_si hotmod-ipmi-si.3: IPMI kcs interface initialized > > -Tony > > On 4/29/26 19:20, Corey Minyard wrote: > > On Wed, Apr 29, 2026 at 04:22:55PM -0700, Tony Hutter wrote: > >> Add driver to control the NVMe slot LEDs on the Cray ClusterStor E1000. > >> The driver provides hotplug attention status callbacks for the 24 NVMe > >> slots on the E1000. This allows users to access the E1000's locate and > >> fault LEDs via the normal /sys/bus/pci/slots//attention sysfs > >> entries. This driver uses IPMI to communicate with the E1000 controller > >> to toggle the LEDs. > >> > >> Signed-off-by: Tony Hutter > > > > For the IPMI portions: > > Reviewed-by: Corey Minyard > > > > Have you tested removing and adding the IPMI interface while this is up? > > You can do that with the hotmod interface on IPMI. I didn't see any > > issues, but it's always good to test. > > > > -corey > > > >> --- > >> Changes in v8: > >> - Remove unused variable in craye1k_get_attention_status(). > >> > >> Changes in v7: > >> - Update sysfs-bus-pci text from feedback. > >> - Add DMI dependency to Kconfig. > >> - Refactor pciehp_core.c to remove CONFIG_HOTPLUG_PCI_PCIE_CRAY_E1000 > >> code blocks. > >> - Move errno.h #include into correct alphabetical order. > >> - Add comment describing the reasoning for the debugfs counters. > >> - Move craye1k_init() call from pcie_hp_init() to init_slot(). > >> - Make craye1k mutex global rather than in craye1k->lock. This enables > >> handling of craye1k_[get|set]_attention_status() calls before the craye1k > >> driver is initialized. > >> - Do driver cleanup on craye1k_smi_gone(). > >> > >> Changes in v6: > >> - Change some dev_info_ratelimited() calls to dev_info(). > >> - Don't call craye1k_init() if pcie_port_service_register() fails. > >> - Fix stray indent in #define CRAYE1K_POST_CMD_WAIT_MS > >> > >> Changes in v5: > >> - Removed unnecessary ipmi_smi.h #include. > >> - Added WARN_ON() to craye1k_do_message() to sanity check that craye1k->lock > >> is held. > >> - Added additional comments for when craye1k->lock should be held. > >> > >> Changes in v4: > >> - Fix typo in Kconfig: "is it" -> "it is" > >> - Rename some #defines to CRAYE1K_SUBCMD_* > >> - Use IS_ERR() check in craye1k_debugfs_init() > >> - Return -EIO instead of -EINVAL when LED value check fails > >> > >> Changes in v3: > >> - Add 'attention' values in Documentation/ABI/testing/sysfs-bus-pci. > >> - Remove ACPI_PCI_SLOT dependency. > >> - Cleanup craye1k_do_message() error checking. > >> - Skip unneeded memcpy() on failure in __craye1k_do_command(). > >> - Merge craye1k_do_command_and_netfn() code into craye1k_do_command(). > >> - Make craye1k_is_primary() return boolean. > >> - Return negative error code on failure in craye1k_set_primary(). > >> > >> Changes in v2: > >> - Integrated E1000 code into the pciehp driver as an built-in > >> extention rather than as a standalone module. > >> - Moved debug tunables and counters to debugfs. > >> - Removed forward declarations. > >> - Kept the /sys/bus/pci/slots//attention interface rather > >> than using NPEM/_DSM or led_classdev as suggested. The "attention" > >> interface is more beneficial for our site, since it allows us to > >> control the NVMe slot LEDs agnostically across different enclosure > >> vendors and kernel versions using the same scripts. It is also > >> nice to use the same /sys/bus/pci/slots// sysfs directory for > >> both slot LED toggling ("attention") and slot power control > >> ("power"). > >> --- > >> Documentation/ABI/testing/sysfs-bus-pci | 21 + > >> MAINTAINERS | 5 + > >> drivers/pci/hotplug/Kconfig | 10 + > >> drivers/pci/hotplug/Makefile | 3 + > >> drivers/pci/hotplug/pciehp.h | 20 + > >> drivers/pci/hotplug/pciehp_core.c | 20 +- > >> drivers/pci/hotplug/pciehp_craye1k.c | 687 ++++++++++++++++++++++++ > >> 7 files changed, 765 insertions(+), 1 deletion(-) > >> create mode 100644 drivers/pci/hotplug/pciehp_craye1k.c > >> > >> diff --git a/Documentation/ABI/testing/sysfs-bus-pci b/Documentation/ABI/testing/sysfs-bus-pci > >> index 92debe879ffb..8536d2ff30d1 100644 > >> --- a/Documentation/ABI/testing/sysfs-bus-pci > >> +++ b/Documentation/ABI/testing/sysfs-bus-pci > >> @@ -231,6 +231,27 @@ Description: > >> - scXX contains the device subclass; > >> - iXX contains the device class programming interface. > >> > >> +What: /sys/bus/pci/slots/.../attention > >> +Date: February 2025 > >> +Contact: linux-pci@vger.kernel.org > >> +Description: > >> + The attention attribute is used to read or write the attention > >> + status for an enclosure slot. This is often used to set the > >> + slot LED value on a NVMe storage enclosure. > >> + > >> + Common values: > >> + 0 = OFF > >> + 1 = ON > >> + 2 = blink > >> + > >> + Using the Cray ClusterStor E1000 extensions: > >> + 0 = fault LED OFF, locate LED OFF > >> + 1 = fault LED ON, locate LED OFF > >> + 2 = fault LED OFF, locate LED ON > >> + 3 = fault LED ON, locate LED ON > >> + > >> + Other values are no-op, OFF, or ON depending on the driver. > >> + > >> What: /sys/bus/pci/slots/.../module > >> Date: June 2009 > >> Contact: linux-pci@vger.kernel.org > >> diff --git a/MAINTAINERS b/MAINTAINERS > >> index 9ac254f4ec41..861576d60a36 100644 > >> --- a/MAINTAINERS > >> +++ b/MAINTAINERS > >> @@ -6543,6 +6543,11 @@ S: Maintained > >> F: Documentation/filesystems/cramfs.rst > >> F: fs/cramfs/ > >> > >> +CRAY CLUSTERSTOR E1000 NVME SLOT LED DRIVER EXTENSIONS > >> +M: Tony Hutter > >> +S: Maintained > >> +F: drivers/pci/hotplug/pciehp_craye1k.c > >> + > >> CRC LIBRARY > >> M: Eric Biggers > >> R: Ard Biesheuvel > >> diff --git a/drivers/pci/hotplug/Kconfig b/drivers/pci/hotplug/Kconfig > >> index 3207860b52e4..3cb84e5e13e9 100644 > >> --- a/drivers/pci/hotplug/Kconfig > >> +++ b/drivers/pci/hotplug/Kconfig > >> @@ -183,4 +183,14 @@ config HOTPLUG_PCI_S390 > >> > >> When in doubt, say Y. > >> > >> +config HOTPLUG_PCI_PCIE_CRAY_E1000 > >> + bool "PCIe Hotplug extensions for Cray ClusterStor E1000" > >> + depends on DMI && HOTPLUG_PCI_PCIE && IPMI_HANDLER=y > >> + help > >> + Say Y here if you have a Cray ClusterStor E1000 and want to control > >> + your NVMe slot LEDs. Without this driver it is not possible > >> + to control the fault and locate LEDs on the E1000's 24 NVMe slots. > >> + > >> + When in doubt, say N. > >> + > >> endif # HOTPLUG_PCI > >> diff --git a/drivers/pci/hotplug/Makefile b/drivers/pci/hotplug/Makefile > >> index 40aaf31fe338..82a1f0592d0a 100644 > >> --- a/drivers/pci/hotplug/Makefile > >> +++ b/drivers/pci/hotplug/Makefile > >> @@ -66,6 +66,9 @@ pciehp-objs := pciehp_core.o \ > >> pciehp_ctrl.o \ > >> pciehp_pci.o \ > >> pciehp_hpc.o > >> +ifdef CONFIG_HOTPLUG_PCI_PCIE_CRAY_E1000 > >> +pciehp-objs += pciehp_craye1k.o > >> +endif > >> > >> shpchp-objs := shpchp_core.o \ > >> shpchp_ctrl.o \ > >> diff --git a/drivers/pci/hotplug/pciehp.h b/drivers/pci/hotplug/pciehp.h > >> index debc79b0adfb..3a8173f3e159 100644 > >> --- a/drivers/pci/hotplug/pciehp.h > >> +++ b/drivers/pci/hotplug/pciehp.h > >> @@ -199,6 +199,17 @@ int pciehp_get_raw_indicator_status(struct hotplug_slot *h_slot, u8 *status); > >> > >> int pciehp_slot_reset(struct pcie_device *dev); > >> > >> +#ifdef CONFIG_HOTPLUG_PCI_PCIE_CRAY_E1000 > >> +int craye1k_init(void); > >> +bool is_craye1k_board(void); > >> +int craye1k_get_attention_status(struct hotplug_slot *hotplug_slot, u8 *status); > >> +int craye1k_set_attention_status(struct hotplug_slot *hotplug_slot, u8 status); > >> +#else > >> +#define craye1k_init() (0) > >> +#define craye1k_get_attention_status NULL > >> +#define craye1k_set_attention_status NULL > >> +#endif > >> + > >> static inline const char *slot_name(struct controller *ctrl) > >> { > >> return hotplug_slot_name(&ctrl->hotplug_slot); > >> @@ -209,4 +220,13 @@ static inline struct controller *to_ctrl(struct hotplug_slot *hotplug_slot) > >> return container_of(hotplug_slot, struct controller, hotplug_slot); > >> } > >> > >> +static inline bool is_craye1k_slot(struct controller *ctrl) > >> +{ > >> +#ifdef CONFIG_HOTPLUG_PCI_PCIE_CRAY_E1000 > >> + return (PSN(ctrl) >= 1 && PSN(ctrl) <= 24 && is_craye1k_board()); > >> +#else > >> + return false; > >> +#endif > >> +} > >> + > >> #endif /* _PCIEHP_H */ > >> diff --git a/drivers/pci/hotplug/pciehp_core.c b/drivers/pci/hotplug/pciehp_core.c > >> index f59baa912970..3e8e2b3069bf 100644 > >> --- a/drivers/pci/hotplug/pciehp_core.c > >> +++ b/drivers/pci/hotplug/pciehp_core.c > >> @@ -72,6 +72,22 @@ static int init_slot(struct controller *ctrl) > >> } else if (ctrl->pcie->port->hotplug_user_indicators) { > >> ops->get_attention_status = pciehp_get_raw_indicator_status; > >> ops->set_attention_status = pciehp_set_raw_indicator_status; > >> + } else if (is_craye1k_slot(ctrl)) { > >> + /* > >> + * The Cray E1000 driver controls slots 1-24. Initialize the > >> + * Cray E1000 driver when slot 1 is seen. > >> + */ > >> + if (PSN(ctrl) == 1) { > >> + retval = craye1k_init(); > >> + if (retval) { > >> + ctrl_err(ctrl, > >> + "Error loading Cray E1000 extensions"); > >> + kfree(ops); > >> + return retval; > >> + } > >> + } > >> + ops->get_attention_status = craye1k_get_attention_status; > >> + ops->set_attention_status = craye1k_set_attention_status; > >> } > >> > >> /* register this slot with the hotplug pci core */ > >> @@ -376,8 +392,10 @@ int __init pcie_hp_init(void) > >> > >> retval = pcie_port_service_register(&hpdriver_portdrv); > >> pr_debug("pcie_port_service_register = %d\n", retval); > >> - if (retval) > >> + if (retval) { > >> pr_debug("Failure to register service\n"); > >> + return retval; > >> + } > >> > >> return retval; > >> } > >> diff --git a/drivers/pci/hotplug/pciehp_craye1k.c b/drivers/pci/hotplug/pciehp_craye1k.c > >> new file mode 100644 > >> index 000000000000..9c5bee81fdf8 > >> --- /dev/null > >> +++ b/drivers/pci/hotplug/pciehp_craye1k.c > >> @@ -0,0 +1,687 @@ > >> +// SPDX-License-Identifier: GPL-2.0 > >> +/* > >> + * Copyright 2022-2024 Lawrence Livermore National Security, LLC > >> + */ > >> +/* > >> + * Cray ClusterStor E1000 hotplug slot LED driver extensions > >> + * > >> + * This driver controls the NVMe slot LEDs on the Cray ClusterStore E1000. > >> + * It provides hotplug attention status callbacks for the 24 NVMe slots on > >> + * the E1000. This allows users to access the E1000's locate and fault > >> + * LEDs via the normal /sys/bus/pci/slots//attention sysfs entries. > >> + * This driver uses IPMI to communicate with the E1000 controller to toggle > >> + * the LEDs. > >> + * > >> + * This driver is based off of ibmpex.c > >> + */ > >> + > >> +#include > >> +#include > >> +#include > >> +#include > >> +#include > >> +#include > >> +#include > >> +#include > >> +#include > >> +#include "pciehp.h" > >> + > >> +/* Cray E1000 commands */ > >> +#define CRAYE1K_CMD_NETFN 0x3c > >> +#define CRAYE1K_CMD_PRIMARY 0x33 > >> +#define CRAYE1K_CMD_FAULT_LED 0x39 > >> +#define CRAYE1K_CMD_LOCATE_LED 0x22 > >> + > >> +/* Subcommands */ > >> +#define CRAYE1K_SUBCMD_GET_LED 0x0 > >> +#define CRAYE1K_SUBCMD_SET_LED 0x1 > >> +#define CRAYE1K_SUBCMD_SET_PRIMARY 0x1 > >> + > >> +/* > >> + * Milliseconds to wait after get/set LED command. 200ms value found though > >> + * experimentation > >> + */ > >> +#define CRAYE1K_POST_CMD_WAIT_MS 200 > >> + > >> +struct craye1k { > >> + struct device *dev; /* BMC device */ > >> + struct mutex lock; > >> + struct completion read_complete; > >> + struct ipmi_addr address; > >> + struct ipmi_user *user; > >> + int iface; > >> + > >> + long tx_msg_id; > >> + struct kernel_ipmi_msg tx_msg; > >> + unsigned char tx_msg_data[IPMI_MAX_MSG_LENGTH]; > >> + unsigned char rx_msg_data[IPMI_MAX_MSG_LENGTH]; > >> + unsigned long rx_msg_len; > >> + unsigned char rx_result; /* IPMI completion code */ > >> + > >> + /* Parent dir for all our debugfs entries */ > >> + struct dentry *parent; > >> + > >> + /* debugfs stats */ > >> + u64 check_primary; > >> + u64 check_primary_failed; > >> + u64 was_already_primary; > >> + u64 was_not_already_primary; > >> + u64 set_primary; > >> + u64 set_initial_primary_failed; > >> + u64 set_primary_failed; > >> + u64 set_led_locate_failed; > >> + u64 set_led_fault_failed; > >> + u64 set_led_readback_failed; > >> + u64 set_led_failed; > >> + u64 get_led_failed; > >> + u64 completion_timeout; > >> + u64 wrong_msgid; > >> + u64 request_failed; > >> + > >> + /* debugfs configuration options */ > >> + > >> + /* Print info on spurious IPMI messages */ > >> + bool print_errors; > >> + > >> + /* Retries for kernel IPMI layer */ > >> + u32 ipmi_retries; > >> + > >> + /* Timeout in ms for IPMI (0 = use IPMI default_retry_ms) */ > >> + u32 ipmi_timeout_ms; > >> + > >> + /* Timeout in ms to wait for E1000 message completion */ > >> + u32 completion_timeout_ms; > >> +}; > >> + > >> +/* > >> + * Make our craye1k a global so get/set_attention_status() can access it. > >> + * This is safe since there's only one node controller on the board, and so it's > >> + * impossible to instantiate more than one craye1k. > >> + */ > >> +static struct craye1k *craye1k_global; > >> +static DEFINE_MUTEX(craye1k_lock); > >> + > >> +/* > >> + * The E1000 command timeout and retry values were found though experimentation > >> + * by looking at the error counters. Keep the counters around to troubleshoot > >> + * any issues with our current timeout/retry values. > >> + */ > >> +static struct dentry * > >> +craye1k_debugfs_init(struct craye1k *craye1k) > >> +{ > >> + umode_t mode = 0644; > >> + struct dentry *parent = debugfs_create_dir("pciehp_craye1k", NULL); > >> + > >> + if (IS_ERR(parent)) > >> + return NULL; > >> + > >> + debugfs_create_x64("check_primary", mode, parent, > >> + &craye1k->check_primary); > >> + debugfs_create_x64("check_primary_failed", mode, parent, > >> + &craye1k->check_primary_failed); > >> + debugfs_create_x64("was_already_primary", mode, parent, > >> + &craye1k->was_already_primary); > >> + debugfs_create_x64("was_not_already_primary", mode, parent, > >> + &craye1k->was_not_already_primary); > >> + debugfs_create_x64("set_primary", mode, parent, > >> + &craye1k->set_primary); > >> + debugfs_create_x64("set_initial_primary_failed", mode, parent, > >> + &craye1k->set_initial_primary_failed); > >> + debugfs_create_x64("set_primary_failed", mode, parent, > >> + &craye1k->set_primary_failed); > >> + debugfs_create_x64("set_led_locate_failed", mode, parent, > >> + &craye1k->set_led_locate_failed); > >> + debugfs_create_x64("set_led_fault_failed", mode, parent, > >> + &craye1k->set_led_fault_failed); > >> + debugfs_create_x64("set_led_readback_failed", mode, parent, > >> + &craye1k->set_led_readback_failed); > >> + debugfs_create_x64("set_led_failed", mode, parent, > >> + &craye1k->set_led_failed); > >> + debugfs_create_x64("get_led_failed", mode, parent, > >> + &craye1k->get_led_failed); > >> + debugfs_create_x64("completion_timeout", mode, parent, > >> + &craye1k->completion_timeout); > >> + debugfs_create_x64("wrong_msgid", mode, parent, > >> + &craye1k->wrong_msgid); > >> + debugfs_create_x64("request_failed", mode, parent, > >> + &craye1k->request_failed); > >> + > >> + debugfs_create_x32("ipmi_retries", mode, parent, > >> + &craye1k->ipmi_retries); > >> + debugfs_create_x32("ipmi_timeout_ms", mode, parent, > >> + &craye1k->ipmi_timeout_ms); > >> + debugfs_create_x32("completion_timeout_ms", mode, parent, > >> + &craye1k->completion_timeout_ms); > >> + debugfs_create_bool("print_errors", mode, parent, > >> + &craye1k->print_errors); > >> + > >> + /* Return parent dir dentry */ > >> + return parent; > >> +} > >> + > >> +/* > >> + * craye1k_msg_handler() - IPMI message response handler > >> + */ > >> +static void craye1k_msg_handler(struct ipmi_recv_msg *msg, void *user_msg_data) > >> +{ > >> + struct craye1k *craye1k = user_msg_data; > >> + > >> + if (msg->msgid != craye1k->tx_msg_id) { > >> + craye1k->wrong_msgid++; > >> + if (craye1k->print_errors) { > >> + dev_warn_ratelimited(craye1k->dev, > >> + "rx msgid %ld != %ld", > >> + msg->msgid, craye1k->tx_msg_id); > >> + } > >> + ipmi_free_recv_msg(msg); > >> + return; > >> + } > >> + > >> + /* Set rx_result to the IPMI completion code */ > >> + if (msg->msg.data_len > 0) > >> + craye1k->rx_result = msg->msg.data[0]; > >> + else > >> + craye1k->rx_result = IPMI_UNKNOWN_ERR_COMPLETION_CODE; > >> + > >> + if (msg->msg.data_len > 1) { > >> + /* Exclude completion code from data bytes */ > >> + craye1k->rx_msg_len = msg->msg.data_len - 1; > >> + memcpy(craye1k->rx_msg_data, msg->msg.data + 1, > >> + craye1k->rx_msg_len); > >> + } else { > >> + craye1k->rx_msg_len = 0; > >> + } > >> + > >> + ipmi_free_recv_msg(msg); > >> + > >> + complete(&craye1k->read_complete); > >> +} > >> + > >> +static const struct ipmi_user_hndl craye1k_user_hndl = { > >> + .ipmi_recv_hndl = craye1k_msg_handler > >> +}; > >> + > >> +static void craye1k_new_smi(int iface, struct device *dev) > >> +{ > >> + int rc; > >> + struct craye1k *craye1k; > >> + > >> + craye1k = kzalloc(sizeof(*craye1k), GFP_KERNEL); > >> + if (!craye1k) > >> + return; > >> + > >> + craye1k->address.addr_type = IPMI_SYSTEM_INTERFACE_ADDR_TYPE; > >> + craye1k->address.channel = IPMI_BMC_CHANNEL; > >> + craye1k->iface = iface; > >> + craye1k->dev = dev; > >> + craye1k->tx_msg.data = craye1k->tx_msg_data; > >> + craye1k->ipmi_retries = 4; > >> + craye1k->ipmi_timeout_ms = 500; > >> + craye1k->completion_timeout_ms = 300; > >> + > >> + init_completion(&craye1k->read_complete); > >> + > >> + dev_set_drvdata(dev, craye1k); > >> + > >> + rc = ipmi_create_user(craye1k->iface, &craye1k_user_hndl, craye1k, > >> + &craye1k->user); > >> + if (rc < 0) { > >> + dev_err(dev, "Unable to register IPMI user, iface %d\n", > >> + craye1k->iface); > >> + kfree(craye1k); > >> + dev_set_drvdata(dev, NULL); > >> + return; > >> + } > >> + > >> + mutex_lock(&craye1k_lock); > >> + > >> + /* There's only one node controller so driver data should not be set */ > >> + WARN_ON(craye1k_global); > >> + > >> + craye1k_global = craye1k; > >> + craye1k->parent = craye1k_debugfs_init(craye1k); > >> + mutex_unlock(&craye1k_lock); > >> + if (!craye1k->parent) > >> + dev_warn(dev, "Cannot create debugfs"); > >> + > >> + dev_info(dev, "Cray ClusterStor E1000 slot LEDs registered"); > >> +} > >> + > >> +static void craye1k_smi_gone(int iface) > >> +{ > >> + pr_warn("craye1k: Got unexpected smi_gone, iface=%d", iface); > >> + > >> + mutex_lock(&craye1k_lock); > >> + if (craye1k_global) { > >> + debugfs_remove_recursive(craye1k_global->parent); > >> + kfree(craye1k_global); > >> + craye1k_global = NULL; > >> + } > >> + mutex_unlock(&craye1k_lock); > >> +} > >> + > >> +static struct ipmi_smi_watcher craye1k_smi_watcher = { > >> + .owner = THIS_MODULE, > >> + .new_smi = craye1k_new_smi, > >> + .smi_gone = craye1k_smi_gone > >> +}; > >> + > >> +/* > >> + * craye1k_send_message() - Send the message already setup in 'craye1k' > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Return: 0 on success, non-zero on error. > >> + */ > >> +static int craye1k_send_message(struct craye1k *craye1k) > >> +{ > >> + int rc; > >> + > >> + rc = ipmi_validate_addr(&craye1k->address, sizeof(craye1k->address)); > >> + if (rc) { > >> + dev_err_ratelimited(craye1k->dev, "ipmi_validate_addr() = %d\n", > >> + rc); > >> + return rc; > >> + } > >> + > >> + craye1k->tx_msg_id++; > >> + > >> + rc = ipmi_request_settime(craye1k->user, &craye1k->address, > >> + craye1k->tx_msg_id, &craye1k->tx_msg, craye1k, > >> + 0, craye1k->ipmi_retries, > >> + craye1k->ipmi_timeout_ms); > >> + > >> + if (rc) { > >> + craye1k->request_failed++; > >> + return rc; > >> + } > >> + > >> + return 0; > >> +} > >> + > >> +/* > >> + * craye1k_do_message() - Send the message in 'craye1k' and wait for a response > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Return: 0 on success, non-zero on error. > >> + */ > >> +static int craye1k_do_message(struct craye1k *craye1k) > >> +{ > >> + int rc; > >> + struct completion *read_complete = &craye1k->read_complete; > >> + unsigned long tout = msecs_to_jiffies(craye1k->completion_timeout_ms); > >> + > >> + WARN_ON(!mutex_is_locked(&craye1k_lock)); > >> + > >> + rc = craye1k_send_message(craye1k); > >> + if (rc) > >> + return rc; > >> + > >> + rc = wait_for_completion_killable_timeout(read_complete, tout); > >> + if (rc == 0) { > >> + /* timed out */ > >> + craye1k->completion_timeout++; > >> + return -ETIME; > >> + } > >> + > >> + return 0; > >> +} > >> + > >> +/* > >> + * __craye1k_do_command() - Do an IPMI command > >> + * > >> + * Send a command with optional data bytes, and read back response bytes. > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Returns: 0 on success, non-zero on error. > >> + */ > >> +static int __craye1k_do_command(struct craye1k *craye1k, u8 netfn, u8 cmd, > >> + u8 *send_data, u8 send_data_len, u8 *recv_data, > >> + u8 recv_data_len) > >> +{ > >> + int rc; > >> + > >> + craye1k->tx_msg.netfn = netfn; > >> + craye1k->tx_msg.cmd = cmd; > >> + > >> + if (send_data) { > >> + memcpy(&craye1k->tx_msg_data[0], send_data, send_data_len); > >> + craye1k->tx_msg.data_len = send_data_len; > >> + } else { > >> + craye1k->tx_msg_data[0] = 0; > >> + craye1k->tx_msg.data_len = 0; > >> + } > >> + > >> + rc = craye1k_do_message(craye1k); > >> + if (rc == 0) > >> + memcpy(recv_data, craye1k->rx_msg_data, recv_data_len); > >> + > >> + return rc; > >> +} > >> + > >> +/* > >> + * craye1k_do_command() - Do a Cray E1000 specific IPMI command. > >> + * @cmd: Cray E1000 specific command > >> + * @send_data: Data to send after the command > >> + * @send_data_len: Data length > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Returns: the last byte from the response or 0 if response had no response > >> + * data bytes, else -1 on error. > >> + */ > >> +static int craye1k_do_command(struct craye1k *craye1k, u8 cmd, u8 *send_data, > >> + u8 send_data_len) > >> +{ > >> + int rc; > >> + > >> + rc = __craye1k_do_command(craye1k, CRAYE1K_CMD_NETFN, cmd, send_data, > >> + send_data_len, NULL, 0); > >> + if (rc != 0) { > >> + /* Error attempting command */ > >> + return -1; > >> + } > >> + > >> + if (craye1k->tx_msg.data_len == 0) > >> + return 0; > >> + > >> + /* Return last received byte value */ > >> + return craye1k->rx_msg_data[craye1k->rx_msg_len - 1]; > >> +} > >> + > >> +/* > >> + * __craye1k_set_primary() - Tell the BMC we want to be the primary server > >> + * > >> + * An E1000 board has two physical servers on it. In order to set a slot > >> + * NVMe LED, this server needs to first tell the BMC that it's the primary > >> + * server. > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Returns: 0 on success, non-zero on error. > >> + */ > >> +static int __craye1k_set_primary(struct craye1k *craye1k) > >> +{ > >> + u8 bytes[2] = {CRAYE1K_SUBCMD_SET_PRIMARY, 1}; /* set primary to 1 */ > >> + > >> + craye1k->set_primary++; > >> + return craye1k_do_command(craye1k, CRAYE1K_CMD_PRIMARY, bytes, 2); > >> +} > >> + > >> +/* > >> + * craye1k_is_primary() - Are we the primary server? > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Returns: true if we are the primary server, false otherwise. > >> + */ > >> +static bool craye1k_is_primary(struct craye1k *craye1k) > >> +{ > >> + u8 byte = 0; > >> + int rc; > >> + > >> + /* Response byte is 0x1 on success */ > >> + rc = craye1k_do_command(craye1k, CRAYE1K_CMD_PRIMARY, &byte, 1); > >> + craye1k->check_primary++; > >> + if (rc == 0x1) > >> + return true; /* success */ > >> + > >> + craye1k->check_primary_failed++; > >> + return false; /* We are not the primary server node */ > >> +} > >> + > >> +/* > >> + * craye1k_set_primary() - Attempt to set ourselves as the primary server > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Returns: 0 on success, -1 otherwise. > >> + */ > >> +static int craye1k_set_primary(struct craye1k *craye1k) > >> +{ > >> + int tries = 10; > >> + > >> + if (craye1k_is_primary(craye1k)) { > >> + craye1k->was_already_primary++; > >> + return 0; > >> + } > >> + craye1k->was_not_already_primary++; > >> + > >> + /* delay found through experimentation */ > >> + msleep(300); > >> + > >> + if (__craye1k_set_primary(craye1k) != 0) { > >> + craye1k->set_initial_primary_failed++; > >> + return -1; /* error */ > >> + } > >> + > >> + /* > >> + * It can take 2 to 3 seconds after setting primary for the controller > >> + * to report that it is the primary. > >> + */ > >> + while (tries--) { > >> + msleep(500); > >> + if (craye1k_is_primary(craye1k)) > >> + break; > >> + } > >> + > >> + if (tries == 0) { > >> + craye1k->set_primary_failed++; > >> + return -1; /* never reported that it's primary */ > >> + } > >> + > >> + /* Wait for primary switch to finish */ > >> + msleep(1500); > >> + > >> + return 0; > >> +} > >> + > >> +/* > >> + * craye1k_get_slot_led() - Get slot LED value > >> + * @slot: Slot number (1-24) > >> + * @is_locate_led: 0 = get fault LED value, 1 = get locate LED value > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Returns: slot value on success, -1 on failure. > >> + */ > >> +static int craye1k_get_slot_led(struct craye1k *craye1k, unsigned char slot, > >> + bool is_locate_led) > >> +{ > >> + u8 bytes[2]; > >> + u8 cmd; > >> + > >> + bytes[0] = CRAYE1K_SUBCMD_GET_LED; > >> + bytes[1] = slot; > >> + > >> + cmd = is_locate_led ? CRAYE1K_CMD_LOCATE_LED : CRAYE1K_CMD_FAULT_LED; > >> + > >> + return craye1k_do_command(craye1k, cmd, bytes, 2); > >> +} > >> + > >> +/* > >> + * craye1k_set_slot_led() - Attempt to set the locate/fault LED to a value > >> + * @slot: Slot number (1-24) > >> + * @is_locate_led: 0 = use fault LED, 1 = use locate LED > >> + * @value: Value to set (0 or 1) > >> + * > >> + * Check the LED value after calling this function to ensure it has been set > >> + * properly. > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Returns: 0 on success, non-zero on failure. > >> + */ > >> +static int craye1k_set_slot_led(struct craye1k *craye1k, unsigned char slot, > >> + unsigned char is_locate_led, > >> + unsigned char value) > >> +{ > >> + u8 bytes[3]; > >> + u8 cmd; > >> + > >> + bytes[0] = CRAYE1K_SUBCMD_SET_LED; > >> + bytes[1] = slot; > >> + bytes[2] = value; > >> + > >> + cmd = is_locate_led ? CRAYE1K_CMD_LOCATE_LED : CRAYE1K_CMD_FAULT_LED; > >> + > >> + return craye1k_do_command(craye1k, cmd, bytes, 3); > >> +} > >> + > >> +/* > >> + * __craye1k_get_attention_status() - Get LED value > >> + * > >> + * Context: craye1k_lock is already held. > >> + * Returns: 0 on success, -EIO on failure. > >> + */ > >> +static int __craye1k_get_attention_status(struct hotplug_slot *hotplug_slot, > >> + u8 *status, bool set_primary) > >> +{ > >> + unsigned char slot; > >> + int locate, fault; > >> + struct craye1k *craye1k; > >> + > >> + craye1k = craye1k_global; > >> + slot = PSN(to_ctrl(hotplug_slot)); > >> + > >> + if (set_primary) { > >> + if (craye1k_set_primary(craye1k) != 0) { > >> + craye1k->get_led_failed++; > >> + return -EIO; > >> + } > >> + } > >> + > >> + locate = craye1k_get_slot_led(craye1k, slot, true); > >> + if (locate == -1) { > >> + craye1k->get_led_failed++; > >> + return -EIO; > >> + } > >> + msleep(CRAYE1K_POST_CMD_WAIT_MS); > >> + > >> + fault = craye1k_get_slot_led(craye1k, slot, false); > >> + if (fault == -1) { > >> + craye1k->get_led_failed++; > >> + return -EIO; > >> + } > >> + msleep(CRAYE1K_POST_CMD_WAIT_MS); > >> + > >> + *status = locate << 1 | fault; > >> + > >> + return 0; > >> +} > >> + > >> +int craye1k_get_attention_status(struct hotplug_slot *hotplug_slot, > >> + u8 *status) > >> +{ > >> + int rc; > >> + > >> + if (mutex_lock_interruptible(&craye1k_lock) != 0) > >> + return -EINTR; > >> + > >> + if (!craye1k_global) { > >> + /* Driver isn't initialized yet */ > >> + mutex_unlock(&craye1k_lock); > >> + return -EOPNOTSUPP; > >> + } > >> + > >> + rc = __craye1k_get_attention_status(hotplug_slot, status, true); > >> + > >> + mutex_unlock(&craye1k_lock); > >> + return rc; > >> +} > >> + > >> +int craye1k_set_attention_status(struct hotplug_slot *hotplug_slot, > >> + u8 status) > >> +{ > >> + unsigned char slot; > >> + int tries = 4; > >> + int rc; > >> + u8 new_status; > >> + struct craye1k *craye1k; > >> + bool locate, fault; > >> + > >> + if (mutex_lock_interruptible(&craye1k_lock) != 0) > >> + return -EINTR; > >> + > >> + if (!craye1k_global) { > >> + /* Driver isn't initialized yet */ > >> + mutex_unlock(&craye1k_lock); > >> + return -EOPNOTSUPP; > >> + } > >> + > >> + craye1k = craye1k_global; > >> + > >> + slot = PSN(to_ctrl(hotplug_slot)); > >> + > >> + /* Retry to ensure all LEDs are set */ > >> + while (tries--) { > >> + /* > >> + * The node must first set itself to be the primary node before > >> + * setting the slot LEDs (each board has two nodes, or > >> + * "servers" as they're called by the manufacturer). This can > >> + * lead to contention if both nodes are trying to set the LEDs > >> + * at the same time. > >> + */ > >> + rc = craye1k_set_primary(craye1k); > >> + if (rc != 0) { > >> + /* Could not set as primary node. Just retry again. */ > >> + continue; > >> + } > >> + > >> + /* Write value twice to increase success rate */ > >> + locate = (status & 0x2) >> 1; > >> + craye1k_set_slot_led(craye1k, slot, 1, locate); > >> + if (craye1k_set_slot_led(craye1k, slot, 1, locate) != 0) { > >> + craye1k->set_led_locate_failed++; > >> + continue; /* fail, retry */ > >> + } > >> + > >> + msleep(CRAYE1K_POST_CMD_WAIT_MS); > >> + > >> + fault = status & 0x1; > >> + craye1k_set_slot_led(craye1k, slot, 0, fault); > >> + if (craye1k_set_slot_led(craye1k, slot, 0, fault) != 0) { > >> + craye1k->set_led_fault_failed++; > >> + continue; /* fail, retry */ > >> + } > >> + > >> + msleep(CRAYE1K_POST_CMD_WAIT_MS); > >> + > >> + rc = __craye1k_get_attention_status(hotplug_slot, &new_status, > >> + false); > >> + > >> + msleep(CRAYE1K_POST_CMD_WAIT_MS); > >> + > >> + if (rc == 0 && new_status == status) > >> + break; /* success */ > >> + > >> + craye1k->set_led_readback_failed++; > >> + > >> + /* > >> + * At this point we weren't successful in setting the LED and > >> + * need to try again. > >> + * > >> + * Do a random back-off to reduce contention with other server > >> + * node in the unlikely case that both server nodes are trying to > >> + * trying to set a LED at the same time. > >> + * > >> + * The 500ms minimum in the back-off reduced the chance of this > >> + * whole retry loop failing from 1 in 700 to none in 10000. > >> + */ > >> + msleep(500 + (get_random_long() % 500)); > >> + } > >> + mutex_unlock(&craye1k_lock); > >> + if (tries == 0) { > >> + craye1k->set_led_failed++; > >> + return -EIO; > >> + } > >> + > >> + return 0; > >> +} > >> + > >> +bool is_craye1k_board(void) > >> +{ > >> + return dmi_match(DMI_PRODUCT_NAME, "VSSEP1EC"); > >> +} > >> + > >> +int craye1k_init(void) > >> +{ > >> + return ipmi_smi_watcher_register(&craye1k_smi_watcher); > >> +} > >> + > >> +MODULE_LICENSE("GPL"); > >> +MODULE_AUTHOR("Tony Hutter "); > >> +MODULE_DESCRIPTION("Cray E1000 NVMe Slot LED driver"); > >> -- > >> 2.43.7 > >> > >> >