Firmware: add iSCSI iBFT Support
authorKonrad Rzeszutek <ketuzsezr@darnok.org>
Thu, 10 Apr 2008 02:50:41 +0000 (19:50 -0700)
committerGreg Kroah-Hartman <gregkh@suse.de>
Sun, 20 Apr 2008 02:10:28 +0000 (19:10 -0700)
Add /sysfs/firmware/ibft/[initiator|targetX|ethernetX] directories along with
text properties which export the the iSCSI Boot Firmware Table (iBFT)
structure.

What is iSCSI Boot Firmware Table?  It is a mechanism for the iSCSI tools to
extract from the machine NICs the iSCSI connection information so that they
can automagically mount the iSCSI share/target.  Currently the iSCSI
information is hard-coded in the initrd.  The /sysfs entries are read-only
one-name-and-value fields.

The usual set of data exposed is:

# for a in `find /sys/firmware/ibft/ -type f -print`; do  echo -n "$a: ";  cat $a; done
/sys/firmware/ibft/target0/target-name: iqn.2007.com.intel-sbx44:storage-10gb
/sys/firmware/ibft/target0/nic-assoc: 0
/sys/firmware/ibft/target0/chap-type: 0
/sys/firmware/ibft/target0/lun: 00000000
/sys/firmware/ibft/target0/port: 3260
/sys/firmware/ibft/target0/ip-addr: 192.168.79.116
/sys/firmware/ibft/target0/flags: 3
/sys/firmware/ibft/target0/index: 0
/sys/firmware/ibft/ethernet0/mac: 00:11:25:9d:8b:01
/sys/firmware/ibft/ethernet0/vlan: 0
/sys/firmware/ibft/ethernet0/gateway: 192.168.79.254
/sys/firmware/ibft/ethernet0/origin: 0
/sys/firmware/ibft/ethernet0/subnet-mask: 255.255.252.0
/sys/firmware/ibft/ethernet0/ip-addr: 192.168.77.41
/sys/firmware/ibft/ethernet0/flags: 7
/sys/firmware/ibft/ethernet0/index: 0
/sys/firmware/ibft/initiator/initiator-name: iqn.2007-07.com:konrad.initiator
/sys/firmware/ibft/initiator/flags: 3
/sys/firmware/ibft/initiator/index: 0

For full details of the IBFT structure please take a look at:
ftp://ftp.software.ibm.com/systems/support/system_x_pdf/ibm_iscsi_boot_firmware_table_v1.02.pdf

[akpm@linux-foundation.org: fix build]
Signed-off-by: Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
Cc: Mike Christie <michaelc@cs.wisc.edu>
Cc: Peter Jones <pjones@redhat.com>
Cc: James Bottomley <James.Bottomley@HansenPartnership.com>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Thomas Gleixner <tglx@linutronix.de>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Documentation/ABI/testing/sysfs-ibft [new file with mode: 0644]
arch/x86/kernel/setup_32.c
arch/x86/kernel/setup_64.c
drivers/firmware/Kconfig
drivers/firmware/Makefile
drivers/firmware/iscsi_ibft.c [new file with mode: 0644]
drivers/firmware/iscsi_ibft_find.c [new file with mode: 0644]
include/linux/iscsi_ibft.h [new file with mode: 0644]

diff --git a/Documentation/ABI/testing/sysfs-ibft b/Documentation/ABI/testing/sysfs-ibft
new file mode 100644 (file)
index 0000000..c2b7d11
--- /dev/null
@@ -0,0 +1,23 @@
+What:          /sys/firmware/ibft/initiator
+Date:          November 2007
+Contact:       Konrad Rzeszutek <ketuzsezr@darnok.org>
+Description:   The /sys/firmware/ibft/initiator directory will contain
+               files that expose the iSCSI Boot Firmware Table initiator data.
+               Usually this contains the Initiator name.
+
+What:          /sys/firmware/ibft/targetX
+Date:          November 2007
+Contact:       Konrad Rzeszutek <ketuzsezr@darnok.org>
+Description:   The /sys/firmware/ibft/targetX directory will contain
+               files that expose the iSCSI Boot Firmware Table target data.
+               Usually this contains the target's IP address, boot LUN,
+               target name, and what NIC it is associated with. It can also
+               contain the CHAP name (and password), the reverse CHAP
+               name (and password)
+
+What:          /sys/firmware/ibft/ethernetX
+Date:          November 2007
+Contact:       Konrad Rzeszutek <ketuzsezr@darnok.org>
+Description:   The /sys/firmware/ibft/ethernetX directory will contain
+               files that expose the iSCSI Boot Firmware Table NIC data.
+               This can this can the IP address, MAC, and gateway of the NIC.
index 5b0bffb7fcc91d9aa237367a396dd1e5321aba6c..4ef91749959ef239f6bbf419a62aee92e9643b5b 100644 (file)
@@ -39,6 +39,7 @@
 #include <linux/efi.h>
 #include <linux/init.h>
 #include <linux/edd.h>
+#include <linux/iscsi_ibft.h>
 #include <linux/nodemask.h>
 #include <linux/kexec.h>
 #include <linux/crash_dump.h>
@@ -689,6 +690,8 @@ void __init setup_bootmem_allocator(void)
 #endif
        numa_kva_reserve();
        reserve_crashkernel();
+
+       reserve_ibft_region();
 }
 
 /*
index 674ef3510cdfd66972a08f716e374a0d96884dea..216c93bd9993af70ead09aaa155d865a428f907e 100644 (file)
@@ -33,6 +33,7 @@
 #include <linux/acpi.h>
 #include <linux/kallsyms.h>
 #include <linux/edd.h>
+#include <linux/iscsi_ibft.h>
 #include <linux/mmzone.h>
 #include <linux/kexec.h>
 #include <linux/cpufreq.h>
@@ -434,6 +435,9 @@ void __init setup_arch(char **cmdline_p)
        }
 #endif
        reserve_crashkernel();
+
+       reserve_ibft_region();
+
        paging_init();
        map_vsyscall();
 
index 05f02a326f1c6283277964f307873df9cb5a5489..40ffd767647d623baa1259618b8912ef8ecc863d 100644 (file)
@@ -93,4 +93,24 @@ config DMIID
          information from userspace through /sys/class/dmi/id/ or if you want
          DMI-based module auto-loading.
 
+config ISCSI_IBFT_FIND
+       bool "iSCSI Boot Firmware Table Attributes"
+       depends on X86
+       default n
+       help
+         This option enables the kernel to find the region of memory
+         in which the ISCSI Boot Firmware Table (iBFT) resides. This
+         is necessary for iSCSI Boot Firmware Table Attributes module to work
+         properly.
+
+config ISCSI_IBFT
+       tristate "iSCSI Boot Firmware Table Attributes module"
+       depends on ISCSI_IBFT_FIND
+       default n
+       help
+         This option enables support for detection and exposing of iSCSI
+         Boot Firmware Table (iBFT) via sysfs to userspace. If you wish to
+         detect iSCSI boot parameters dynamically during system boot, say Y.
+         Otherwise, say N.
+
 endmenu
index 8d4ebc805a505cae9ff188fc24604e88e058be60..4c9147154df891d0659cd1f68582a9f6fca4a196 100644 (file)
@@ -8,3 +8,5 @@ obj-$(CONFIG_EFI_PCDP)          += pcdp.o
 obj-$(CONFIG_DELL_RBU)          += dell_rbu.o
 obj-$(CONFIG_DCDBAS)           += dcdbas.o
 obj-$(CONFIG_DMIID)            += dmi-id.o
+obj-$(CONFIG_ISCSI_IBFT_FIND)  += iscsi_ibft_find.o
+obj-$(CONFIG_ISCSI_IBFT)       += iscsi_ibft.o
diff --git a/drivers/firmware/iscsi_ibft.c b/drivers/firmware/iscsi_ibft.c
new file mode 100644 (file)
index 0000000..8024e3b
--- /dev/null
@@ -0,0 +1,982 @@
+/*
+ *  Copyright 2007 Red Hat, Inc.
+ *  by Peter Jones <pjones@redhat.com>
+ *  Copyright 2008 IBM, Inc.
+ *  by Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *  Copyright 2008
+ *  by Konrad Rzeszutek <ketuzsezr@darnok.org>
+ *
+ * This code exposes the iSCSI Boot Format Table to userland via sysfs.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License v2.0 as published by
+ * the Free Software Foundation
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * Changelog:
+ *
+ *  14 Mar 2008 - Konrad Rzeszutek <ketuzsezr@darnok.org>
+ *    Updated comments and copyrights. (v0.4.9)
+ *
+ *  11 Feb 2008 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *    Converted to using ibft_addr. (v0.4.8)
+ *
+ *   8 Feb 2008 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *    Combined two functions in one: reserve_ibft_region. (v0.4.7)
+ *
+ *  30 Jan 2008 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   Added logic to handle IPv6 addresses. (v0.4.6)
+ *
+ *  25 Jan 2008 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   Added logic to handle badly not-to-spec iBFT. (v0.4.5)
+ *
+ *   4 Jan 2008 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   Added __init to function declarations. (v0.4.4)
+ *
+ *  21 Dec 2007 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   Updated kobject registration, combined unregister functions in one
+ *   and code and style cleanup. (v0.4.3)
+ *
+ *   5 Dec 2007 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   Added end-markers to enums and re-organized kobject registration. (v0.4.2)
+ *
+ *   4 Dec 2007 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   Created 'device' sysfs link to the NIC and style cleanup. (v0.4.1)
+ *
+ *  28 Nov 2007 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   Added sysfs-ibft documentation, moved 'find_ibft' function to
+ *   in its own file and added text attributes for every struct field.  (v0.4)
+ *
+ *  21 Nov 2007 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   Added text attributes emulating OpenFirmware /proc/device-tree naming.
+ *   Removed binary /sysfs interface (v0.3)
+ *
+ *  29 Aug 2007 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   Added functionality in setup.c to reserve iBFT region. (v0.2)
+ *
+ *  27 Aug 2007 - Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *   First version exposing iBFT data via a binary /sysfs. (v0.1)
+ *
+ */
+
+
+#include <linux/blkdev.h>
+#include <linux/capability.h>
+#include <linux/ctype.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/iscsi_ibft.h>
+#include <linux/limits.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/slab.h>
+#include <linux/stat.h>
+#include <linux/string.h>
+#include <linux/types.h>
+
+#define IBFT_ISCSI_VERSION "0.4.9"
+#define IBFT_ISCSI_DATE "2008-Mar-14"
+
+MODULE_AUTHOR("Peter Jones <pjones@redhat.com> and \
+Konrad Rzeszutek <ketuzsezr@darnok.org>");
+MODULE_DESCRIPTION("sysfs interface to BIOS iBFT information");
+MODULE_LICENSE("GPL");
+MODULE_VERSION(IBFT_ISCSI_VERSION);
+
+struct ibft_hdr {
+       u8 id;
+       u8 version;
+       u16 length;
+       u8 index;
+       u8 flags;
+} __attribute__((__packed__));
+
+struct ibft_control {
+       struct ibft_hdr hdr;
+       u16 extensions;
+       u16 initiator_off;
+       u16 nic0_off;
+       u16 tgt0_off;
+       u16 nic1_off;
+       u16 tgt1_off;
+} __attribute__((__packed__));
+
+struct ibft_initiator {
+       struct ibft_hdr hdr;
+       char isns_server[16];
+       char slp_server[16];
+       char pri_radius_server[16];
+       char sec_radius_server[16];
+       u16 initiator_name_len;
+       u16 initiator_name_off;
+} __attribute__((__packed__));
+
+struct ibft_nic {
+       struct ibft_hdr hdr;
+       char ip_addr[16];
+       u8 subnet_mask_prefix;
+       u8 origin;
+       char gateway[16];
+       char primary_dns[16];
+       char secondary_dns[16];
+       char dhcp[16];
+       u16 vlan;
+       char mac[6];
+       u16 pci_bdf;
+       u16 hostname_len;
+       u16 hostname_off;
+} __attribute__((__packed__));
+
+struct ibft_tgt {
+       struct ibft_hdr hdr;
+       char ip_addr[16];
+       u16 port;
+       char lun[8];
+       u8 chap_type;
+       u8 nic_assoc;
+       u16 tgt_name_len;
+       u16 tgt_name_off;
+       u16 chap_name_len;
+       u16 chap_name_off;
+       u16 chap_secret_len;
+       u16 chap_secret_off;
+       u16 rev_chap_name_len;
+       u16 rev_chap_name_off;
+       u16 rev_chap_secret_len;
+       u16 rev_chap_secret_off;
+} __attribute__((__packed__));
+
+/*
+ * The kobject different types and its names.
+ *
+*/
+enum ibft_id {
+       id_reserved = 0, /* We don't support. */
+       id_control = 1, /* Should show up only once and is not exported. */
+       id_initiator = 2,
+       id_nic = 3,
+       id_target = 4,
+       id_extensions = 5, /* We don't support. */
+       id_end_marker,
+};
+
+/*
+ * We do not support the other types, hence the usage of NULL.
+ * This maps to the enum ibft_id.
+ */
+static const char *ibft_id_names[] =
+       {NULL, NULL, "initiator", "ethernet%d", "target%d", NULL, NULL};
+
+/*
+ * The text attributes names for each of the kobjects.
+*/
+enum ibft_eth_properties_enum {
+       ibft_eth_index,
+       ibft_eth_flags,
+       ibft_eth_ip_addr,
+       ibft_eth_subnet_mask,
+       ibft_eth_origin,
+       ibft_eth_gateway,
+       ibft_eth_primary_dns,
+       ibft_eth_secondary_dns,
+       ibft_eth_dhcp,
+       ibft_eth_vlan,
+       ibft_eth_mac,
+       /* ibft_eth_pci_bdf - this is replaced by link to the device itself. */
+       ibft_eth_hostname,
+       ibft_eth_end_marker,
+};
+
+static const char *ibft_eth_properties[] =
+       {"index", "flags", "ip-addr", "subnet-mask", "origin", "gateway",
+       "primary-dns", "secondary-dns", "dhcp", "vlan", "mac", "hostname",
+       NULL};
+
+enum ibft_tgt_properties_enum {
+       ibft_tgt_index,
+       ibft_tgt_flags,
+       ibft_tgt_ip_addr,
+       ibft_tgt_port,
+       ibft_tgt_lun,
+       ibft_tgt_chap_type,
+       ibft_tgt_nic_assoc,
+       ibft_tgt_name,
+       ibft_tgt_chap_name,
+       ibft_tgt_chap_secret,
+       ibft_tgt_rev_chap_name,
+       ibft_tgt_rev_chap_secret,
+       ibft_tgt_end_marker,
+};
+
+static const char *ibft_tgt_properties[] =
+       {"index", "flags", "ip-addr", "port", "lun", "chap-type", "nic-assoc",
+       "target-name", "chap-name", "chap-secret", "rev-chap-name",
+       "rev-chap-name-secret", NULL};
+
+enum ibft_initiator_properties_enum {
+       ibft_init_index,
+       ibft_init_flags,
+       ibft_init_isns_server,
+       ibft_init_slp_server,
+       ibft_init_pri_radius_server,
+       ibft_init_sec_radius_server,
+       ibft_init_initiator_name,
+       ibft_init_end_marker,
+};
+
+static const char *ibft_initiator_properties[] =
+       {"index", "flags", "isns-server", "slp-server", "pri-radius-server",
+       "sec-radius-server", "initiator-name", NULL};
+
+/*
+ * The kobject and attribute structures.
+ */
+
+struct ibft_kobject {
+       struct ibft_table_header *header;
+       union {
+               struct ibft_initiator *initiator;
+               struct ibft_nic *nic;
+               struct ibft_tgt *tgt;
+               struct ibft_hdr *hdr;
+       };
+       struct kobject kobj;
+       struct list_head node;
+};
+
+struct ibft_attribute {
+       struct attribute attr;
+       ssize_t (*show) (struct  ibft_kobject *entry,
+                        struct ibft_attribute *attr, char *buf);
+       union {
+               struct ibft_initiator *initiator;
+               struct ibft_nic *nic;
+               struct ibft_tgt *tgt;
+               struct ibft_hdr *hdr;
+       };
+       struct kobject *kobj;
+       int type; /* The enum of the type. This can be any value of:
+               ibft_eth_properties_enum, ibft_tgt_properties_enum,
+               or ibft_initiator_properties_enum. */
+       struct list_head node;
+};
+
+static LIST_HEAD(ibft_attr_list);
+static LIST_HEAD(ibft_kobject_list);
+
+static const char nulls[16];
+
+/*
+ * Helper functions to parse data properly.
+ */
+static ssize_t sprintf_ipaddr(char *buf, u8 *ip)
+{
+       char *str = buf;
+
+       if (ip[0] == 0 && ip[1] == 0 && ip[2] == 0 && ip[3] == 0 &&
+           ip[4] == 0 && ip[5] == 0 && ip[6] == 0 && ip[7] == 0 &&
+           ip[8] == 0 && ip[9] == 0 && ip[10] == 0xff && ip[11] == 0xff) {
+               /*
+                * IPV4
+                */
+               str += sprintf(buf, NIPQUAD_FMT, ip[12],
+                              ip[13], ip[14], ip[15]);
+       } else {
+               /*
+                * IPv6
+                */
+               str += sprintf(str, NIP6_FMT, ntohs(ip[0]), ntohs(ip[1]),
+                              ntohs(ip[2]), ntohs(ip[3]), ntohs(ip[4]),
+                              ntohs(ip[5]), ntohs(ip[6]), ntohs(ip[7]));
+       }
+       str += sprintf(str, "\n");
+       return str - buf;
+}
+
+static ssize_t sprintf_string(char *str, int len, char *buf)
+{
+       return sprintf(str, "%.*s\n", len, buf);
+}
+
+/*
+ * Helper function to verify the IBFT header.
+ */
+static int ibft_verify_hdr(char *t, struct ibft_hdr *hdr, int id, int length)
+{
+       if (hdr->id != id) {
+               printk(KERN_ERR "iBFT error: We expected the " \
+                               "field header.id to have %d but " \
+                               "found %d instead!\n", id, hdr->id);
+               return -ENODEV;
+       }
+       if (hdr->length != length) {
+               printk(KERN_ERR "iBFT error: We expected the " \
+                               "field header.length to have %d but " \
+                               "found %d instead!\n", length, hdr->length);
+               return -ENODEV;
+       }
+
+       return 0;
+}
+
+static void ibft_release(struct kobject *kobj)
+{
+       struct ibft_kobject *ibft =
+               container_of(kobj, struct ibft_kobject, kobj);
+       kfree(ibft);
+}
+
+/*
+ *  Routines for parsing the iBFT data to be human readable.
+ */
+ssize_t ibft_attr_show_initiator(struct ibft_kobject *entry,
+                                 struct ibft_attribute *attr,
+                                 char *buf)
+{
+       struct ibft_initiator *initiator = entry->initiator;
+       void *ibft_loc = entry->header;
+       char *str = buf;
+
+       if (!initiator)
+               return 0;
+
+       switch (attr->type) {
+       case ibft_init_index:
+               str += sprintf(str, "%d\n", initiator->hdr.index);
+               break;
+       case ibft_init_flags:
+               str += sprintf(str, "%d\n", initiator->hdr.flags);
+               break;
+       case ibft_init_isns_server:
+               str += sprintf_ipaddr(str, initiator->isns_server);
+               break;
+       case ibft_init_slp_server:
+               str += sprintf_ipaddr(str, initiator->slp_server);
+               break;
+       case ibft_init_pri_radius_server:
+               str += sprintf_ipaddr(str, initiator->pri_radius_server);
+               break;
+       case ibft_init_sec_radius_server:
+               str += sprintf_ipaddr(str, initiator->sec_radius_server);
+               break;
+       case ibft_init_initiator_name:
+               str += sprintf_string(str, initiator->initiator_name_len,
+                                     (char *)ibft_loc +
+                                     initiator->initiator_name_off);
+               break;
+       default:
+               break;
+       }
+
+       return str - buf;
+}
+
+ssize_t ibft_attr_show_nic(struct ibft_kobject *entry,
+                           struct ibft_attribute *attr,
+                           char *buf)
+{
+       struct ibft_nic *nic = entry->nic;
+       void *ibft_loc = entry->header;
+       char *str = buf;
+       char *mac;
+       int val;
+
+       if (!nic)
+               return 0;
+
+       switch (attr->type) {
+       case ibft_eth_index:
+               str += sprintf(str, "%d\n", nic->hdr.index);
+               break;
+       case ibft_eth_flags:
+               str += sprintf(str, "%d\n", nic->hdr.flags);
+               break;
+       case ibft_eth_ip_addr:
+               str += sprintf_ipaddr(str, nic->ip_addr);
+               break;
+       case ibft_eth_subnet_mask:
+               val = ~((1 << (32-nic->subnet_mask_prefix))-1);
+               str += sprintf(str, NIPQUAD_FMT,
+                              (u8)(val >> 24), (u8)(val >> 16),
+                              (u8)(val >> 8), (u8)(val));
+               break;
+       case ibft_eth_origin:
+               str += sprintf(str, "%d\n", nic->origin);
+               break;
+       case ibft_eth_gateway:
+               str += sprintf_ipaddr(str, nic->gateway);
+               break;
+       case ibft_eth_primary_dns:
+               str += sprintf_ipaddr(str, nic->primary_dns);
+               break;
+       case ibft_eth_secondary_dns:
+               str += sprintf_ipaddr(str, nic->secondary_dns);
+               break;
+       case ibft_eth_dhcp:
+               str += sprintf_ipaddr(str, nic->dhcp);
+               break;
+       case ibft_eth_vlan:
+               str += sprintf(str, "%d\n", nic->vlan);
+               break;
+       case ibft_eth_mac:
+               mac = nic->mac;
+               str += sprintf(str, "%02x:%02x:%02x:%02x:%02x:%02x\n",
+                              (u8)mac[0], (u8)mac[1], (u8)mac[2],
+                              (u8)mac[3], (u8)mac[4], (u8)mac[5]);
+               break;
+       case ibft_eth_hostname:
+               str += sprintf_string(str, nic->hostname_len,
+                                     (char *)ibft_loc + nic->hostname_off);
+               break;
+       default:
+               break;
+       }
+
+       return str - buf;
+};
+
+ssize_t ibft_attr_show_target(struct ibft_kobject *entry,
+                              struct ibft_attribute *attr,
+                              char *buf)
+{
+       struct ibft_tgt *tgt = entry->tgt;
+       void *ibft_loc = entry->header;
+       char *str = buf;
+       int i;
+
+       if (!tgt)
+               return 0;
+
+       switch (attr->type) {
+       case ibft_tgt_index:
+               str += sprintf(str, "%d\n", tgt->hdr.index);
+               break;
+       case ibft_tgt_flags:
+               str += sprintf(str, "%d\n", tgt->hdr.flags);
+               break;
+       case ibft_tgt_ip_addr:
+               str += sprintf_ipaddr(str, tgt->ip_addr);
+               break;
+       case ibft_tgt_port:
+               str += sprintf(str, "%d\n", tgt->port);
+               break;
+       case ibft_tgt_lun:
+               for (i = 0; i < 8; i++)
+                       str += sprintf(str, "%x", (u8)tgt->lun[i]);
+               str += sprintf(str, "\n");
+               break;
+       case ibft_tgt_nic_assoc:
+               str += sprintf(str, "%d\n", tgt->nic_assoc);
+               break;
+       case ibft_tgt_chap_type:
+               str += sprintf(str, "%d\n", tgt->chap_type);
+               break;
+       case ibft_tgt_name:
+               str += sprintf_string(str, tgt->tgt_name_len,
+                                     (char *)ibft_loc + tgt->tgt_name_off);
+               break;
+       case ibft_tgt_chap_name:
+               str += sprintf_string(str, tgt->chap_name_len,
+                                     (char *)ibft_loc + tgt->chap_name_off);
+               break;
+       case ibft_tgt_chap_secret:
+               str += sprintf_string(str, tgt->chap_secret_len,
+                                     (char *)ibft_loc + tgt->chap_secret_off);
+               break;
+       case ibft_tgt_rev_chap_name:
+               str += sprintf_string(str, tgt->rev_chap_name_len,
+                                     (char *)ibft_loc +
+                                     tgt->rev_chap_name_off);
+               break;
+       case ibft_tgt_rev_chap_secret:
+               str += sprintf_string(str, tgt->rev_chap_secret_len,
+                                     (char *)ibft_loc +
+                                     tgt->rev_chap_secret_off);
+               break;
+       default:
+               break;
+       }
+
+       return str - buf;
+}
+
+/*
+ * The routine called for all sysfs attributes.
+ */
+static ssize_t ibft_show_attribute(struct kobject *kobj,
+                                   struct attribute *attr,
+                                   char *buf)
+{
+       struct ibft_kobject *dev =
+               container_of(kobj, struct ibft_kobject, kobj);
+       struct ibft_attribute *ibft_attr =
+               container_of(attr, struct ibft_attribute, attr);
+       ssize_t ret = -EIO;
+       char *str = buf;
+
+       if (!capable(CAP_SYS_ADMIN))
+               return -EACCES;
+
+       if (ibft_attr->show)
+               ret = ibft_attr->show(dev, ibft_attr, str);
+
+       return ret;
+}
+
+static struct sysfs_ops ibft_attr_ops = {
+       .show = ibft_show_attribute,
+};
+
+static struct kobj_type ibft_ktype = {
+       .release = ibft_release,
+       .sysfs_ops = &ibft_attr_ops,
+};
+
+static struct kset *ibft_kset;
+
+static int __init ibft_check_device(void)
+{
+       int len;
+       u8 *pos;
+       u8 csum = 0;
+
+       len = ibft_addr->length;
+
+       /* Sanity checking of iBFT. */
+       if (ibft_addr->revision != 1) {
+               printk(KERN_ERR "iBFT module supports only revision 1, " \
+                               "while this is %d.\n", ibft_addr->revision);
+               return -ENOENT;
+       }
+       for (pos = (u8 *)ibft_addr; pos < (u8 *)ibft_addr + len; pos++)
+               csum += *pos;
+
+       if (csum) {
+               printk(KERN_ERR "iBFT has incorrect checksum (0x%x)!\n", csum);
+               return -ENOENT;
+       }
+
+       return 0;
+}
+
+/*
+ * Helper function for ibft_register_kobjects.
+ */
+static int __init ibft_create_kobject(struct ibft_table_header *header,
+                                      struct ibft_hdr *hdr,
+                                      struct list_head *list)
+{
+       struct ibft_kobject *ibft_kobj = NULL;
+       struct ibft_nic *nic = (struct ibft_nic *)hdr;
+       struct pci_dev *pci_dev;
+       int rc = 0;
+
+       ibft_kobj = kzalloc(sizeof(*ibft_kobj), GFP_KERNEL);
+       if (!ibft_kobj)
+               return -ENOMEM;
+
+       ibft_kobj->header = header;
+       ibft_kobj->hdr = hdr;
+
+       switch (hdr->id) {
+       case id_initiator:
+               rc = ibft_verify_hdr("initiator", hdr, id_initiator,
+                                    sizeof(*ibft_kobj->initiator));
+               break;
+       case id_nic:
+               rc = ibft_verify_hdr("ethernet", hdr, id_nic,
+                                    sizeof(*ibft_kobj->nic));
+               break;
+       case id_target:
+               rc = ibft_verify_hdr("target", hdr, id_target,
+                                    sizeof(*ibft_kobj->tgt));
+               break;
+       case id_reserved:
+       case id_control:
+       case id_extensions:
+               /* Fields which we don't support. Ignore them */
+               rc = 1;
+               break;
+       default:
+               printk(KERN_ERR "iBFT has unknown structure type (%d). " \
+                               "Report this bug to %.6s!\n", hdr->id,
+                               header->oem_id);
+               rc = 1;
+               break;
+       }
+
+       if (rc) {
+               /* Skip adding this kobject, but exit with non-fatal error. */
+               kfree(ibft_kobj);
+               goto out_invalid_struct;
+       }
+
+       ibft_kobj->kobj.kset = ibft_kset;
+
+       rc = kobject_init_and_add(&ibft_kobj->kobj, &ibft_ktype,
+                                 NULL, ibft_id_names[hdr->id], hdr->index);
+
+       if (rc) {
+               kfree(ibft_kobj);
+               goto out;
+       }
+
+       kobject_uevent(&ibft_kobj->kobj, KOBJ_ADD);
+
+       if (hdr->id == id_nic) {
+               /*
+               * We don't search for the device in other domains than
+               * zero. This is because on x86 platforms the BIOS
+               * executes only devices which are in domain 0. Furthermore, the
+               * iBFT spec doesn't have a domain id field :-(
+               */
+               pci_dev = pci_get_bus_and_slot((nic->pci_bdf & 0xff00) >> 8,
+                                              (nic->pci_bdf & 0xff));
+               if (pci_dev) {
+                       rc = sysfs_create_link(&ibft_kobj->kobj,
+                                              &pci_dev->dev.kobj, "device");
+                       pci_dev_put(pci_dev);
+               }
+       }
+
+       /* Nothing broke so lets add it to the list. */
+       list_add_tail(&ibft_kobj->node, list);
+out:
+       return rc;
+out_invalid_struct:
+       /* Unsupported structs are skipped. */
+       return 0;
+}
+
+/*
+ * Scan the IBFT table structure for the NIC and Target fields. When
+ * found add them on the passed-in list. We do not support the other
+ * fields at this point, so they are skipped.
+ */
+static int __init ibft_register_kobjects(struct ibft_table_header *header,
+                                         struct list_head *list)
+{
+       struct ibft_control *control = NULL;
+       void *ptr, *end;
+       int rc = 0;
+       u16 offset;
+       u16 eot_offset;
+
+       control = (void *)header + sizeof(*header);
+       end = (void *)control + control->hdr.length;
+       eot_offset = (void *)header + header->length -
+                    (void *)control - sizeof(*header);
+       rc = ibft_verify_hdr("control", (struct ibft_hdr *)control, id_control,
+                            sizeof(*control));
+
+       /* iBFT table safety checking */
+       rc |= ((control->hdr.index) ? -ENODEV : 0);
+       if (rc) {
+               printk(KERN_ERR "iBFT error: Control header is invalid!\n");
+               return rc;
+       }
+       for (ptr = &control->initiator_off; ptr < end; ptr += sizeof(u16)) {
+               offset = *(u16 *)ptr;
+               if (offset && offset < header->length && offset < eot_offset) {
+                       rc = ibft_create_kobject(header,
+                                                (void *)header + offset,
+                                                list);
+                       if (rc)
+                               break;
+               }
+       }
+
+       return rc;
+}
+
+static void ibft_unregister(struct list_head *attr_list,
+                            struct list_head *kobj_list)
+{
+       struct ibft_kobject *data = NULL, *n;
+       struct ibft_attribute *attr = NULL, *m;
+
+       list_for_each_entry_safe(attr, m, attr_list, node) {
+               sysfs_remove_file(attr->kobj, &attr->attr);
+               list_del(&attr->node);
+               kfree(attr);
+       };
+       list_del_init(attr_list);
+
+       list_for_each_entry_safe(data, n, kobj_list, node) {
+               list_del(&data->node);
+               if (data->hdr->id == id_nic)
+                       sysfs_remove_link(&data->kobj, "device");
+               kobject_put(&data->kobj);
+       };
+       list_del_init(kobj_list);
+}
+
+static int __init ibft_create_attribute(struct ibft_kobject *kobj_data,
+                                        int type,
+                                        const char *name,
+                                        ssize_t (*show)(struct ibft_kobject *,
+                                                        struct ibft_attribute*,
+                                                        char *buf),
+                                        struct list_head *list)
+{
+       struct ibft_attribute *attr = NULL;
+       struct ibft_hdr *hdr = kobj_data->hdr;
+
+       attr = kmalloc(sizeof(*attr), GFP_KERNEL);
+       if (!attr)
+               return -ENOMEM;
+
+       attr->attr.name = name;
+       attr->attr.mode = S_IRUSR;
+       attr->attr.owner = THIS_MODULE;
+
+       attr->hdr = hdr;
+       attr->show = show;
+       attr->kobj = &kobj_data->kobj;
+       attr->type = type;
+
+       list_add_tail(&attr->node, list);
+
+       return 0;
+}
+
+/*
+ * Helper routiners to check to determine if the entry is valid
+ * in the proper iBFT structure.
+ */
+static int __init ibft_check_nic_for(struct ibft_nic *nic, int entry)
+{
+       int rc = 0;
+
+       switch (entry) {
+       case ibft_eth_index:
+       case ibft_eth_flags:
+               rc = 1;
+               break;
+       case ibft_eth_ip_addr:
+               if (!memcmp(nic->dhcp, nulls, sizeof(nic->dhcp)))
+                       rc = 1;
+               break;
+       case ibft_eth_subnet_mask:
+               if (!memcmp(nic->dhcp, nulls, sizeof(nic->dhcp)))
+                       rc = 1;
+               break;
+       case ibft_eth_origin:
+               rc = 1;
+               break;
+       case ibft_eth_gateway:
+               if (memcmp(nic->gateway, nulls, sizeof(nic->gateway)))
+                       rc = 1;
+               break;
+       case ibft_eth_primary_dns:
+               if (memcmp(nic->primary_dns, nulls,
+                          sizeof(nic->primary_dns)))
+                       rc = 1;
+               break;
+       case ibft_eth_secondary_dns:
+               if (memcmp(nic->secondary_dns, nulls,
+                          sizeof(nic->secondary_dns)))
+                       rc = 1;
+               break;
+       case ibft_eth_dhcp:
+               if (memcmp(nic->dhcp, nulls, sizeof(nic->dhcp)))
+                       rc = 1;
+               break;
+       case ibft_eth_vlan:
+       case ibft_eth_mac:
+               rc = 1;
+               break;
+       case ibft_eth_hostname:
+               if (nic->hostname_off)
+                       rc = 1;
+               break;
+       default:
+               break;
+       }
+
+       return rc;
+}
+
+static int __init ibft_check_tgt_for(struct ibft_tgt *tgt, int entry)
+{
+       int rc = 0;
+
+       switch (entry) {
+       case ibft_tgt_index:
+       case ibft_tgt_flags:
+       case ibft_tgt_ip_addr:
+       case ibft_tgt_port:
+       case ibft_tgt_lun:
+       case ibft_tgt_nic_assoc:
+       case ibft_tgt_chap_type:
+               rc = 1;
+       case ibft_tgt_name:
+               if (tgt->tgt_name_len)
+                       rc = 1;
+               break;
+       case ibft_tgt_chap_name:
+       case ibft_tgt_chap_secret:
+               if (tgt->chap_name_len)
+                       rc = 1;
+               break;
+       case ibft_tgt_rev_chap_name:
+       case ibft_tgt_rev_chap_secret:
+               if (tgt->rev_chap_name_len)
+                       rc = 1;
+               break;
+       default:
+               break;
+       }
+
+       return rc;
+}
+
+static int __init ibft_check_initiator_for(struct ibft_initiator *init,
+                                           int entry)
+{
+       int rc = 0;
+
+       switch (entry) {
+       case ibft_init_index:
+       case ibft_init_flags:
+               rc = 1;
+               break;
+       case ibft_init_isns_server:
+               if (memcmp(init->isns_server, nulls,
+                          sizeof(init->isns_server)))
+                       rc = 1;
+               break;
+       case ibft_init_slp_server:
+               if (memcmp(init->slp_server, nulls,
+                          sizeof(init->slp_server)))
+                       rc = 1;
+               break;
+       case ibft_init_pri_radius_server:
+               if (memcmp(init->pri_radius_server, nulls,
+                          sizeof(init->pri_radius_server)))
+                       rc = 1;
+               break;
+       case ibft_init_sec_radius_server:
+               if (memcmp(init->sec_radius_server, nulls,
+                          sizeof(init->sec_radius_server)))
+                       rc = 1;
+               break;
+       case ibft_init_initiator_name:
+               if (init->initiator_name_len)
+                       rc = 1;
+               break;
+       default:
+               break;
+       }
+
+       return rc;
+}
+
+/*
+ *  Register the attributes for all of the kobjects.
+ */
+static int __init ibft_register_attributes(struct list_head *kobject_list,
+                                           struct list_head *attr_list)
+{
+       int rc = 0, i = 0;
+       struct ibft_kobject *data = NULL;
+       struct ibft_attribute *attr = NULL, *m;
+
+       list_for_each_entry(data, kobject_list, node) {
+               switch (data->hdr->id) {
+               case id_nic:
+                       for (i = 0; i < ibft_eth_end_marker && !rc; i++)
+                               if (ibft_check_nic_for(data->nic, i))
+                                       rc = ibft_create_attribute(data, i,
+                                               ibft_eth_properties[i],
+                                               ibft_attr_show_nic, attr_list);
+                       break;
+               case id_target:
+                       for (i = 0; i < ibft_tgt_end_marker && !rc; i++)
+                               if (ibft_check_tgt_for(data->tgt, i))
+                                       rc = ibft_create_attribute(data, i,
+                                               ibft_tgt_properties[i],
+                                               ibft_attr_show_target,
+                                               attr_list);
+                       break;
+               case id_initiator:
+                       for (i = 0; i < ibft_init_end_marker && !rc; i++)
+                               if (ibft_check_initiator_for(
+                                       data->initiator, i))
+                                       rc = ibft_create_attribute(data, i,
+                                               ibft_initiator_properties[i],
+                                               ibft_attr_show_initiator,
+                                               attr_list);
+                       break;
+               default:
+                       break;
+               }
+               if (rc)
+                       break;
+       }
+       list_for_each_entry_safe(attr, m, attr_list, node) {
+               rc = sysfs_create_file(attr->kobj, &attr->attr);
+               if (rc) {
+                       list_del(&attr->node);
+                       kfree(attr);
+                       break;
+               }
+       }
+
+       return rc;
+}
+
+/*
+ * ibft_init() - creates sysfs tree entries for the iBFT data.
+ */
+static int __init ibft_init(void)
+{
+       int rc = 0;
+
+       ibft_kset = kset_create_and_add("ibft", NULL, firmware_kobj);
+       if (!ibft_kset)
+               return -ENOMEM;
+
+       if (ibft_addr) {
+               printk(KERN_INFO "iBFT detected at 0x%lx.\n",
+                      virt_to_phys((void *)ibft_addr));
+
+               rc = ibft_check_device();
+               if (rc)
+                       goto out_firmware_unregister;
+
+               /* Scan the IBFT for data and register the kobjects. */
+               rc = ibft_register_kobjects(ibft_addr, &ibft_kobject_list);
+               if (rc)
+                       goto out_free;
+
+               /* Register the attributes */
+               rc = ibft_register_attributes(&ibft_kobject_list,
+                                             &ibft_attr_list);
+               if (rc)
+                       goto out_free;
+       } else
+               printk(KERN_INFO "No iBFT detected.\n");
+
+       return 0;
+
+out_free:
+       ibft_unregister(&ibft_attr_list, &ibft_kobject_list);
+out_firmware_unregister:
+       kset_unregister(ibft_kset);
+       return rc;
+}
+
+static void __exit ibft_exit(void)
+{
+       ibft_unregister(&ibft_attr_list, &ibft_kobject_list);
+       kset_unregister(ibft_kset);
+}
+
+module_init(ibft_init);
+module_exit(ibft_exit);
diff --git a/drivers/firmware/iscsi_ibft_find.c b/drivers/firmware/iscsi_ibft_find.c
new file mode 100644 (file)
index 0000000..d0e5fa4
--- /dev/null
@@ -0,0 +1,84 @@
+/*
+ *  Copyright 2007 Red Hat, Inc.
+ *  by Peter Jones <pjones@redhat.com>
+ *  Copyright 2007 IBM, Inc.
+ *  by Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *  Copyright 2008
+ *  by Konrad Rzeszutek <ketuzsezr@darnok.org>
+ *
+ * This code finds the iSCSI Boot Format Table.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License v2.0 as published by
+ * the Free Software Foundation
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#include <linux/bootmem.h>
+#include <linux/blkdev.h>
+#include <linux/ctype.h>
+#include <linux/device.h>
+#include <linux/err.h>
+#include <linux/init.h>
+#include <linux/limits.h>
+#include <linux/module.h>
+#include <linux/pci.h>
+#include <linux/slab.h>
+#include <linux/stat.h>
+#include <linux/string.h>
+#include <linux/types.h>
+
+#include <asm/mmzone.h>
+
+/*
+ * Physical location of iSCSI Boot Format Table.
+ */
+struct ibft_table_header *ibft_addr;
+EXPORT_SYMBOL_GPL(ibft_addr);
+
+#define IBFT_SIGN "iBFT"
+#define IBFT_SIGN_LEN 4
+#define IBFT_START 0x80000 /* 512kB */
+#define IBFT_END 0x100000 /* 1MB */
+#define VGA_MEM 0xA0000 /* VGA buffer */
+#define VGA_SIZE 0x20000 /* 128kB */
+
+
+/*
+ * Routine used to find the iSCSI Boot Format Table. The logical
+ * kernel address is set in the ibft_addr global variable.
+ */
+void __init reserve_ibft_region(void)
+{
+       unsigned long pos;
+       unsigned int len = 0;
+       void *virt;
+
+       ibft_addr = 0;
+
+       for (pos = IBFT_START; pos < IBFT_END; pos += 16) {
+               /* The table can't be inside the VGA BIOS reserved space,
+                * so skip that area */
+               if (pos == VGA_MEM)
+                       pos += VGA_SIZE;
+               virt = phys_to_virt(pos);
+               if (memcmp(virt, IBFT_SIGN, IBFT_SIGN_LEN) == 0) {
+                       unsigned long *addr =
+                           (unsigned long *)phys_to_virt(pos + 4);
+                       len = *addr;
+                       /* if the length of the table extends past 1M,
+                        * the table cannot be valid. */
+                       if (pos + len <= (IBFT_END-1)) {
+                               ibft_addr = (struct ibft_table_header *)virt;
+                               break;
+                       }
+               }
+       }
+       if (ibft_addr)
+               reserve_bootmem(pos, PAGE_ALIGN(len), BOOTMEM_DEFAULT);
+}
+EXPORT_SYMBOL_GPL(reserve_ibft_region);
diff --git a/include/linux/iscsi_ibft.h b/include/linux/iscsi_ibft.h
new file mode 100644 (file)
index 0000000..6092487
--- /dev/null
@@ -0,0 +1,50 @@
+/*
+ *  Copyright 2007 Red Hat, Inc.
+ *  by Peter Jones <pjones@redhat.com>
+ *  Copyright 2007 IBM, Inc.
+ *  by Konrad Rzeszutek <konradr@linux.vnet.ibm.com>
+ *  Copyright 2008
+ *  by Konrad Rzeszutek <ketuzsezr@darnok.org>
+ *
+ * This code exposes the iSCSI Boot Format Table to userland via sysfs.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License v2.0 as published by
+ * the Free Software Foundation
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ */
+
+#ifndef ISCSI_IBFT_H
+#define ISCSI_IBFT_H
+
+struct ibft_table_header {
+       char signature[4];
+       u32 length;
+       u8 revision;
+       u8 checksum;
+       char oem_id[6];
+       char oem_table_id[8];
+       char reserved[24];
+} __attribute__((__packed__));
+
+/*
+ * Logical location of iSCSI Boot Format Table.
+ * If the value is NULL there is no iBFT on the machine.
+ */
+extern struct ibft_table_header *ibft_addr;
+
+/*
+ * Routine used to find and reserve the iSCSI Boot Format Table. The
+ * mapped address is set in the ibft_addr variable.
+ */
+#ifdef CONFIG_ISCSI_IBFT_FIND
+extern void __init reserve_ibft_region(void);
+#else
+static inline void reserve_ibft_region(void) { }
+#endif
+
+#endif /* ISCSI_IBFT_H */