[PATCH] ipmi: OEM flag handling and hacks for some Dell machines

The ipmi driver does not have a way to handle firmware-generated events which have the OEM[012] Data Available flags set. In such a case, the SMS_ATN bit may never get cleared by firmware, leaving the driver looping infinitely but never able to make any progress. This patch first simplifies storage and use of the data returned from an IPMI Get Device ID command. It then creates a new per-OEM handler hook, which should know how to handle events with the OEM[012] Data Available flags set. It then uses this to implement a workaround for IPMI 1.5-capable Dell PowerEdge servers which are susceptable to setting the OEM[012] Data Available flags when the driver can't handle it. Signed-off-by: Matt Domsch <Matt_Domsch@dell.com> Signed-off-by: Corey Minyard <minyard@acm.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>

[PATCH] ipmi: OEM flag handling and hacks for some Dell machines
The ipmi driver does not have a way to handle firmware-generated events which have the OEM[012] Data Available flags set. In such a case, the SMS_ATN bit may never get cleared by firmware, leaving the driver looping infinitely but never able to make any progress. This patch first simplifies storage and use of the data returned from an IPMI Get Device ID command. It then creates a new per-OEM handler hook, which should know how to handle events with the OEM[012] Data Available flags set. It then uses this to implement a workaround for IPMI 1.5-capable Dell PowerEdge servers which are susceptable to setting the OEM[012] Data Available flags when the driver can't handle it. Signed-off-by: Matt Domsch <Matt_Domsch@dell.com> Signed-off-by: Corey Minyard <minyard@acm.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
Corey Minyard · Linus Torvalds
1 parent 07766f241b
Showing 1 changed file with 105 additions and 12 deletions Side-by-side Diff
drivers/char/ipmi/ipmi_si_intf.c
@@ -110,6 +110,21 @@
     SI_KCS, SI_SMIC, SI_BT
 };
  
+struct ipmi_device_id {
+	unsigned char device_id;
+	unsigned char device_revision;
+	unsigned char firmware_revision_1;
+	unsigned char firmware_revision_2;
+	unsigned char ipmi_version;
+	unsigned char additional_device_support;
+	unsigned char manufacturer_id[3];
+	unsigned char product_id[2];
+	unsigned char aux_firmware_revision[4];
+} __attribute__((packed));
+
+#define ipmi_version_major(v) ((v)->ipmi_version & 0xf)
+#define ipmi_version_minor(v) ((v)->ipmi_version >> 4)
+
 struct smi_info
 {
 	ipmi_smi_t             intf;
  
@@ -132,12 +147,24 @@
 	void (*irq_cleanup)(struct smi_info *info);
 	unsigned int io_size;
  
+	/* Per-OEM handler, called from handle_flags().
+	   Returns 1 when handle_flags() needs to be re-run
+	   or 0 indicating it set si_state itself.
+	*/
+	int (*oem_data_avail_handler)(struct smi_info *smi_info);
+
 	/* Flags from the last GET_MSG_FLAGS command, used when an ATTN
 	   is set to hold the flags until we are done handling everything
 	   from the flags. */
 #define RECEIVE_MSG_AVAIL	0x01
 #define EVENT_MSG_BUFFER_FULL	0x02
 #define WDT_PRE_TIMEOUT_INT	0x08
+#define OEM0_DATA_AVAIL     0x20
+#define OEM1_DATA_AVAIL     0x40
+#define OEM2_DATA_AVAIL     0x80
+#define OEM_DATA_AVAIL      (OEM0_DATA_AVAIL | \
+                             OEM1_DATA_AVAIL | \
+                             OEM2_DATA_AVAIL)
 	unsigned char       msg_flags;
  
 	/* If set to true, this will request events the next time the
@@ -176,11 +203,7 @@
 	   interrupts. */
 	int interrupt_disabled;
  
-	unsigned char ipmi_si_dev_rev;
-	unsigned char ipmi_si_fw_rev_major;
-	unsigned char ipmi_si_fw_rev_minor;
-	unsigned char ipmi_version_major;
-	unsigned char ipmi_version_minor;
+	struct ipmi_device_id device_id;
  
 	/* Slave address, could be reported from DMI. */
 	unsigned char slave_addr;
@@ -323,6 +346,7 @@
  
 static void handle_flags(struct smi_info *smi_info)
 {
+ retry:
 	if (smi_info->msg_flags & WDT_PRE_TIMEOUT_INT) {
 		/* Watchdog pre-timeout */
 		spin_lock(&smi_info->count_lock);
@@ -372,6 +396,10 @@
 			smi_info->curr_msg->data,
 			smi_info->curr_msg->data_size);
 		smi_info->si_state = SI_GETTING_EVENTS;
+	} else if (smi_info->msg_flags & OEM_DATA_AVAIL) {
+		if (smi_info->oem_data_avail_handler)
+			if (smi_info->oem_data_avail_handler(smi_info))
+				goto retry;
 	} else {
 		smi_info->si_state = SI_NORMAL;
 	}
@@ -1927,11 +1955,8 @@
 	}
  
 	/* Record info from the get device id, in case we need it. */
-	smi_info->ipmi_si_dev_rev = resp[4] & 0xf;
-	smi_info->ipmi_si_fw_rev_major = resp[5] & 0x7f;
-	smi_info->ipmi_si_fw_rev_minor = resp[6];
-	smi_info->ipmi_version_major = resp[7] & 0xf;
-	smi_info->ipmi_version_minor = resp[7] >> 4;
+	memcpy(&smi_info->device_id, &resp[3],
+	       min_t(unsigned long, resp_len-3, sizeof(smi_info->device_id)));
  
  out:
 	kfree(resp);
@@ -1992,6 +2017,72 @@
 	return (out - ((char *) page));
 }
  
+/*
+ * oem_data_avail_to_receive_msg_avail
+ * @info - smi_info structure with msg_flags set
+ *
+ * Converts flags from OEM_DATA_AVAIL to RECEIVE_MSG_AVAIL
+ * Returns 1 indicating need to re-run handle_flags().
+ */
+static int oem_data_avail_to_receive_msg_avail(struct smi_info *smi_info)
+{
+	smi_info->msg_flags = (smi_info->msg_flags & ~OEM_DATA_AVAIL) |
+		RECEIVE_MSG_AVAIL;
+	return 1;
+}
+
+/*
+ * setup_dell_poweredge_oem_data_handler
+ * @info - smi_info.device_id must be populated
+ *
+ * Systems that match, but have firmware version < 1.40 may assert
+ * OEM0_DATA_AVAIL on their own, without being told via Set Flags that
+ * it's safe to do so.  Such systems will de-assert OEM1_DATA_AVAIL
+ * upon receipt of IPMI_GET_MSG_CMD, so we should treat these flags
+ * as RECEIVE_MSG_AVAIL instead.
+ *
+ * As Dell has no plans to release IPMI 1.5 firmware that *ever*
+ * assert the OEM[012] bits, and if it did, the driver would have to
+ * change to handle that properly, we don't actually check for the
+ * firmware version.
+ * Device ID = 0x20                BMC on PowerEdge 8G servers
+ * Device Revision = 0x80
+ * Firmware Revision1 = 0x01       BMC version 1.40
+ * Firmware Revision2 = 0x40       BCD encoded
+ * IPMI Version = 0x51             IPMI 1.5
+ * Manufacturer ID = A2 02 00      Dell IANA
+ *
+ */
+#define DELL_POWEREDGE_8G_BMC_DEVICE_ID  0x20
+#define DELL_POWEREDGE_8G_BMC_DEVICE_REV 0x80
+#define DELL_POWEREDGE_8G_BMC_IPMI_VERSION 0x51
+#define DELL_IANA_MFR_ID {0xA2, 0x02, 0x00}
+static void setup_dell_poweredge_oem_data_handler(struct smi_info *smi_info)
+{
+	struct ipmi_device_id *id = &smi_info->device_id;
+	const char mfr[3]=DELL_IANA_MFR_ID;
+	if (!memcmp(mfr, id->manufacturer_id, sizeof(mfr)) &&
+	    id->device_id       == DELL_POWEREDGE_8G_BMC_DEVICE_ID &&
+	    id->device_revision == DELL_POWEREDGE_8G_BMC_DEVICE_REV &&
+	    id->ipmi_version    == DELL_POWEREDGE_8G_BMC_IPMI_VERSION) {
+		smi_info->oem_data_avail_handler =
+			oem_data_avail_to_receive_msg_avail;
+	}
+}
+
+/*
+ * setup_oem_data_handler
+ * @info - smi_info.device_id must be filled in already
+ *
+ * Fills in smi_info.device_id.oem_data_available_handler
+ * when we know what function to use there.
+ */
+
+static void setup_oem_data_handler(struct smi_info *smi_info)
+{
+	setup_dell_poweredge_oem_data_handler(smi_info);
+}
+
 /* Returns 0 if initialized, or negative on an error. */
 static int init_one_smi(int intf_num, struct smi_info **smi)
 {
@@ -2090,6 +2181,8 @@
 	if (rv)
 		goto out_err;
  
+	setup_oem_data_handler(new_smi);
+
 	/* Try to claim any interrupts. */
 	new_smi->irq_setup(new_smi);
  
@@ -2123,8 +2216,8 @@
  
 	rv = ipmi_register_smi(&handlers,
 			       new_smi,
-			       new_smi->ipmi_version_major,
-			       new_smi->ipmi_version_minor,
+			       ipmi_version_major(&new_smi->device_id),
+			       ipmi_version_minor(&new_smi->device_id),
 			       new_smi->slave_addr,
 			       &(new_smi->intf));
 	if (rv) {
...	...	@@ -110,6 +110,21 @@
110	110	SI_KCS, SI_SMIC, SI_BT
111	111	};
112	112
	113	+struct ipmi_device_id {
	114	+ unsigned char device_id;
	115	+ unsigned char device_revision;
	116	+ unsigned char firmware_revision_1;
	117	+ unsigned char firmware_revision_2;
	118	+ unsigned char ipmi_version;
	119	+ unsigned char additional_device_support;
	120	+ unsigned char manufacturer_id[3];
	121	+ unsigned char product_id[2];
	122	+ unsigned char aux_firmware_revision[4];
	123	+} __attribute__((packed));
	124	+
	125	+#define ipmi_version_major(v) ((v)->ipmi_version & 0xf)
	126	+#define ipmi_version_minor(v) ((v)->ipmi_version >> 4)
	127	+
113	128	struct smi_info
114	129	{
115	130	ipmi_smi_t intf;
116	131
...	...	@@ -132,12 +147,24 @@
132	147	void (irq_cleanup)(struct smi_info info);
133	148	unsigned int io_size;
134	149
	150	+ /* Per-OEM handler, called from handle_flags().
	151	+ Returns 1 when handle_flags() needs to be re-run
	152	+ or 0 indicating it set si_state itself.
	153	+ */
	154	+ int (oem_data_avail_handler)(struct smi_info smi_info);
	155	+
135	156	/* Flags from the last GET_MSG_FLAGS command, used when an ATTN
136	157	is set to hold the flags until we are done handling everything
137	158	from the flags. */
138	159	#define RECEIVE_MSG_AVAIL 0x01
139	160	#define EVENT_MSG_BUFFER_FULL 0x02
140	161	#define WDT_PRE_TIMEOUT_INT 0x08
	162	+#define OEM0_DATA_AVAIL 0x20
	163	+#define OEM1_DATA_AVAIL 0x40
	164	+#define OEM2_DATA_AVAIL 0x80
	165	+#define OEM_DATA_AVAIL (OEM0_DATA_AVAIL \| \
	166	+ OEM1_DATA_AVAIL \| \
	167	+ OEM2_DATA_AVAIL)
141	168	unsigned char msg_flags;
142	169
143	170	/* If set to true, this will request events the next time the
...	...	@@ -176,11 +203,7 @@
176	203	interrupts. */
177	204	int interrupt_disabled;
178	205
179		- unsigned char ipmi_si_dev_rev;
180		- unsigned char ipmi_si_fw_rev_major;
181		- unsigned char ipmi_si_fw_rev_minor;
182		- unsigned char ipmi_version_major;
183		- unsigned char ipmi_version_minor;
	206	+ struct ipmi_device_id device_id;
184	207
185	208	/* Slave address, could be reported from DMI. */
186	209	unsigned char slave_addr;
...	...	@@ -323,6 +346,7 @@
323	346
324	347	static void handle_flags(struct smi_info *smi_info)
325	348	{
	349	+ retry:
326	350	if (smi_info->msg_flags & WDT_PRE_TIMEOUT_INT) {
327	351	/* Watchdog pre-timeout */
328	352	spin_lock(&smi_info->count_lock);
...	...	@@ -372,6 +396,10 @@
372	396	smi_info->curr_msg->data,
373	397	smi_info->curr_msg->data_size);
374	398	smi_info->si_state = SI_GETTING_EVENTS;
	399	+ } else if (smi_info->msg_flags & OEM_DATA_AVAIL) {
	400	+ if (smi_info->oem_data_avail_handler)
	401	+ if (smi_info->oem_data_avail_handler(smi_info))
	402	+ goto retry;
375	403	} else {
376	404	smi_info->si_state = SI_NORMAL;
377	405	}
...	...	@@ -1927,11 +1955,8 @@
1927	1955	}
1928	1956
1929	1957	/* Record info from the get device id, in case we need it. */
1930		- smi_info->ipmi_si_dev_rev = resp[4] & 0xf;
1931		- smi_info->ipmi_si_fw_rev_major = resp[5] & 0x7f;
1932		- smi_info->ipmi_si_fw_rev_minor = resp[6];
1933		- smi_info->ipmi_version_major = resp[7] & 0xf;
1934		- smi_info->ipmi_version_minor = resp[7] >> 4;
	1958	+ memcpy(&smi_info->device_id, &resp[3],
	1959	+ min_t(unsigned long, resp_len-3, sizeof(smi_info->device_id)));
1935	1960
1936	1961	out:
1937	1962	kfree(resp);
...	...	@@ -1992,6 +2017,72 @@
1992	2017	return (out - ((char *) page));
1993	2018	}
1994	2019
	2020	+/*
	2021	+ * oem_data_avail_to_receive_msg_avail
	2022	+ * @info - smi_info structure with msg_flags set
	2023	+ *
	2024	+ * Converts flags from OEM_DATA_AVAIL to RECEIVE_MSG_AVAIL
	2025	+ * Returns 1 indicating need to re-run handle_flags().
	2026	+ */
	2027	+static int oem_data_avail_to_receive_msg_avail(struct smi_info *smi_info)
	2028	+{
	2029	+ smi_info->msg_flags = (smi_info->msg_flags & ~OEM_DATA_AVAIL) \|
	2030	+ RECEIVE_MSG_AVAIL;
	2031	+ return 1;
	2032	+}
	2033	+
	2034	+/*
	2035	+ * setup_dell_poweredge_oem_data_handler
	2036	+ * @info - smi_info.device_id must be populated
	2037	+ *
	2038	+ * Systems that match, but have firmware version < 1.40 may assert
	2039	+ * OEM0_DATA_AVAIL on their own, without being told via Set Flags that
	2040	+ * it's safe to do so. Such systems will de-assert OEM1_DATA_AVAIL
	2041	+ * upon receipt of IPMI_GET_MSG_CMD, so we should treat these flags
	2042	+ * as RECEIVE_MSG_AVAIL instead.
	2043	+ *
	2044	+ * As Dell has no plans to release IPMI 1.5 firmware that ever
	2045	+ * assert the OEM[012] bits, and if it did, the driver would have to
	2046	+ * change to handle that properly, we don't actually check for the
	2047	+ * firmware version.
	2048	+ * Device ID = 0x20 BMC on PowerEdge 8G servers
	2049	+ * Device Revision = 0x80
	2050	+ * Firmware Revision1 = 0x01 BMC version 1.40
	2051	+ * Firmware Revision2 = 0x40 BCD encoded
	2052	+ * IPMI Version = 0x51 IPMI 1.5
	2053	+ * Manufacturer ID = A2 02 00 Dell IANA
	2054	+ *
	2055	+ */
	2056	+#define DELL_POWEREDGE_8G_BMC_DEVICE_ID 0x20
	2057	+#define DELL_POWEREDGE_8G_BMC_DEVICE_REV 0x80
	2058	+#define DELL_POWEREDGE_8G_BMC_IPMI_VERSION 0x51
	2059	+#define DELL_IANA_MFR_ID {0xA2, 0x02, 0x00}
	2060	+static void setup_dell_poweredge_oem_data_handler(struct smi_info *smi_info)
	2061	+{
	2062	+ struct ipmi_device_id *id = &smi_info->device_id;
	2063	+ const char mfr[3]=DELL_IANA_MFR_ID;
	2064	+ if (!memcmp(mfr, id->manufacturer_id, sizeof(mfr)) &&
	2065	+ id->device_id == DELL_POWEREDGE_8G_BMC_DEVICE_ID &&
	2066	+ id->device_revision == DELL_POWEREDGE_8G_BMC_DEVICE_REV &&
	2067	+ id->ipmi_version == DELL_POWEREDGE_8G_BMC_IPMI_VERSION) {
	2068	+ smi_info->oem_data_avail_handler =
	2069	+ oem_data_avail_to_receive_msg_avail;
	2070	+ }
	2071	+}
	2072	+
	2073	+/*
	2074	+ * setup_oem_data_handler
	2075	+ * @info - smi_info.device_id must be filled in already
	2076	+ *
	2077	+ * Fills in smi_info.device_id.oem_data_available_handler
	2078	+ * when we know what function to use there.
	2079	+ */
	2080	+
	2081	+static void setup_oem_data_handler(struct smi_info *smi_info)
	2082	+{
	2083	+ setup_dell_poweredge_oem_data_handler(smi_info);
	2084	+}
	2085	+
1995	2086	/* Returns 0 if initialized, or negative on an error. */
1996	2087	static int init_one_smi(int intf_num, struct smi_info **smi)
1997	2088	{
...	...	@@ -2090,6 +2181,8 @@
2090	2181	if (rv)
2091	2182	goto out_err;
2092	2183
	2184	+ setup_oem_data_handler(new_smi);
	2185	+
2093	2186	/* Try to claim any interrupts. */
2094	2187	new_smi->irq_setup(new_smi);
2095	2188
...	...	@@ -2123,8 +2216,8 @@
2123	2216
2124	2217	rv = ipmi_register_smi(&handlers,
2125	2218	new_smi,
2126		- new_smi->ipmi_version_major,
2127		- new_smi->ipmi_version_minor,
	2219	+ ipmi_version_major(&new_smi->device_id),
	2220	+ ipmi_version_minor(&new_smi->device_id),
2128	2221	new_smi->slave_addr,
2129	2222	&(new_smi->intf));
2130	2223	if (rv) {