blob: 14e656ff73ecde3d825b8d30c7d1019905082e43 [file] [log] [blame]
/* Copyright (C) 2010 - 2013 UNISYS CORPORATION */
/* All rights reserved. */
#ifndef __IOCHANNEL_H__
#define __IOCHANNEL_H__
/*
* Everything needed for IOPart-GuestPart communication is define in
* this file. Note: Everything is OS-independent because this file is
* used by Windows, Linux and possible EFI drivers. */
/*
* Communication flow between the IOPart and GuestPart uses the channel headers
* channel state. The following states are currently being used:
* UNINIT(All Zeroes), CHANNEL_ATTACHING, CHANNEL_ATTACHED, CHANNEL_OPENED
*
* additional states will be used later. No locking is needed to switch between
* states due to the following rules:
*
* 1. IOPart is only the only partition allowed to change from UNIT
* 2. IOPart is only the only partition allowed to change from
* CHANNEL_ATTACHING
* 3. GuestPart is only the only partition allowed to change from
* CHANNEL_ATTACHED
*
* The state changes are the following: IOPart sees the channel is in UNINIT,
* UNINIT -> CHANNEL_ATTACHING (performed only by IOPart)
* CHANNEL_ATTACHING -> CHANNEL_ATTACHED (performed only by IOPart)
* CHANNEL_ATTACHED -> CHANNEL_OPENED (performed only by GuestPart)
*/
#include <linux/uuid.h>
#include <linux/dma-direction.h>
#include "channel.h"
#include "channel_guid.h"
#define ULTRA_VHBA_CHANNEL_PROTOCOL_SIGNATURE ULTRA_CHANNEL_PROTOCOL_SIGNATURE
#define ULTRA_VNIC_CHANNEL_PROTOCOL_SIGNATURE ULTRA_CHANNEL_PROTOCOL_SIGNATURE
#define ULTRA_VSWITCH_CHANNEL_PROTOCOL_SIGNATURE \
ULTRA_CHANNEL_PROTOCOL_SIGNATURE
/* Must increment these whenever you insert or delete fields within this channel
* struct. Also increment whenever you change the meaning of fields within this
* channel struct so as to break pre-existing software. Note that you can
* usually add fields to the END of the channel struct withOUT needing to
* increment this.
*/
#define ULTRA_VHBA_CHANNEL_PROTOCOL_VERSIONID 2
#define ULTRA_VNIC_CHANNEL_PROTOCOL_VERSIONID 2
#define ULTRA_VSWITCH_CHANNEL_PROTOCOL_VERSIONID 1
#define SPAR_VHBA_CHANNEL_OK_CLIENT(ch) \
(spar_check_channel_client(ch, spar_vhba_channel_protocol_uuid, \
"vhba", MIN_IO_CHANNEL_SIZE, \
ULTRA_VHBA_CHANNEL_PROTOCOL_VERSIONID, \
ULTRA_VHBA_CHANNEL_PROTOCOL_SIGNATURE))
#define SPAR_VNIC_CHANNEL_OK_CLIENT(ch) \
(spar_check_channel_client(ch, spar_vnic_channel_protocol_uuid, \
"vnic", MIN_IO_CHANNEL_SIZE, \
ULTRA_VNIC_CHANNEL_PROTOCOL_VERSIONID, \
ULTRA_VNIC_CHANNEL_PROTOCOL_SIGNATURE))
/*
* Everything necessary to handle SCSI & NIC traffic between Guest Partition and
* IO Partition is defined below.
*/
/*
* Defines and enums.
*/
#define MINNUM(a, b) (((a) < (b)) ? (a) : (b))
#define MAXNUM(a, b) (((a) > (b)) ? (a) : (b))
/* these define the two queues per data channel between iopart and
* ioguestparts
*/
#define IOCHAN_TO_IOPART 0 /* used by ioguestpart to 'insert' signals to
* iopart */
#define IOCHAN_FROM_IOPART 1 /* used by ioguestpart to 'remove' signals from
* iopart - same queue as previous queue */
/* size of cdb - i.e., scsi cmnd */
#define MAX_CMND_SIZE 16
#define MAX_SENSE_SIZE 64
#define MAX_PHYS_INFO 64
/* various types of network packets that can be sent in cmdrsp */
enum net_types {
NET_RCV_POST = 0, /* submit buffer to hold receiving
* incoming packet */
/* virtnic -> uisnic */
NET_RCV, /* incoming packet received */
/* uisnic -> virtpci */
NET_XMIT, /* for outgoing net packets */
/* virtnic -> uisnic */
NET_XMIT_DONE, /* outgoing packet xmitted */
/* uisnic -> virtpci */
NET_RCV_ENBDIS, /* enable/disable packet reception */
/* virtnic -> uisnic */
NET_RCV_ENBDIS_ACK, /* acknowledge enable/disable packet
* reception */
/* uisnic -> virtnic */
NET_RCV_PROMISC, /* enable/disable promiscuous mode */
/* virtnic -> uisnic */
NET_CONNECT_STATUS, /* indicate the loss or restoration of a network
* connection */
/* uisnic -> virtnic */
NET_MACADDR, /* indicates the client has requested to update
* its MAC addr */
NET_MACADDR_ACK, /* MAC address */
};
#define ETH_HEADER_SIZE 14 /* size of ethernet header */
#define ETH_MIN_DATA_SIZE 46 /* minimum eth data size */
#define ETH_MIN_PACKET_SIZE (ETH_HEADER_SIZE + ETH_MIN_DATA_SIZE)
#define ETH_MAX_MTU 16384 /* maximum data size */
#ifndef MAX_MACADDR_LEN
#define MAX_MACADDR_LEN 6 /* number of bytes in MAC address */
#endif /* MAX_MACADDR_LEN */
/* various types of scsi task mgmt commands */
enum task_mgmt_types {
TASK_MGMT_ABORT_TASK = 1,
TASK_MGMT_BUS_RESET,
TASK_MGMT_LUN_RESET,
TASK_MGMT_TARGET_RESET,
};
/* various types of vdisk mgmt commands */
enum vdisk_mgmt_types {
VDISK_MGMT_ACQUIRE = 1,
VDISK_MGMT_RELEASE,
};
struct phys_info {
u64 pi_pfn;
u16 pi_off;
u16 pi_len;
} __packed;
#define MIN_NUMSIGNALS 64
/* structs with pragma pack */
struct guest_phys_info {
u64 address;
u64 length;
} __packed;
#define GPI_ENTRIES_PER_PAGE (PAGE_SIZE / sizeof(struct guest_phys_info))
struct uisscsi_dest {
u32 channel; /* channel == bus number */
u32 id; /* id == target number */
u32 lun; /* lun == logical unit number */
} __packed;
struct vhba_wwnn {
u32 wwnn1;
u32 wwnn2;
} __packed;
/* WARNING: Values stired in this structure must contain maximum counts (not
* maximum values). */
struct vhba_config_max { /* 20 bytes */
u32 max_channel; /* maximum channel for devices attached to this
* bus */
u32 max_id; /* maximum SCSI ID for devices attached to this
* bus */
u32 max_lun; /* maximum SCSI LUN for devices attached to this
* bus */
u32 cmd_per_lun; /* maximum number of outstanding commands per
* lun that are allowed at one time */
u32 max_io_size; /* maximum io size for devices attached to this
* bus */
/* max io size is often determined by the resource of the hba. e.g */
/* max scatter gather list length * page size / sector size */
} __packed;
struct uiscmdrsp_scsi {
u64 handle; /* the handle to the cmd that was received -
* send it back as is in the rsp packet. */
u8 cmnd[MAX_CMND_SIZE]; /* the cdb for the command */
u32 bufflen; /* length of data to be transferred out or in */
u16 guest_phys_entries; /* Number of entries in scatter-gather (sg)
* list */
struct guest_phys_info gpi_list[MAX_PHYS_INFO]; /* physical address
* information for each
* fragment */
enum dma_data_direction data_dir; /* direction of the data, if any */
struct uisscsi_dest vdest; /* identifies the virtual hba, id,
* channel, lun to which cmd was sent */
/* the following fields are needed to queue the rsp back to cmd
* originator */
int linuxstat; /* the original Linux status - for use by linux
* vdisk code */
u8 scsistat; /* the scsi status */
u8 addlstat; /* non-scsi status - covers cases like timeout
* needed by windows guests */
#define ADDL_SEL_TIMEOUT 4
/* the following fields are need to determine the result of command */
u8 sensebuf[MAX_SENSE_SIZE]; /* sense info in case cmd failed; */
/* it holds the sense_data struct; */
/* see that struct for details. */
void *vdisk; /* contains pointer to the vdisk so that we can clean up
* when the IO completes. */
int no_disk_result;
/* used to return no disk inquiry result
* when no_disk_result is set to 1,
* scsi.scsistat is SAM_STAT_GOOD
* scsi.addlstat is 0
* scsi.linuxstat is SAM_STAT_GOOD
* That is, there is NO error.
*/
} __packed;
/* Defines to support sending correct inquiry result when no disk is
* configured.
*/
/* From SCSI SPC2 -
*
* If the target is not capable of supporting a device on this logical unit, the
* device server shall set this field to 7Fh (PERIPHERAL QUALIFIER set to 011b
* and PERIPHERAL DEVICE TYPE set to 1Fh).
*
*The device server is capable of supporting the specified peripheral device
*type on this logical unit. However, the physical device is not currently
*connected to this logical unit.
*/
#define DEV_NOT_CAPABLE 0x7f /* peripheral qualifier of 0x3 */
/* peripheral type of 0x1f */
/* specifies no device but target present */
#define DEV_DISK_CAPABLE_NOT_PRESENT 0x20 /* peripheral qualifier of 0x1 */
/* peripheral type of 0 - disk */
/* specifies device capable, but not present */
#define DEV_HISUPPORT 0x10 /* HiSup = 1; shows support for report luns */
/* must be returned for lun 0. */
/* NOTE: Linux code assumes inquiry contains 36 bytes. Without checking length
* in buf[4] some linux code accesses bytes beyond 5 to retrieve vendor, product
* & revision. Yikes! So let us always send back 36 bytes, the minimum for
* inquiry result.
*/
#define NO_DISK_INQUIRY_RESULT_LEN 36
#define MIN_INQUIRY_RESULT_LEN 5 /* we need at least 5 bytes minimum for inquiry
* result */
/* SCSI device version for no disk inquiry result */
#define SCSI_SPC2_VER 4 /* indicates SCSI SPC2 (SPC3 is 5) */
/* Windows and Linux want different things for a non-existent lun. So, we'll let
* caller pass in the peripheral qualifier and type.
* NOTE:[4] SCSI returns (n-4); so we return length-1-4 or length-5. */
#define SET_NO_DISK_INQUIRY_RESULT(buf, len, lun, lun0notpresent, notpresent) \
do { \
memset(buf, 0, \
MINNUM(len, \
(unsigned int)NO_DISK_INQUIRY_RESULT_LEN)); \
buf[2] = (u8)SCSI_SPC2_VER; \
if (lun == 0) { \
buf[0] = (u8)lun0notpresent; \
buf[3] = (u8)DEV_HISUPPORT; \
} else \
buf[0] = (u8)notpresent; \
buf[4] = (u8)( \
MINNUM(len, \
(unsigned int)NO_DISK_INQUIRY_RESULT_LEN) - 5);\
if (len >= NO_DISK_INQUIRY_RESULT_LEN) { \
buf[8] = 'D'; \
buf[9] = 'E'; \
buf[10] = 'L'; \
buf[11] = 'L'; \
buf[16] = 'P'; \
buf[17] = 'S'; \
buf[18] = 'E'; \
buf[19] = 'U'; \
buf[20] = 'D'; \
buf[21] = 'O'; \
buf[22] = ' '; \
buf[23] = 'D'; \
buf[24] = 'E'; \
buf[25] = 'V'; \
buf[26] = 'I'; \
buf[27] = 'C'; \
buf[28] = 'E'; \
buf[30] = ' '; \
buf[31] = '.'; \
} \
} while (0)
/*
* Struct & Defines to support sense information.
*/
/* The following struct is returned in sensebuf field in uiscmdrsp_scsi. It is
* initialized in exactly the manner that is recommended in Windows (hence the
* odd values).
* When set, these fields will have the following values:
* ErrorCode = 0x70 indicates current error
* Valid = 1 indicates sense info is valid
* SenseKey contains sense key as defined by SCSI specs.
* AdditionalSenseCode contains sense key as defined by SCSI specs.
* AdditionalSenseCodeQualifier contains qualifier to sense code as defined by
* scsi docs.
* AdditionalSenseLength contains will be sizeof(sense_data)-8=10.
*/
struct sense_data {
u8 errorcode:7;
u8 valid:1;
u8 segment_number;
u8 sense_key:4;
u8 reserved:1;
u8 incorrect_length:1;
u8 end_of_media:1;
u8 file_mark:1;
u8 information[4];
u8 additional_sense_length;
u8 command_specific_information[4];
u8 additional_sense_code;
u8 additional_sense_code_qualifier;
u8 fru_code;
u8 sense_key_specific[3];
} __packed;
struct net_pkt_xmt {
int len; /* full length of data in the packet */
int num_frags; /* number of fragments in frags containing data */
struct phys_info frags[MAX_PHYS_INFO]; /* physical page information for
* each fragment */
char ethhdr[ETH_HEADER_SIZE]; /* the ethernet header */
struct {
/* these are needed for csum at uisnic end */
u8 valid; /* 1 = rest of this struct is valid - else
* ignore */
u8 hrawoffv; /* 1 = hwrafoff is valid */
u8 nhrawoffv; /* 1 = nhwrafoff is valid */
u16 protocol; /* specifies packet protocol */
u32 csum; /* value used to set skb->csum at IOPart */
u32 hrawoff; /* value used to set skb->h.raw at IOPart */
/* hrawoff points to the start of the TRANSPORT LAYER HEADER */
u32 nhrawoff; /* value used to set skb->nh.raw at IOPart */
/* nhrawoff points to the start of the NETWORK LAYER HEADER */
} lincsum;
/* **** NOTE ****
* The full packet is described in frags but the ethernet header is
* separately kept in ethhdr so that uisnic doesn't have "MAP" the
* guest memory to get to the header. uisnic needs ethhdr to
* determine how to route the packet.
*/
} __packed;
struct net_pkt_xmtdone {
u32 xmt_done_result; /* result of NET_XMIT */
} __packed;
/* RCVPOST_BUF_SIZe must be at most page_size(4096) - cache_line_size (64) The
* reason is because dev_skb_alloc which is used to generate RCV_POST skbs in
* virtnic requires that there is "overhead" in the buffer, and pads 16 bytes. I
* prefer to use 1 full cache line size for "overhead" so that transfers are
* better. IOVM requires that a buffer be represented by 1 phys_info structure
* which can only cover page_size.
*/
#define RCVPOST_BUF_SIZE 4032
#define MAX_NET_RCV_CHAIN \
((ETH_MAX_MTU+ETH_HEADER_SIZE + RCVPOST_BUF_SIZE-1) / RCVPOST_BUF_SIZE)
struct net_pkt_rcvpost {
/* rcv buf size must be large enough to include ethernet data len +
* ethernet header len - we are choosing 2K because it is guaranteed
* to be describable */
struct phys_info frag; /* physical page information for the
* single fragment 2K rcv buf */
u64 unique_num; /* This is used to make sure that
* receive posts are returned to */
/* the Adapter which we sent them originally. */
} __packed;
struct net_pkt_rcv {
/* the number of receive buffers that can be chained */
/* is based on max mtu and size of each rcv buf */
u32 rcv_done_len; /* length of received data */
u8 numrcvbufs; /* number of receive buffers that contain the */
/* incoming data; guest end MUST chain these together. */
void *rcvbuf[MAX_NET_RCV_CHAIN]; /* the list of receive buffers
* that must be chained; */
/* each entry is a receive buffer provided by NET_RCV_POST. */
/* NOTE: first rcvbuf in the chain will also be provided in net.buf. */
u64 unique_num;
u32 rcvs_dropped_delta;
} __packed;
struct net_pkt_enbdis {
void *context;
u16 enable; /* 1 = enable, 0 = disable */
} __packed;
struct net_pkt_macaddr {
void *context;
u8 macaddr[MAX_MACADDR_LEN]; /* 6 bytes */
} __packed;
/* cmd rsp packet used for VNIC network traffic */
struct uiscmdrsp_net {
enum net_types type;
void *buf;
union {
struct net_pkt_xmt xmt; /* used for NET_XMIT */
struct net_pkt_xmtdone xmtdone; /* used for NET_XMIT_DONE */
struct net_pkt_rcvpost rcvpost; /* used for NET_RCV_POST */
struct net_pkt_rcv rcv; /* used for NET_RCV */
struct net_pkt_enbdis enbdis; /* used for NET_RCV_ENBDIS, */
/* NET_RCV_ENBDIS_ACK, */
/* NET_RCV_PROMSIC, */
/* and NET_CONNECT_STATUS */
struct net_pkt_macaddr macaddr;
};
} __packed;
struct uiscmdrsp_scsitaskmgmt {
enum task_mgmt_types tasktype;
/* the type of task */
struct uisscsi_dest vdest;
/* the vdisk for which this task mgmt is generated */
u64 handle;
/* This is a handle that the guest has saved off for its own use.
* Its value is preserved by iopart & returned as is in the task
* mgmt rsp.
*/
u64 notify_handle;
/* For linux guests, this is a pointer to wait_queue_head that a
* thread is waiting on to see if the taskmgmt command has completed.
* When the rsp is received by guest, the thread receiving the
* response uses this to notify the thread waiting for taskmgmt
* command completion. Its value is preserved by iopart & returned
* as is in the task mgmt rsp.
*/
u64 notifyresult_handle;
/* this is a handle to location in guest where the result of the
* taskmgmt command (result field) is to saved off when the response
* is handled. Its value is preserved by iopart & returned as is in
* the task mgmt rsp.
*/
char result;
/* result of taskmgmt command - set by IOPart - values are: */
#define TASK_MGMT_FAILED 0
} __packed;
/* The following is used by uissd to send disk add/remove notifications to
* Guest */
/* Note that the vHba pointer is not used by the Client/Guest side. */
struct uiscmdrsp_disknotify {
u8 add; /* 0-remove, 1-add */
void *v_hba; /* Pointer to vhba_info for channel info to
* route msg */
u32 channel, id, lun; /* SCSI Path of Disk to added or removed */
} __packed;
/* The following is used by virthba/vSCSI to send the Acquire/Release commands
* to the IOVM. */
struct uiscmdrsp_vdiskmgmt {
enum vdisk_mgmt_types vdisktype;
/* the type of task */
struct uisscsi_dest vdest;
/* the vdisk for which this task mgmt is generated */
u64 handle;
/* This is a handle that the guest has saved off for its own use.
* Its value is preserved by iopart & returned as is in the task
* mgmt rsp.
*/
u64 notify_handle;
/* For linux guests, this is a pointer to wait_queue_head that a
* thread is waiting on to see if the tskmgmt command has completed.
* When the rsp is received by guest, the thread receiving the
* response uses this to notify the thread waiting for taskmgmt
* command completion. Its value is preserved by iopart & returned
* as is in the task mgmt rsp.
*/
u64 notifyresult_handle;
/* this is a handle to location in guest where the result of the
* taskmgmt command (result field) is to saved off when the response
* is handled. Its value is preserved by iopart & returned as is in
* the task mgmt rsp.
*/
char result;
/* result of taskmgmt command - set by IOPart - values are: */
#define VDISK_MGMT_FAILED 0
} __packed;
/* keeping cmd & rsp info in one structure for now cmd rsp packet for scsi */
struct uiscmdrsp {
char cmdtype;
/* describes what type of information is in the struct */
#define CMD_SCSI_TYPE 1
#define CMD_NET_TYPE 2
#define CMD_SCSITASKMGMT_TYPE 3
#define CMD_NOTIFYGUEST_TYPE 4
#define CMD_VDISKMGMT_TYPE 5
union {
struct uiscmdrsp_scsi scsi;
struct uiscmdrsp_net net;
struct uiscmdrsp_scsitaskmgmt scsitaskmgmt;
struct uiscmdrsp_disknotify disknotify;
struct uiscmdrsp_vdiskmgmt vdiskmgmt;
};
void *private_data; /* used to send the response when the cmd is
* done (scsi & scsittaskmgmt). */
struct uiscmdrsp *next; /* General Purpose Queue Link */
struct uiscmdrsp *activeQ_next; /* Used to track active commands */
struct uiscmdrsp *activeQ_prev; /* Used to track active commands */
} __packed;
struct iochannel_vhba {
struct vhba_wwnn wwnn; /* 8 bytes */
struct vhba_config_max max; /* 20 bytes */
} __packed; /* total = 28 bytes */
struct iochannel_vnic {
u8 macaddr[6]; /* 6 bytes */
u32 num_rcv_bufs; /* 4 bytes */
u32 mtu; /* 4 bytes */
uuid_le zone_uuid; /* 16 bytes */
} __packed;
/* This is just the header of the IO channel. It is assumed that directly after
* this header there is a large region of memory which contains the command and
* response queues as specified in cmd_q and rsp_q SIGNAL_QUEUE_HEADERS.
*/
struct spar_io_channel_protocol {
struct channel_header channel_header;
struct signal_queue_header cmd_q;
struct signal_queue_header rsp_q;
union {
struct iochannel_vhba vhba;
struct iochannel_vnic vnic;
} __packed;
#define MAX_CLIENTSTRING_LEN 1024
u8 client_string[MAX_CLIENTSTRING_LEN];/* NULL terminated - so holds
* max - 1 bytes */
} __packed;
/*
* INLINE functions for initializing and accessing I/O data channels
*/
#define SIZEOF_PROTOCOL (COVER(sizeof(struct spar_io_channel_protocol), 64))
#define SIZEOF_CMDRSP (COVER(sizeof(struct uiscmdrsp), 64))
#define MIN_IO_CHANNEL_SIZE COVER(SIZEOF_PROTOCOL + \
2 * MIN_NUMSIGNALS * SIZEOF_CMDRSP, 4096)
/*
* INLINE function for expanding a guest's pfn-off-size into multiple 4K page
* pfn-off-size entires.
*/
/* we deal with 4K page sizes when we it comes to passing page information
* between */
/* Guest and IOPartition. */
#define PI_PAGE_SIZE 0x1000
#define PI_PAGE_MASK 0x0FFF
/* returns next non-zero index on success or zero on failure (i.e. out of
* room)
*/
static inline u16
add_physinfo_entries(u32 inp_pfn, /* input - specifies the pfn to be used
* to add entries */
u16 inp_off, /* input - specifies the off to be used
* to add entries */
u32 inp_len, /* input - specifies the len to be used
* to add entries */
u16 index, /* input - index in array at which new
* entries are added */
u16 max_pi_arr_entries, /* input - specifies the maximum
* entries pi_arr can hold */
struct phys_info pi_arr[]) /* input & output - array to
* which entries are added */
{
u32 len;
u16 i, firstlen;
firstlen = PI_PAGE_SIZE - inp_off;
if (inp_len <= firstlen) {
/* the input entry spans only one page - add as is */
if (index >= max_pi_arr_entries)
return 0;
pi_arr[index].pi_pfn = inp_pfn;
pi_arr[index].pi_off = (u16)inp_off;
pi_arr[index].pi_len = (u16)inp_len;
return index + 1;
}
/* this entry spans multiple pages */
for (len = inp_len, i = 0; len;
len -= pi_arr[index + i].pi_len, i++) {
if (index + i >= max_pi_arr_entries)
return 0;
pi_arr[index + i].pi_pfn = inp_pfn + i;
if (i == 0) {
pi_arr[index].pi_off = inp_off;
pi_arr[index].pi_len = firstlen;
}
else {
pi_arr[index + i].pi_off = 0;
pi_arr[index + i].pi_len =
(u16)MINNUM(len, (u32)PI_PAGE_SIZE);
}
}
return index + i;
}
#endif /* __IOCHANNEL_H__ */