linux/drivers/net/e100.c

2943 lines
86 KiB
C
Raw Normal View History

/*******************************************************************************
Intel PRO/100 Linux driver
Copyright(c) 1999 - 2006 Intel Corporation.
This program is free software; you can redistribute it and/or modify it
under the terms and conditions of the GNU General Public License,
version 2, as published by the Free Software Foundation.
This program is distributed in the hope 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.
You should have received a copy of the GNU General Public License along with
this program; if not, write to the Free Software Foundation, Inc.,
51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
The full GNU General Public License is included in this distribution in
the file called "COPYING".
Contact Information:
Linux NICS <linux.nics@intel.com>
e1000-devel Mailing List <e1000-devel@lists.sourceforge.net>
Intel Corporation, 5200 N.E. Elam Young Parkway, Hillsboro, OR 97124-6497
*******************************************************************************/
/*
* e100.c: Intel(R) PRO/100 ethernet driver
*
* (Re)written 2003 by scott.feldman@intel.com. Based loosely on
* original e100 driver, but better described as a munging of
* e100, e1000, eepro100, tg3, 8139cp, and other drivers.
*
* References:
* Intel 8255x 10/100 Mbps Ethernet Controller Family,
* Open Source Software Developers Manual,
* http://sourceforge.net/projects/e1000
*
*
* Theory of Operation
*
* I. General
*
* The driver supports Intel(R) 10/100 Mbps PCI Fast Ethernet
* controller family, which includes the 82557, 82558, 82559, 82550,
* 82551, and 82562 devices. 82558 and greater controllers
* integrate the Intel 82555 PHY. The controllers are used in
* server and client network interface cards, as well as in
* LAN-On-Motherboard (LOM), CardBus, MiniPCI, and ICHx
* configurations. 8255x supports a 32-bit linear addressing
* mode and operates at 33Mhz PCI clock rate.
*
* II. Driver Operation
*
* Memory-mapped mode is used exclusively to access the device's
* shared-memory structure, the Control/Status Registers (CSR). All
* setup, configuration, and control of the device, including queuing
* of Tx, Rx, and configuration commands is through the CSR.
* cmd_lock serializes accesses to the CSR command register. cb_lock
* protects the shared Command Block List (CBL).
*
* 8255x is highly MII-compliant and all access to the PHY go
* through the Management Data Interface (MDI). Consequently, the
* driver leverages the mii.c library shared with other MII-compliant
* devices.
*
* Big- and Little-Endian byte order as well as 32- and 64-bit
* archs are supported. Weak-ordered memory and non-cache-coherent
* archs are supported.
*
* III. Transmit
*
* A Tx skb is mapped and hangs off of a TCB. TCBs are linked
* together in a fixed-size ring (CBL) thus forming the flexible mode
* memory structure. A TCB marked with the suspend-bit indicates
* the end of the ring. The last TCB processed suspends the
* controller, and the controller can be restarted by issue a CU
* resume command to continue from the suspend point, or a CU start
* command to start at a given position in the ring.
*
* Non-Tx commands (config, multicast setup, etc) are linked
* into the CBL ring along with Tx commands. The common structure
* used for both Tx and non-Tx commands is the Command Block (CB).
*
* cb_to_use is the next CB to use for queuing a command; cb_to_clean
* is the next CB to check for completion; cb_to_send is the first
* CB to start on in case of a previous failure to resume. CB clean
* up happens in interrupt context in response to a CU interrupt.
* cbs_avail keeps track of number of free CB resources available.
*
* Hardware padding of short packets to minimum packet size is
* enabled. 82557 pads with 7Eh, while the later controllers pad
* with 00h.
*
* IV. Receive
*
* The Receive Frame Area (RFA) comprises a ring of Receive Frame
* Descriptors (RFD) + data buffer, thus forming the simplified mode
* memory structure. Rx skbs are allocated to contain both the RFD
* and the data buffer, but the RFD is pulled off before the skb is
* indicated. The data buffer is aligned such that encapsulated
* protocol headers are u32-aligned. Since the RFD is part of the
* mapped shared memory, and completion status is contained within
* the RFD, the RFD must be dma_sync'ed to maintain a consistent
* view from software and hardware.
*
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
* In order to keep updates to the RFD link field from colliding with
* hardware writes to mark packets complete, we use the feature that
* hardware will not write to a size 0 descriptor and mark the previous
* packet as end-of-list (EL). After updating the link, we remove EL
* and only then restore the size such that hardware may use the
* previous-to-end RFD.
*
* Under typical operation, the receive unit (RU) is start once,
* and the controller happily fills RFDs as frames arrive. If
* replacement RFDs cannot be allocated, or the RU goes non-active,
* the RU must be restarted. Frame arrival generates an interrupt,
* and Rx indication and re-allocation happen in the same context,
* therefore no locking is required. A software-generated interrupt
* is generated from the watchdog to recover from a failed allocation
* scenario where all Rx resources have been indicated and none re-
* placed.
*
* V. Miscellaneous
*
* VLAN offloading of tagging, stripping and filtering is not
* supported, but driver will accommodate the extra 4-byte VLAN tag
* for processing by upper layers. Tx/Rx Checksum offloading is not
* supported. Tx Scatter/Gather is not supported. Jumbo Frames is
* not supported (hardware limitation).
*
* MagicPacket(tm) WoL support is enabled/disabled via ethtool.
*
* Thanks to JC (jchapman@katalix.com) for helping with
* testing/troubleshooting the development driver.
*
* TODO:
* o several entry points race with dev->close
* o check for tx-no-resources/stop Q races with tx clean/wake Q
*
* FIXES:
* 2005/12/02 - Michael O'Donnell <Michael.ODonnell at stratus dot com>
* - Stratus87247: protect MDI control register manipulations
*/
#include <linux/module.h>
#include <linux/moduleparam.h>
#include <linux/kernel.h>
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/delay.h>
#include <linux/init.h>
#include <linux/pci.h>
#include <linux/dma-mapping.h>
#include <linux/netdevice.h>
#include <linux/etherdevice.h>
#include <linux/mii.h>
#include <linux/if_vlan.h>
#include <linux/skbuff.h>
#include <linux/ethtool.h>
#include <linux/string.h>
#include <asm/unaligned.h>
#define DRV_NAME "e100"
#define DRV_EXT "-NAPI"
#define DRV_VERSION "3.5.23-k4"DRV_EXT
#define DRV_DESCRIPTION "Intel(R) PRO/100 Network Driver"
#define DRV_COPYRIGHT "Copyright(c) 1999-2006 Intel Corporation"
#define PFX DRV_NAME ": "
#define E100_WATCHDOG_PERIOD (2 * HZ)
#define E100_NAPI_WEIGHT 16
MODULE_DESCRIPTION(DRV_DESCRIPTION);
MODULE_AUTHOR(DRV_COPYRIGHT);
MODULE_LICENSE("GPL");
MODULE_VERSION(DRV_VERSION);
static int debug = 3;
static int eeprom_bad_csum_allow = 0;
static int use_io = 0;
module_param(debug, int, 0);
module_param(eeprom_bad_csum_allow, int, 0);
module_param(use_io, int, 0);
MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
MODULE_PARM_DESC(eeprom_bad_csum_allow, "Allow bad eeprom checksums");
MODULE_PARM_DESC(use_io, "Force use of i/o access mode");
#define DPRINTK(nlevel, klevel, fmt, args...) \
(void)((NETIF_MSG_##nlevel & nic->msg_enable) && \
printk(KERN_##klevel PFX "%s: %s: " fmt, nic->netdev->name, \
__func__ , ## args))
#define INTEL_8255X_ETHERNET_DEVICE(device_id, ich) {\
PCI_VENDOR_ID_INTEL, device_id, PCI_ANY_ID, PCI_ANY_ID, \
PCI_CLASS_NETWORK_ETHERNET << 8, 0xFFFF00, ich }
static struct pci_device_id e100_id_table[] = {
INTEL_8255X_ETHERNET_DEVICE(0x1029, 0),
INTEL_8255X_ETHERNET_DEVICE(0x1030, 0),
INTEL_8255X_ETHERNET_DEVICE(0x1031, 3),
INTEL_8255X_ETHERNET_DEVICE(0x1032, 3),
INTEL_8255X_ETHERNET_DEVICE(0x1033, 3),
INTEL_8255X_ETHERNET_DEVICE(0x1034, 3),
INTEL_8255X_ETHERNET_DEVICE(0x1038, 3),
INTEL_8255X_ETHERNET_DEVICE(0x1039, 4),
INTEL_8255X_ETHERNET_DEVICE(0x103A, 4),
INTEL_8255X_ETHERNET_DEVICE(0x103B, 4),
INTEL_8255X_ETHERNET_DEVICE(0x103C, 4),
INTEL_8255X_ETHERNET_DEVICE(0x103D, 4),
INTEL_8255X_ETHERNET_DEVICE(0x103E, 4),
INTEL_8255X_ETHERNET_DEVICE(0x1050, 5),
INTEL_8255X_ETHERNET_DEVICE(0x1051, 5),
INTEL_8255X_ETHERNET_DEVICE(0x1052, 5),
INTEL_8255X_ETHERNET_DEVICE(0x1053, 5),
INTEL_8255X_ETHERNET_DEVICE(0x1054, 5),
INTEL_8255X_ETHERNET_DEVICE(0x1055, 5),
INTEL_8255X_ETHERNET_DEVICE(0x1056, 5),
INTEL_8255X_ETHERNET_DEVICE(0x1057, 5),
INTEL_8255X_ETHERNET_DEVICE(0x1059, 0),
INTEL_8255X_ETHERNET_DEVICE(0x1064, 6),
INTEL_8255X_ETHERNET_DEVICE(0x1065, 6),
INTEL_8255X_ETHERNET_DEVICE(0x1066, 6),
INTEL_8255X_ETHERNET_DEVICE(0x1067, 6),
INTEL_8255X_ETHERNET_DEVICE(0x1068, 6),
INTEL_8255X_ETHERNET_DEVICE(0x1069, 6),
INTEL_8255X_ETHERNET_DEVICE(0x106A, 6),
INTEL_8255X_ETHERNET_DEVICE(0x106B, 6),
INTEL_8255X_ETHERNET_DEVICE(0x1091, 7),
INTEL_8255X_ETHERNET_DEVICE(0x1092, 7),
INTEL_8255X_ETHERNET_DEVICE(0x1093, 7),
INTEL_8255X_ETHERNET_DEVICE(0x1094, 7),
INTEL_8255X_ETHERNET_DEVICE(0x1095, 7),
INTEL_8255X_ETHERNET_DEVICE(0x1209, 0),
INTEL_8255X_ETHERNET_DEVICE(0x1229, 0),
INTEL_8255X_ETHERNET_DEVICE(0x2449, 2),
INTEL_8255X_ETHERNET_DEVICE(0x2459, 2),
INTEL_8255X_ETHERNET_DEVICE(0x245D, 2),
INTEL_8255X_ETHERNET_DEVICE(0x27DC, 7),
{ 0, }
};
MODULE_DEVICE_TABLE(pci, e100_id_table);
enum mac {
mac_82557_D100_A = 0,
mac_82557_D100_B = 1,
mac_82557_D100_C = 2,
mac_82558_D101_A4 = 4,
mac_82558_D101_B0 = 5,
mac_82559_D101M = 8,
mac_82559_D101S = 9,
mac_82550_D102 = 12,
mac_82550_D102_C = 13,
mac_82551_E = 14,
mac_82551_F = 15,
mac_82551_10 = 16,
mac_unknown = 0xFF,
};
enum phy {
phy_100a = 0x000003E0,
phy_100c = 0x035002A8,
phy_82555_tx = 0x015002A8,
phy_nsc_tx = 0x5C002000,
phy_82562_et = 0x033002A8,
phy_82562_em = 0x032002A8,
phy_82562_ek = 0x031002A8,
phy_82562_eh = 0x017002A8,
phy_unknown = 0xFFFFFFFF,
};
/* CSR (Control/Status Registers) */
struct csr {
struct {
u8 status;
u8 stat_ack;
u8 cmd_lo;
u8 cmd_hi;
u32 gen_ptr;
} scb;
u32 port;
u16 flash_ctrl;
u8 eeprom_ctrl_lo;
u8 eeprom_ctrl_hi;
u32 mdi_ctrl;
u32 rx_dma_count;
};
enum scb_status {
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
rus_no_res = 0x08,
rus_ready = 0x10,
rus_mask = 0x3C,
};
enum ru_state {
RU_SUSPENDED = 0,
RU_RUNNING = 1,
RU_UNINITIALIZED = -1,
};
enum scb_stat_ack {
stat_ack_not_ours = 0x00,
stat_ack_sw_gen = 0x04,
stat_ack_rnr = 0x10,
stat_ack_cu_idle = 0x20,
stat_ack_frame_rx = 0x40,
stat_ack_cu_cmd_done = 0x80,
stat_ack_not_present = 0xFF,
stat_ack_rx = (stat_ack_sw_gen | stat_ack_rnr | stat_ack_frame_rx),
stat_ack_tx = (stat_ack_cu_idle | stat_ack_cu_cmd_done),
};
enum scb_cmd_hi {
irq_mask_none = 0x00,
irq_mask_all = 0x01,
irq_sw_gen = 0x02,
};
enum scb_cmd_lo {
cuc_nop = 0x00,
ruc_start = 0x01,
ruc_load_base = 0x06,
cuc_start = 0x10,
cuc_resume = 0x20,
cuc_dump_addr = 0x40,
cuc_dump_stats = 0x50,
cuc_load_base = 0x60,
cuc_dump_reset = 0x70,
};
enum cuc_dump {
cuc_dump_complete = 0x0000A005,
cuc_dump_reset_complete = 0x0000A007,
};
enum port {
software_reset = 0x0000,
selftest = 0x0001,
selective_reset = 0x0002,
};
enum eeprom_ctrl_lo {
eesk = 0x01,
eecs = 0x02,
eedi = 0x04,
eedo = 0x08,
};
enum mdi_ctrl {
mdi_write = 0x04000000,
mdi_read = 0x08000000,
mdi_ready = 0x10000000,
};
enum eeprom_op {
op_write = 0x05,
op_read = 0x06,
op_ewds = 0x10,
op_ewen = 0x13,
};
enum eeprom_offsets {
eeprom_cnfg_mdix = 0x03,
eeprom_id = 0x0A,
eeprom_config_asf = 0x0D,
eeprom_smbus_addr = 0x90,
};
enum eeprom_cnfg_mdix {
eeprom_mdix_enabled = 0x0080,
};
enum eeprom_id {
eeprom_id_wol = 0x0020,
};
enum eeprom_config_asf {
eeprom_asf = 0x8000,
eeprom_gcl = 0x4000,
};
enum cb_status {
cb_complete = 0x8000,
cb_ok = 0x2000,
};
enum cb_command {
cb_nop = 0x0000,
cb_iaaddr = 0x0001,
cb_config = 0x0002,
cb_multi = 0x0003,
cb_tx = 0x0004,
cb_ucode = 0x0005,
cb_dump = 0x0006,
cb_tx_sf = 0x0008,
cb_cid = 0x1f00,
cb_i = 0x2000,
cb_s = 0x4000,
cb_el = 0x8000,
};
struct rfd {
__le16 status;
__le16 command;
__le32 link;
__le32 rbd;
__le16 actual_size;
__le16 size;
};
struct rx {
struct rx *next, *prev;
struct sk_buff *skb;
dma_addr_t dma_addr;
};
#if defined(__BIG_ENDIAN_BITFIELD)
#define X(a,b) b,a
#else
#define X(a,b) a,b
#endif
struct config {
/*0*/ u8 X(byte_count:6, pad0:2);
/*1*/ u8 X(X(rx_fifo_limit:4, tx_fifo_limit:3), pad1:1);
/*2*/ u8 adaptive_ifs;
/*3*/ u8 X(X(X(X(mwi_enable:1, type_enable:1), read_align_enable:1),
term_write_cache_line:1), pad3:4);
/*4*/ u8 X(rx_dma_max_count:7, pad4:1);
/*5*/ u8 X(tx_dma_max_count:7, dma_max_count_enable:1);
/*6*/ u8 X(X(X(X(X(X(X(late_scb_update:1, direct_rx_dma:1),
tno_intr:1), cna_intr:1), standard_tcb:1), standard_stat_counter:1),
rx_discard_overruns:1), rx_save_bad_frames:1);
/*7*/ u8 X(X(X(X(X(rx_discard_short_frames:1, tx_underrun_retry:2),
pad7:2), rx_extended_rfd:1), tx_two_frames_in_fifo:1),
tx_dynamic_tbd:1);
/*8*/ u8 X(X(mii_mode:1, pad8:6), csma_disabled:1);
/*9*/ u8 X(X(X(X(X(rx_tcpudp_checksum:1, pad9:3), vlan_arp_tco:1),
link_status_wake:1), arp_wake:1), mcmatch_wake:1);
/*10*/ u8 X(X(X(pad10:3, no_source_addr_insertion:1), preamble_length:2),
loopback:2);
/*11*/ u8 X(linear_priority:3, pad11:5);
/*12*/ u8 X(X(linear_priority_mode:1, pad12:3), ifs:4);
/*13*/ u8 ip_addr_lo;
/*14*/ u8 ip_addr_hi;
/*15*/ u8 X(X(X(X(X(X(X(promiscuous_mode:1, broadcast_disabled:1),
wait_after_win:1), pad15_1:1), ignore_ul_bit:1), crc_16_bit:1),
pad15_2:1), crs_or_cdt:1);
/*16*/ u8 fc_delay_lo;
/*17*/ u8 fc_delay_hi;
/*18*/ u8 X(X(X(X(X(rx_stripping:1, tx_padding:1), rx_crc_transfer:1),
rx_long_ok:1), fc_priority_threshold:3), pad18:1);
/*19*/ u8 X(X(X(X(X(X(X(addr_wake:1, magic_packet_disable:1),
fc_disable:1), fc_restop:1), fc_restart:1), fc_reject:1),
full_duplex_force:1), full_duplex_pin:1);
/*20*/ u8 X(X(X(pad20_1:5, fc_priority_location:1), multi_ia:1), pad20_2:1);
/*21*/ u8 X(X(pad21_1:3, multicast_all:1), pad21_2:4);
/*22*/ u8 X(X(rx_d102_mode:1, rx_vlan_drop:1), pad22:6);
u8 pad_d102[9];
};
#define E100_MAX_MULTICAST_ADDRS 64
struct multi {
__le16 count;
u8 addr[E100_MAX_MULTICAST_ADDRS * ETH_ALEN + 2/*pad*/];
};
/* Important: keep total struct u32-aligned */
#define UCODE_SIZE 134
struct cb {
__le16 status;
__le16 command;
__le32 link;
union {
u8 iaaddr[ETH_ALEN];
__le32 ucode[UCODE_SIZE];
struct config config;
struct multi multi;
struct {
u32 tbd_array;
u16 tcb_byte_count;
u8 threshold;
u8 tbd_count;
struct {
__le32 buf_addr;
__le16 size;
u16 eol;
} tbd;
} tcb;
__le32 dump_buffer_addr;
} u;
struct cb *next, *prev;
dma_addr_t dma_addr;
struct sk_buff *skb;
};
enum loopback {
lb_none = 0, lb_mac = 1, lb_phy = 3,
};
struct stats {
__le32 tx_good_frames, tx_max_collisions, tx_late_collisions,
tx_underruns, tx_lost_crs, tx_deferred, tx_single_collisions,
tx_multiple_collisions, tx_total_collisions;
__le32 rx_good_frames, rx_crc_errors, rx_alignment_errors,
rx_resource_errors, rx_overrun_errors, rx_cdt_errors,
rx_short_frame_errors;
__le32 fc_xmt_pause, fc_rcv_pause, fc_rcv_unsupported;
__le16 xmt_tco_frames, rcv_tco_frames;
__le32 complete;
};
struct mem {
struct {
u32 signature;
u32 result;
} selftest;
struct stats stats;
u8 dump_buf[596];
};
struct param_range {
u32 min;
u32 max;
u32 count;
};
struct params {
struct param_range rfds;
struct param_range cbs;
};
struct nic {
/* Begin: frequently used values: keep adjacent for cache effect */
u32 msg_enable ____cacheline_aligned;
struct net_device *netdev;
struct pci_dev *pdev;
struct rx *rxs ____cacheline_aligned;
struct rx *rx_to_use;
struct rx *rx_to_clean;
struct rfd blank_rfd;
enum ru_state ru_running;
spinlock_t cb_lock ____cacheline_aligned;
spinlock_t cmd_lock;
struct csr __iomem *csr;
enum scb_cmd_lo cuc_cmd;
unsigned int cbs_avail;
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
struct napi_struct napi;
struct cb *cbs;
struct cb *cb_to_use;
struct cb *cb_to_send;
struct cb *cb_to_clean;
__le16 tx_command;
/* End: frequently used values: keep adjacent for cache effect */
enum {
ich = (1 << 0),
promiscuous = (1 << 1),
multicast_all = (1 << 2),
wol_magic = (1 << 3),
ich_10h_workaround = (1 << 4),
} flags ____cacheline_aligned;
enum mac mac;
enum phy phy;
struct params params;
struct timer_list watchdog;
struct timer_list blink_timer;
struct mii_if_info mii;
struct work_struct tx_timeout_task;
enum loopback loopback;
struct mem *mem;
dma_addr_t dma_addr;
dma_addr_t cbs_dma_addr;
u8 adaptive_ifs;
u8 tx_threshold;
u32 tx_frames;
u32 tx_collisions;
u32 tx_deferred;
u32 tx_single_collisions;
u32 tx_multiple_collisions;
u32 tx_fc_pause;
u32 tx_tco_frames;
u32 rx_fc_pause;
u32 rx_fc_unsupported;
u32 rx_tco_frames;
u32 rx_over_length_errors;
u16 leds;
u16 eeprom_wc;
__le16 eeprom[256];
spinlock_t mdio_lock;
};
static inline void e100_write_flush(struct nic *nic)
{
/* Flush previous PCI writes through intermediate bridges
* by doing a benign read */
(void)ioread8(&nic->csr->scb.status);
}
static void e100_enable_irq(struct nic *nic)
{
unsigned long flags;
spin_lock_irqsave(&nic->cmd_lock, flags);
iowrite8(irq_mask_none, &nic->csr->scb.cmd_hi);
e100_write_flush(nic);
spin_unlock_irqrestore(&nic->cmd_lock, flags);
}
static void e100_disable_irq(struct nic *nic)
{
unsigned long flags;
spin_lock_irqsave(&nic->cmd_lock, flags);
iowrite8(irq_mask_all, &nic->csr->scb.cmd_hi);
e100_write_flush(nic);
spin_unlock_irqrestore(&nic->cmd_lock, flags);
}
static void e100_hw_reset(struct nic *nic)
{
/* Put CU and RU into idle with a selective reset to get
* device off of PCI bus */
iowrite32(selective_reset, &nic->csr->port);
e100_write_flush(nic); udelay(20);
/* Now fully reset device */
iowrite32(software_reset, &nic->csr->port);
e100_write_flush(nic); udelay(20);
/* Mask off our interrupt line - it's unmasked after reset */
e100_disable_irq(nic);
}
static int e100_self_test(struct nic *nic)
{
u32 dma_addr = nic->dma_addr + offsetof(struct mem, selftest);
/* Passing the self-test is a pretty good indication
* that the device can DMA to/from host memory */
nic->mem->selftest.signature = 0;
nic->mem->selftest.result = 0xFFFFFFFF;
iowrite32(selftest | dma_addr, &nic->csr->port);
e100_write_flush(nic);
/* Wait 10 msec for self-test to complete */
msleep(10);
/* Interrupts are enabled after self-test */
e100_disable_irq(nic);
/* Check results of self-test */
if(nic->mem->selftest.result != 0) {
DPRINTK(HW, ERR, "Self-test failed: result=0x%08X\n",
nic->mem->selftest.result);
return -ETIMEDOUT;
}
if(nic->mem->selftest.signature == 0) {
DPRINTK(HW, ERR, "Self-test failed: timed out\n");
return -ETIMEDOUT;
}
return 0;
}
static void e100_eeprom_write(struct nic *nic, u16 addr_len, u16 addr, __le16 data)
{
u32 cmd_addr_data[3];
u8 ctrl;
int i, j;
/* Three cmds: write/erase enable, write data, write/erase disable */
cmd_addr_data[0] = op_ewen << (addr_len - 2);
cmd_addr_data[1] = (((op_write << addr_len) | addr) << 16) |
le16_to_cpu(data);
cmd_addr_data[2] = op_ewds << (addr_len - 2);
/* Bit-bang cmds to write word to eeprom */
for(j = 0; j < 3; j++) {
/* Chip select */
iowrite8(eecs | eesk, &nic->csr->eeprom_ctrl_lo);
e100_write_flush(nic); udelay(4);
for(i = 31; i >= 0; i--) {
ctrl = (cmd_addr_data[j] & (1 << i)) ?
eecs | eedi : eecs;
iowrite8(ctrl, &nic->csr->eeprom_ctrl_lo);
e100_write_flush(nic); udelay(4);
iowrite8(ctrl | eesk, &nic->csr->eeprom_ctrl_lo);
e100_write_flush(nic); udelay(4);
}
/* Wait 10 msec for cmd to complete */
msleep(10);
/* Chip deselect */
iowrite8(0, &nic->csr->eeprom_ctrl_lo);
e100_write_flush(nic); udelay(4);
}
};
/* General technique stolen from the eepro100 driver - very clever */
static __le16 e100_eeprom_read(struct nic *nic, u16 *addr_len, u16 addr)
{
u32 cmd_addr_data;
u16 data = 0;
u8 ctrl;
int i;
cmd_addr_data = ((op_read << *addr_len) | addr) << 16;
/* Chip select */
iowrite8(eecs | eesk, &nic->csr->eeprom_ctrl_lo);
e100_write_flush(nic); udelay(4);
/* Bit-bang to read word from eeprom */
for(i = 31; i >= 0; i--) {
ctrl = (cmd_addr_data & (1 << i)) ? eecs | eedi : eecs;
iowrite8(ctrl, &nic->csr->eeprom_ctrl_lo);
e100_write_flush(nic); udelay(4);
iowrite8(ctrl | eesk, &nic->csr->eeprom_ctrl_lo);
e100_write_flush(nic); udelay(4);
/* Eeprom drives a dummy zero to EEDO after receiving
* complete address. Use this to adjust addr_len. */
ctrl = ioread8(&nic->csr->eeprom_ctrl_lo);
if(!(ctrl & eedo) && i > 16) {
*addr_len -= (i - 16);
i = 17;
}
data = (data << 1) | (ctrl & eedo ? 1 : 0);
}
/* Chip deselect */
iowrite8(0, &nic->csr->eeprom_ctrl_lo);
e100_write_flush(nic); udelay(4);
return cpu_to_le16(data);
};
/* Load entire EEPROM image into driver cache and validate checksum */
static int e100_eeprom_load(struct nic *nic)
{
u16 addr, addr_len = 8, checksum = 0;
/* Try reading with an 8-bit addr len to discover actual addr len */
e100_eeprom_read(nic, &addr_len, 0);
nic->eeprom_wc = 1 << addr_len;
for(addr = 0; addr < nic->eeprom_wc; addr++) {
nic->eeprom[addr] = e100_eeprom_read(nic, &addr_len, addr);
if(addr < nic->eeprom_wc - 1)
checksum += le16_to_cpu(nic->eeprom[addr]);
}
/* The checksum, stored in the last word, is calculated such that
* the sum of words should be 0xBABA */
if (cpu_to_le16(0xBABA - checksum) != nic->eeprom[nic->eeprom_wc - 1]) {
DPRINTK(PROBE, ERR, "EEPROM corrupted\n");
if (!eeprom_bad_csum_allow)
return -EAGAIN;
}
return 0;
}
/* Save (portion of) driver EEPROM cache to device and update checksum */
static int e100_eeprom_save(struct nic *nic, u16 start, u16 count)
{
u16 addr, addr_len = 8, checksum = 0;
/* Try reading with an 8-bit addr len to discover actual addr len */
e100_eeprom_read(nic, &addr_len, 0);
nic->eeprom_wc = 1 << addr_len;
if(start + count >= nic->eeprom_wc)
return -EINVAL;
for(addr = start; addr < start + count; addr++)
e100_eeprom_write(nic, addr_len, addr, nic->eeprom[addr]);
/* The checksum, stored in the last word, is calculated such that
* the sum of words should be 0xBABA */
for(addr = 0; addr < nic->eeprom_wc - 1; addr++)
checksum += le16_to_cpu(nic->eeprom[addr]);
nic->eeprom[nic->eeprom_wc - 1] = cpu_to_le16(0xBABA - checksum);
e100_eeprom_write(nic, addr_len, nic->eeprom_wc - 1,
nic->eeprom[nic->eeprom_wc - 1]);
return 0;
}
#define E100_WAIT_SCB_TIMEOUT 20000 /* we might have to wait 100ms!!! */
#define E100_WAIT_SCB_FAST 20 /* delay like the old code */
static int e100_exec_cmd(struct nic *nic, u8 cmd, dma_addr_t dma_addr)
{
unsigned long flags;
unsigned int i;
int err = 0;
spin_lock_irqsave(&nic->cmd_lock, flags);
/* Previous command is accepted when SCB clears */
for(i = 0; i < E100_WAIT_SCB_TIMEOUT; i++) {
if(likely(!ioread8(&nic->csr->scb.cmd_lo)))
break;
cpu_relax();
if(unlikely(i > E100_WAIT_SCB_FAST))
udelay(5);
}
if(unlikely(i == E100_WAIT_SCB_TIMEOUT)) {
err = -EAGAIN;
goto err_unlock;
}
if(unlikely(cmd != cuc_resume))
iowrite32(dma_addr, &nic->csr->scb.gen_ptr);
iowrite8(cmd, &nic->csr->scb.cmd_lo);
err_unlock:
spin_unlock_irqrestore(&nic->cmd_lock, flags);
return err;
}
static int e100_exec_cb(struct nic *nic, struct sk_buff *skb,
void (*cb_prepare)(struct nic *, struct cb *, struct sk_buff *))
{
struct cb *cb;
unsigned long flags;
int err = 0;
spin_lock_irqsave(&nic->cb_lock, flags);
if(unlikely(!nic->cbs_avail)) {
err = -ENOMEM;
goto err_unlock;
}
cb = nic->cb_to_use;
nic->cb_to_use = cb->next;
nic->cbs_avail--;
cb->skb = skb;
if(unlikely(!nic->cbs_avail))
err = -ENOSPC;
cb_prepare(nic, cb, skb);
/* Order is important otherwise we'll be in a race with h/w:
* set S-bit in current first, then clear S-bit in previous. */
cb->command |= cpu_to_le16(cb_s);
wmb();
cb->prev->command &= cpu_to_le16(~cb_s);
while(nic->cb_to_send != nic->cb_to_use) {
if(unlikely(e100_exec_cmd(nic, nic->cuc_cmd,
nic->cb_to_send->dma_addr))) {
/* Ok, here's where things get sticky. It's
* possible that we can't schedule the command
* because the controller is too busy, so
* let's just queue the command and try again
* when another command is scheduled. */
if(err == -ENOSPC) {
//request a reset
schedule_work(&nic->tx_timeout_task);
}
break;
} else {
nic->cuc_cmd = cuc_resume;
nic->cb_to_send = nic->cb_to_send->next;
}
}
err_unlock:
spin_unlock_irqrestore(&nic->cb_lock, flags);
return err;
}
static u16 mdio_ctrl(struct nic *nic, u32 addr, u32 dir, u32 reg, u16 data)
{
u32 data_out = 0;
unsigned int i;
unsigned long flags;
/*
* Stratus87247: we shouldn't be writing the MDI control
* register until the Ready bit shows True. Also, since
* manipulation of the MDI control registers is a multi-step
* procedure it should be done under lock.
*/
spin_lock_irqsave(&nic->mdio_lock, flags);
for (i = 100; i; --i) {
if (ioread32(&nic->csr->mdi_ctrl) & mdi_ready)
break;
udelay(20);
}
if (unlikely(!i)) {
printk("e100.mdio_ctrl(%s) won't go Ready\n",
nic->netdev->name );
spin_unlock_irqrestore(&nic->mdio_lock, flags);
return 0; /* No way to indicate timeout error */
}
iowrite32((reg << 16) | (addr << 21) | dir | data, &nic->csr->mdi_ctrl);
for (i = 0; i < 100; i++) {
udelay(20);
if ((data_out = ioread32(&nic->csr->mdi_ctrl)) & mdi_ready)
break;
}
spin_unlock_irqrestore(&nic->mdio_lock, flags);
DPRINTK(HW, DEBUG,
"%s:addr=%d, reg=%d, data_in=0x%04X, data_out=0x%04X\n",
dir == mdi_read ? "READ" : "WRITE", addr, reg, data, data_out);
return (u16)data_out;
}
static int mdio_read(struct net_device *netdev, int addr, int reg)
{
return mdio_ctrl(netdev_priv(netdev), addr, mdi_read, reg, 0);
}
static void mdio_write(struct net_device *netdev, int addr, int reg, int data)
{
mdio_ctrl(netdev_priv(netdev), addr, mdi_write, reg, data);
}
static void e100_get_defaults(struct nic *nic)
{
struct param_range rfds = { .min = 16, .max = 256, .count = 256 };
struct param_range cbs = { .min = 64, .max = 256, .count = 128 };
/* MAC type is encoded as rev ID; exception: ICH is treated as 82559 */
nic->mac = (nic->flags & ich) ? mac_82559_D101M : nic->pdev->revision;
if(nic->mac == mac_unknown)
nic->mac = mac_82557_D100_A;
nic->params.rfds = rfds;
nic->params.cbs = cbs;
/* Quadwords to DMA into FIFO before starting frame transmit */
nic->tx_threshold = 0xE0;
/* no interrupt for every tx completion, delay = 256us if not 557 */
nic->tx_command = cpu_to_le16(cb_tx | cb_tx_sf |
((nic->mac >= mac_82558_D101_A4) ? cb_cid : cb_i));
/* Template for a freshly allocated RFD */
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
nic->blank_rfd.command = 0;
nic->blank_rfd.rbd = cpu_to_le32(0xFFFFFFFF);
nic->blank_rfd.size = cpu_to_le16(VLAN_ETH_FRAME_LEN);
/* MII setup */
nic->mii.phy_id_mask = 0x1F;
nic->mii.reg_num_mask = 0x1F;
nic->mii.dev = nic->netdev;
nic->mii.mdio_read = mdio_read;
nic->mii.mdio_write = mdio_write;
}
static void e100_configure(struct nic *nic, struct cb *cb, struct sk_buff *skb)
{
struct config *config = &cb->u.config;
u8 *c = (u8 *)config;
cb->command = cpu_to_le16(cb_config);
memset(config, 0, sizeof(struct config));
config->byte_count = 0x16; /* bytes in this struct */
config->rx_fifo_limit = 0x8; /* bytes in FIFO before DMA */
config->direct_rx_dma = 0x1; /* reserved */
config->standard_tcb = 0x1; /* 1=standard, 0=extended */
config->standard_stat_counter = 0x1; /* 1=standard, 0=extended */
config->rx_discard_short_frames = 0x1; /* 1=discard, 0=pass */
config->tx_underrun_retry = 0x3; /* # of underrun retries */
config->mii_mode = 0x1; /* 1=MII mode, 0=503 mode */
config->pad10 = 0x6;
config->no_source_addr_insertion = 0x1; /* 1=no, 0=yes */
config->preamble_length = 0x2; /* 0=1, 1=3, 2=7, 3=15 bytes */
config->ifs = 0x6; /* x16 = inter frame spacing */
config->ip_addr_hi = 0xF2; /* ARP IP filter - not used */
config->pad15_1 = 0x1;
config->pad15_2 = 0x1;
config->crs_or_cdt = 0x0; /* 0=CRS only, 1=CRS or CDT */
config->fc_delay_hi = 0x40; /* time delay for fc frame */
config->tx_padding = 0x1; /* 1=pad short frames */
config->fc_priority_threshold = 0x7; /* 7=priority fc disabled */
config->pad18 = 0x1;
config->full_duplex_pin = 0x1; /* 1=examine FDX# pin */
config->pad20_1 = 0x1F;
config->fc_priority_location = 0x1; /* 1=byte#31, 0=byte#19 */
config->pad21_1 = 0x5;
config->adaptive_ifs = nic->adaptive_ifs;
config->loopback = nic->loopback;
if(nic->mii.force_media && nic->mii.full_duplex)
config->full_duplex_force = 0x1; /* 1=force, 0=auto */
if(nic->flags & promiscuous || nic->loopback) {
config->rx_save_bad_frames = 0x1; /* 1=save, 0=discard */
config->rx_discard_short_frames = 0x0; /* 1=discard, 0=save */
config->promiscuous_mode = 0x1; /* 1=on, 0=off */
}
if(nic->flags & multicast_all)
config->multicast_all = 0x1; /* 1=accept, 0=no */
/* disable WoL when up */
if(netif_running(nic->netdev) || !(nic->flags & wol_magic))
config->magic_packet_disable = 0x1; /* 1=off, 0=on */
if(nic->mac >= mac_82558_D101_A4) {
config->fc_disable = 0x1; /* 1=Tx fc off, 0=Tx fc on */
config->mwi_enable = 0x1; /* 1=enable, 0=disable */
config->standard_tcb = 0x0; /* 1=standard, 0=extended */
config->rx_long_ok = 0x1; /* 1=VLANs ok, 0=standard */
if (nic->mac >= mac_82559_D101M) {
config->tno_intr = 0x1; /* TCO stats enable */
/* Enable TCO in extended config */
if (nic->mac >= mac_82551_10) {
config->byte_count = 0x20; /* extended bytes */
config->rx_d102_mode = 0x1; /* GMRC for TCO */
}
} else {
config->standard_stat_counter = 0x0;
}
}
DPRINTK(HW, DEBUG, "[00-07]=%02X:%02X:%02X:%02X:%02X:%02X:%02X:%02X\n",
c[0], c[1], c[2], c[3], c[4], c[5], c[6], c[7]);
DPRINTK(HW, DEBUG, "[08-15]=%02X:%02X:%02X:%02X:%02X:%02X:%02X:%02X\n",
c[8], c[9], c[10], c[11], c[12], c[13], c[14], c[15]);
DPRINTK(HW, DEBUG, "[16-23]=%02X:%02X:%02X:%02X:%02X:%02X:%02X:%02X\n",
c[16], c[17], c[18], c[19], c[20], c[21], c[22], c[23]);
}
/********************************************************/
/* Micro code for 8086:1229 Rev 8 */
/********************************************************/
/* Parameter values for the D101M B-step */
#define D101M_CPUSAVER_TIMER_DWORD 78
#define D101M_CPUSAVER_BUNDLE_DWORD 65
#define D101M_CPUSAVER_MIN_SIZE_DWORD 126
#define D101M_B_RCVBUNDLE_UCODE \
{\
0x00550215, 0xFFFF0437, 0xFFFFFFFF, 0x06A70789, 0xFFFFFFFF, 0x0558FFFF, \
0x000C0001, 0x00101312, 0x000C0008, 0x00380216, \
0x0010009C, 0x00204056, 0x002380CC, 0x00380056, \
0x0010009C, 0x00244C0B, 0x00000800, 0x00124818, \
0x00380438, 0x00000000, 0x00140000, 0x00380555, \
0x00308000, 0x00100662, 0x00100561, 0x000E0408, \
0x00134861, 0x000C0002, 0x00103093, 0x00308000, \
0x00100624, 0x00100561, 0x000E0408, 0x00100861, \
0x000C007E, 0x00222C21, 0x000C0002, 0x00103093, \
0x00380C7A, 0x00080000, 0x00103090, 0x00380C7A, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x0010009C, 0x00244C2D, 0x00010004, 0x00041000, \
0x003A0437, 0x00044010, 0x0038078A, 0x00000000, \
0x00100099, 0x00206C7A, 0x0010009C, 0x00244C48, \
0x00130824, 0x000C0001, 0x00101213, 0x00260C75, \
0x00041000, 0x00010004, 0x00130826, 0x000C0006, \
0x002206A8, 0x0013C926, 0x00101313, 0x003806A8, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00080600, 0x00101B10, 0x00050004, 0x00100826, \
0x00101210, 0x00380C34, 0x00000000, 0x00000000, \
0x0021155B, 0x00100099, 0x00206559, 0x0010009C, \
0x00244559, 0x00130836, 0x000C0000, 0x00220C62, \
0x000C0001, 0x00101B13, 0x00229C0E, 0x00210C0E, \
0x00226C0E, 0x00216C0E, 0x0022FC0E, 0x00215C0E, \
0x00214C0E, 0x00380555, 0x00010004, 0x00041000, \
0x00278C67, 0x00040800, 0x00018100, 0x003A0437, \
0x00130826, 0x000C0001, 0x00220559, 0x00101313, \
0x00380559, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00130831, 0x0010090B, 0x00124813, \
0x000CFF80, 0x002606AB, 0x00041000, 0x00010004, \
0x003806A8, 0x00000000, 0x00000000, 0x00000000, \
}
/********************************************************/
/* Micro code for 8086:1229 Rev 9 */
/********************************************************/
/* Parameter values for the D101S */
#define D101S_CPUSAVER_TIMER_DWORD 78
#define D101S_CPUSAVER_BUNDLE_DWORD 67
#define D101S_CPUSAVER_MIN_SIZE_DWORD 128
#define D101S_RCVBUNDLE_UCODE \
{\
0x00550242, 0xFFFF047E, 0xFFFFFFFF, 0x06FF0818, 0xFFFFFFFF, 0x05A6FFFF, \
0x000C0001, 0x00101312, 0x000C0008, 0x00380243, \
0x0010009C, 0x00204056, 0x002380D0, 0x00380056, \
0x0010009C, 0x00244F8B, 0x00000800, 0x00124818, \
0x0038047F, 0x00000000, 0x00140000, 0x003805A3, \
0x00308000, 0x00100610, 0x00100561, 0x000E0408, \
0x00134861, 0x000C0002, 0x00103093, 0x00308000, \
0x00100624, 0x00100561, 0x000E0408, 0x00100861, \
0x000C007E, 0x00222FA1, 0x000C0002, 0x00103093, \
0x00380F90, 0x00080000, 0x00103090, 0x00380F90, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x0010009C, 0x00244FAD, 0x00010004, 0x00041000, \
0x003A047E, 0x00044010, 0x00380819, 0x00000000, \
0x00100099, 0x00206FFD, 0x0010009A, 0x0020AFFD, \
0x0010009C, 0x00244FC8, 0x00130824, 0x000C0001, \
0x00101213, 0x00260FF7, 0x00041000, 0x00010004, \
0x00130826, 0x000C0006, 0x00220700, 0x0013C926, \
0x00101313, 0x00380700, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00080600, 0x00101B10, 0x00050004, 0x00100826, \
0x00101210, 0x00380FB6, 0x00000000, 0x00000000, \
0x002115A9, 0x00100099, 0x002065A7, 0x0010009A, \
0x0020A5A7, 0x0010009C, 0x002445A7, 0x00130836, \
0x000C0000, 0x00220FE4, 0x000C0001, 0x00101B13, \
0x00229F8E, 0x00210F8E, 0x00226F8E, 0x00216F8E, \
0x0022FF8E, 0x00215F8E, 0x00214F8E, 0x003805A3, \
0x00010004, 0x00041000, 0x00278FE9, 0x00040800, \
0x00018100, 0x003A047E, 0x00130826, 0x000C0001, \
0x002205A7, 0x00101313, 0x003805A7, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00130831, \
0x0010090B, 0x00124813, 0x000CFF80, 0x00260703, \
0x00041000, 0x00010004, 0x00380700 \
}
/********************************************************/
/* Micro code for the 8086:1229 Rev F/10 */
/********************************************************/
/* Parameter values for the D102 E-step */
#define D102_E_CPUSAVER_TIMER_DWORD 42
#define D102_E_CPUSAVER_BUNDLE_DWORD 54
#define D102_E_CPUSAVER_MIN_SIZE_DWORD 46
#define D102_E_RCVBUNDLE_UCODE \
{\
0x007D028F, 0x0E4204F9, 0x14ED0C85, 0x14FA14E9, 0x0EF70E36, 0x1FFF1FFF, \
0x00E014B9, 0x00000000, 0x00000000, 0x00000000, \
0x00E014BD, 0x00000000, 0x00000000, 0x00000000, \
0x00E014D5, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00E014C1, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00E014C8, 0x00000000, 0x00000000, 0x00000000, \
0x00200600, 0x00E014EE, 0x00000000, 0x00000000, \
0x0030FF80, 0x00940E46, 0x00038200, 0x00102000, \
0x00E00E43, 0x00000000, 0x00000000, 0x00000000, \
0x00300006, 0x00E014FB, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00906E41, 0x00800E3C, 0x00E00E39, 0x00000000, \
0x00906EFD, 0x00900EFD, 0x00E00EF8, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
0x00000000, 0x00000000, 0x00000000, 0x00000000, \
}
static void e100_setup_ucode(struct nic *nic, struct cb *cb, struct sk_buff *skb)
{
/* *INDENT-OFF* */
static struct {
u32 ucode[UCODE_SIZE + 1];
u8 mac;
u8 timer_dword;
u8 bundle_dword;
u8 min_size_dword;
} ucode_opts[] = {
{ D101M_B_RCVBUNDLE_UCODE,
mac_82559_D101M,
D101M_CPUSAVER_TIMER_DWORD,
D101M_CPUSAVER_BUNDLE_DWORD,
D101M_CPUSAVER_MIN_SIZE_DWORD },
{ D101S_RCVBUNDLE_UCODE,
mac_82559_D101S,
D101S_CPUSAVER_TIMER_DWORD,
D101S_CPUSAVER_BUNDLE_DWORD,
D101S_CPUSAVER_MIN_SIZE_DWORD },
{ D102_E_RCVBUNDLE_UCODE,
mac_82551_F,
D102_E_CPUSAVER_TIMER_DWORD,
D102_E_CPUSAVER_BUNDLE_DWORD,
D102_E_CPUSAVER_MIN_SIZE_DWORD },
{ D102_E_RCVBUNDLE_UCODE,
mac_82551_10,
D102_E_CPUSAVER_TIMER_DWORD,
D102_E_CPUSAVER_BUNDLE_DWORD,
D102_E_CPUSAVER_MIN_SIZE_DWORD },
{ {0}, 0, 0, 0, 0}
}, *opts;
/* *INDENT-ON* */
/*************************************************************************
* CPUSaver parameters
*
* All CPUSaver parameters are 16-bit literals that are part of a
* "move immediate value" instruction. By changing the value of
* the literal in the instruction before the code is loaded, the
* driver can change the algorithm.
*
* INTDELAY - This loads the dead-man timer with its initial value.
* When this timer expires the interrupt is asserted, and the
* timer is reset each time a new packet is received. (see
* BUNDLEMAX below to set the limit on number of chained packets)
* The current default is 0x600 or 1536. Experiments show that
* the value should probably stay within the 0x200 - 0x1000.
*
* BUNDLEMAX -
* This sets the maximum number of frames that will be bundled. In
* some situations, such as the TCP windowing algorithm, it may be
* better to limit the growth of the bundle size than let it go as
* high as it can, because that could cause too much added latency.
* The default is six, because this is the number of packets in the
* default TCP window size. A value of 1 would make CPUSaver indicate
* an interrupt for every frame received. If you do not want to put
* a limit on the bundle size, set this value to xFFFF.
*
* BUNDLESMALL -
* This contains a bit-mask describing the minimum size frame that
* will be bundled. The default masks the lower 7 bits, which means
* that any frame less than 128 bytes in length will not be bundled,
* but will instead immediately generate an interrupt. This does
* not affect the current bundle in any way. Any frame that is 128
* bytes or large will be bundled normally. This feature is meant
* to provide immediate indication of ACK frames in a TCP environment.
* Customers were seeing poor performance when a machine with CPUSaver
* enabled was sending but not receiving. The delay introduced when
* the ACKs were received was enough to reduce total throughput, because
* the sender would sit idle until the ACK was finally seen.
*
* The current default is 0xFF80, which masks out the lower 7 bits.
* This means that any frame which is x7F (127) bytes or smaller
* will cause an immediate interrupt. Because this value must be a
* bit mask, there are only a few valid values that can be used. To
* turn this feature off, the driver can write the value xFFFF to the
* lower word of this instruction (in the same way that the other
* parameters are used). Likewise, a value of 0xF800 (2047) would
* cause an interrupt to be generated for every frame, because all
* standard Ethernet frames are <= 2047 bytes in length.
*************************************************************************/
/* if you wish to disable the ucode functionality, while maintaining the
* workarounds it provides, set the following defines to:
* BUNDLESMALL 0
* BUNDLEMAX 1
* INTDELAY 1
*/
#define BUNDLESMALL 1
#define BUNDLEMAX (u16)6
#define INTDELAY (u16)1536 /* 0x600 */
/* do not load u-code for ICH devices */
if (nic->flags & ich)
goto noloaducode;
/* Search for ucode match against h/w revision */
for (opts = ucode_opts; opts->mac; opts++) {
int i;
u32 *ucode = opts->ucode;
if (nic->mac != opts->mac)
continue;
/* Insert user-tunable settings */
ucode[opts->timer_dword] &= 0xFFFF0000;
ucode[opts->timer_dword] |= INTDELAY;
ucode[opts->bundle_dword] &= 0xFFFF0000;
ucode[opts->bundle_dword] |= BUNDLEMAX;
ucode[opts->min_size_dword] &= 0xFFFF0000;
ucode[opts->min_size_dword] |= (BUNDLESMALL) ? 0xFFFF : 0xFF80;
for (i = 0; i < UCODE_SIZE; i++)
cb->u.ucode[i] = cpu_to_le32(ucode[i]);
cb->command = cpu_to_le16(cb_ucode | cb_el);
return;
}
noloaducode:
cb->command = cpu_to_le16(cb_nop | cb_el);
}
static inline int e100_exec_cb_wait(struct nic *nic, struct sk_buff *skb,
void (*cb_prepare)(struct nic *, struct cb *, struct sk_buff *))
{
int err = 0, counter = 50;
struct cb *cb = nic->cb_to_clean;
if ((err = e100_exec_cb(nic, NULL, e100_setup_ucode)))
DPRINTK(PROBE,ERR, "ucode cmd failed with error %d\n", err);
/* must restart cuc */
nic->cuc_cmd = cuc_start;
/* wait for completion */
e100_write_flush(nic);
udelay(10);
/* wait for possibly (ouch) 500ms */
while (!(cb->status & cpu_to_le16(cb_complete))) {
msleep(10);
if (!--counter) break;
}
/* ack any interrupts, something could have been set */
iowrite8(~0, &nic->csr->scb.stat_ack);
/* if the command failed, or is not OK, notify and return */
if (!counter || !(cb->status & cpu_to_le16(cb_ok))) {
DPRINTK(PROBE,ERR, "ucode load failed\n");
err = -EPERM;
}
return err;
}
static void e100_setup_iaaddr(struct nic *nic, struct cb *cb,
struct sk_buff *skb)
{
cb->command = cpu_to_le16(cb_iaaddr);
memcpy(cb->u.iaaddr, nic->netdev->dev_addr, ETH_ALEN);
}
static void e100_dump(struct nic *nic, struct cb *cb, struct sk_buff *skb)
{
cb->command = cpu_to_le16(cb_dump);
cb->u.dump_buffer_addr = cpu_to_le32(nic->dma_addr +
offsetof(struct mem, dump_buf));
}
#define NCONFIG_AUTO_SWITCH 0x0080
#define MII_NSC_CONG MII_RESV1
#define NSC_CONG_ENABLE 0x0100
#define NSC_CONG_TXREADY 0x0400
#define ADVERTISE_FC_SUPPORTED 0x0400
static int e100_phy_init(struct nic *nic)
{
struct net_device *netdev = nic->netdev;
u32 addr;
u16 bmcr, stat, id_lo, id_hi, cong;
/* Discover phy addr by searching addrs in order {1,0,2,..., 31} */
for(addr = 0; addr < 32; addr++) {
nic->mii.phy_id = (addr == 0) ? 1 : (addr == 1) ? 0 : addr;
bmcr = mdio_read(netdev, nic->mii.phy_id, MII_BMCR);
stat = mdio_read(netdev, nic->mii.phy_id, MII_BMSR);
stat = mdio_read(netdev, nic->mii.phy_id, MII_BMSR);
if(!((bmcr == 0xFFFF) || ((stat == 0) && (bmcr == 0))))
break;
}
DPRINTK(HW, DEBUG, "phy_addr = %d\n", nic->mii.phy_id);
if(addr == 32)
return -EAGAIN;
/* Selected the phy and isolate the rest */
for(addr = 0; addr < 32; addr++) {
if(addr != nic->mii.phy_id) {
mdio_write(netdev, addr, MII_BMCR, BMCR_ISOLATE);
} else {
bmcr = mdio_read(netdev, addr, MII_BMCR);
mdio_write(netdev, addr, MII_BMCR,
bmcr & ~BMCR_ISOLATE);
}
}
/* Get phy ID */
id_lo = mdio_read(netdev, nic->mii.phy_id, MII_PHYSID1);
id_hi = mdio_read(netdev, nic->mii.phy_id, MII_PHYSID2);
nic->phy = (u32)id_hi << 16 | (u32)id_lo;
DPRINTK(HW, DEBUG, "phy ID = 0x%08X\n", nic->phy);
/* Handle National tx phys */
#define NCS_PHY_MODEL_MASK 0xFFF0FFFF
if((nic->phy & NCS_PHY_MODEL_MASK) == phy_nsc_tx) {
/* Disable congestion control */
cong = mdio_read(netdev, nic->mii.phy_id, MII_NSC_CONG);
cong |= NSC_CONG_TXREADY;
cong &= ~NSC_CONG_ENABLE;
mdio_write(netdev, nic->mii.phy_id, MII_NSC_CONG, cong);
}
if((nic->mac >= mac_82550_D102) || ((nic->flags & ich) &&
(mdio_read(netdev, nic->mii.phy_id, MII_TPISTATUS) & 0x8000) &&
!(nic->eeprom[eeprom_cnfg_mdix] & eeprom_mdix_enabled))) {
/* enable/disable MDI/MDI-X auto-switching. */
mdio_write(netdev, nic->mii.phy_id, MII_NCONFIG,
nic->mii.force_media ? 0 : NCONFIG_AUTO_SWITCH);
}
return 0;
}
static int e100_hw_init(struct nic *nic)
{
int err;
e100_hw_reset(nic);
DPRINTK(HW, ERR, "e100_hw_init\n");
if(!in_interrupt() && (err = e100_self_test(nic)))
return err;
if((err = e100_phy_init(nic)))
return err;
if((err = e100_exec_cmd(nic, cuc_load_base, 0)))
return err;
if((err = e100_exec_cmd(nic, ruc_load_base, 0)))
return err;
if ((err = e100_exec_cb_wait(nic, NULL, e100_setup_ucode)))
return err;
if((err = e100_exec_cb(nic, NULL, e100_configure)))
return err;
if((err = e100_exec_cb(nic, NULL, e100_setup_iaaddr)))
return err;
if((err = e100_exec_cmd(nic, cuc_dump_addr,
nic->dma_addr + offsetof(struct mem, stats))))
return err;
if((err = e100_exec_cmd(nic, cuc_dump_reset, 0)))
return err;
e100_disable_irq(nic);
return 0;
}
static void e100_multi(struct nic *nic, struct cb *cb, struct sk_buff *skb)
{
struct net_device *netdev = nic->netdev;
struct dev_mc_list *list = netdev->mc_list;
u16 i, count = min(netdev->mc_count, E100_MAX_MULTICAST_ADDRS);
cb->command = cpu_to_le16(cb_multi);
cb->u.multi.count = cpu_to_le16(count * ETH_ALEN);
for(i = 0; list && i < count; i++, list = list->next)
memcpy(&cb->u.multi.addr[i*ETH_ALEN], &list->dmi_addr,
ETH_ALEN);
}
static void e100_set_multicast_list(struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
DPRINTK(HW, DEBUG, "mc_count=%d, flags=0x%04X\n",
netdev->mc_count, netdev->flags);
if(netdev->flags & IFF_PROMISC)
nic->flags |= promiscuous;
else
nic->flags &= ~promiscuous;
if(netdev->flags & IFF_ALLMULTI ||
netdev->mc_count > E100_MAX_MULTICAST_ADDRS)
nic->flags |= multicast_all;
else
nic->flags &= ~multicast_all;
e100_exec_cb(nic, NULL, e100_configure);
e100_exec_cb(nic, NULL, e100_multi);
}
static void e100_update_stats(struct nic *nic)
{
struct net_device *dev = nic->netdev;
struct net_device_stats *ns = &dev->stats;
struct stats *s = &nic->mem->stats;
__le32 *complete = (nic->mac < mac_82558_D101_A4) ? &s->fc_xmt_pause :
(nic->mac < mac_82559_D101M) ? (__le32 *)&s->xmt_tco_frames :
&s->complete;
/* Device's stats reporting may take several microseconds to
* complete, so we're always waiting for results of the
* previous command. */
if(*complete == cpu_to_le32(cuc_dump_reset_complete)) {
*complete = 0;
nic->tx_frames = le32_to_cpu(s->tx_good_frames);
nic->tx_collisions = le32_to_cpu(s->tx_total_collisions);
ns->tx_aborted_errors += le32_to_cpu(s->tx_max_collisions);
ns->tx_window_errors += le32_to_cpu(s->tx_late_collisions);
ns->tx_carrier_errors += le32_to_cpu(s->tx_lost_crs);
ns->tx_fifo_errors += le32_to_cpu(s->tx_underruns);
ns->collisions += nic->tx_collisions;
ns->tx_errors += le32_to_cpu(s->tx_max_collisions) +
le32_to_cpu(s->tx_lost_crs);
ns->rx_length_errors += le32_to_cpu(s->rx_short_frame_errors) +
nic->rx_over_length_errors;
ns->rx_crc_errors += le32_to_cpu(s->rx_crc_errors);
ns->rx_frame_errors += le32_to_cpu(s->rx_alignment_errors);
ns->rx_over_errors += le32_to_cpu(s->rx_overrun_errors);
ns->rx_fifo_errors += le32_to_cpu(s->rx_overrun_errors);
ns->rx_missed_errors += le32_to_cpu(s->rx_resource_errors);
ns->rx_errors += le32_to_cpu(s->rx_crc_errors) +
le32_to_cpu(s->rx_alignment_errors) +
le32_to_cpu(s->rx_short_frame_errors) +
le32_to_cpu(s->rx_cdt_errors);
nic->tx_deferred += le32_to_cpu(s->tx_deferred);
nic->tx_single_collisions +=
le32_to_cpu(s->tx_single_collisions);
nic->tx_multiple_collisions +=
le32_to_cpu(s->tx_multiple_collisions);
if(nic->mac >= mac_82558_D101_A4) {
nic->tx_fc_pause += le32_to_cpu(s->fc_xmt_pause);
nic->rx_fc_pause += le32_to_cpu(s->fc_rcv_pause);
nic->rx_fc_unsupported +=
le32_to_cpu(s->fc_rcv_unsupported);
if(nic->mac >= mac_82559_D101M) {
nic->tx_tco_frames +=
le16_to_cpu(s->xmt_tco_frames);
nic->rx_tco_frames +=
le16_to_cpu(s->rcv_tco_frames);
}
}
}
if(e100_exec_cmd(nic, cuc_dump_reset, 0))
DPRINTK(TX_ERR, DEBUG, "exec cuc_dump_reset failed\n");
}
static void e100_adjust_adaptive_ifs(struct nic *nic, int speed, int duplex)
{
/* Adjust inter-frame-spacing (IFS) between two transmits if
* we're getting collisions on a half-duplex connection. */
if(duplex == DUPLEX_HALF) {
u32 prev = nic->adaptive_ifs;
u32 min_frames = (speed == SPEED_100) ? 1000 : 100;
if((nic->tx_frames / 32 < nic->tx_collisions) &&
(nic->tx_frames > min_frames)) {
if(nic->adaptive_ifs < 60)
nic->adaptive_ifs += 5;
} else if (nic->tx_frames < min_frames) {
if(nic->adaptive_ifs >= 5)
nic->adaptive_ifs -= 5;
}
if(nic->adaptive_ifs != prev)
e100_exec_cb(nic, NULL, e100_configure);
}
}
static void e100_watchdog(unsigned long data)
{
struct nic *nic = (struct nic *)data;
struct ethtool_cmd cmd;
DPRINTK(TIMER, DEBUG, "right now = %ld\n", jiffies);
/* mii library handles link maintenance tasks */
mii_ethtool_gset(&nic->mii, &cmd);
if(mii_link_ok(&nic->mii) && !netif_carrier_ok(nic->netdev)) {
DPRINTK(LINK, INFO, "link up, %sMbps, %s-duplex\n",
cmd.speed == SPEED_100 ? "100" : "10",
cmd.duplex == DUPLEX_FULL ? "full" : "half");
} else if(!mii_link_ok(&nic->mii) && netif_carrier_ok(nic->netdev)) {
DPRINTK(LINK, INFO, "link down\n");
}
mii_check_link(&nic->mii);
/* Software generated interrupt to recover from (rare) Rx
* allocation failure.
* Unfortunately have to use a spinlock to not re-enable interrupts
* accidentally, due to hardware that shares a register between the
* interrupt mask bit and the SW Interrupt generation bit */
spin_lock_irq(&nic->cmd_lock);
iowrite8(ioread8(&nic->csr->scb.cmd_hi) | irq_sw_gen,&nic->csr->scb.cmd_hi);
e100_write_flush(nic);
spin_unlock_irq(&nic->cmd_lock);
e100_update_stats(nic);
e100_adjust_adaptive_ifs(nic, cmd.speed, cmd.duplex);
if(nic->mac <= mac_82557_D100_C)
/* Issue a multicast command to workaround a 557 lock up */
e100_set_multicast_list(nic->netdev);
if(nic->flags & ich && cmd.speed==SPEED_10 && cmd.duplex==DUPLEX_HALF)
/* Need SW workaround for ICH[x] 10Mbps/half duplex Tx hang. */
nic->flags |= ich_10h_workaround;
else
nic->flags &= ~ich_10h_workaround;
mod_timer(&nic->watchdog,
round_jiffies(jiffies + E100_WATCHDOG_PERIOD));
}
static void e100_xmit_prepare(struct nic *nic, struct cb *cb,
struct sk_buff *skb)
{
cb->command = nic->tx_command;
/* interrupt every 16 packets regardless of delay */
if((nic->cbs_avail & ~15) == nic->cbs_avail)
cb->command |= cpu_to_le16(cb_i);
cb->u.tcb.tbd_array = cb->dma_addr + offsetof(struct cb, u.tcb.tbd);
cb->u.tcb.tcb_byte_count = 0;
cb->u.tcb.threshold = nic->tx_threshold;
cb->u.tcb.tbd_count = 1;
cb->u.tcb.tbd.buf_addr = cpu_to_le32(pci_map_single(nic->pdev,
skb->data, skb->len, PCI_DMA_TODEVICE));
/* check for mapping failure? */
cb->u.tcb.tbd.size = cpu_to_le16(skb->len);
}
static int e100_xmit_frame(struct sk_buff *skb, struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
int err;
if(nic->flags & ich_10h_workaround) {
/* SW workaround for ICH[x] 10Mbps/half duplex Tx hang.
Issue a NOP command followed by a 1us delay before
issuing the Tx command. */
if(e100_exec_cmd(nic, cuc_nop, 0))
DPRINTK(TX_ERR, DEBUG, "exec cuc_nop failed\n");
udelay(1);
}
err = e100_exec_cb(nic, skb, e100_xmit_prepare);
switch(err) {
case -ENOSPC:
/* We queued the skb, but now we're out of space. */
DPRINTK(TX_ERR, DEBUG, "No space for CB\n");
netif_stop_queue(netdev);
break;
case -ENOMEM:
/* This is a hard error - log it. */
DPRINTK(TX_ERR, DEBUG, "Out of Tx resources, returning skb\n");
netif_stop_queue(netdev);
return 1;
}
netdev->trans_start = jiffies;
return 0;
}
static int e100_tx_clean(struct nic *nic)
{
struct net_device *dev = nic->netdev;
struct cb *cb;
int tx_cleaned = 0;
spin_lock(&nic->cb_lock);
/* Clean CBs marked complete */
for(cb = nic->cb_to_clean;
cb->status & cpu_to_le16(cb_complete);
cb = nic->cb_to_clean = cb->next) {
DPRINTK(TX_DONE, DEBUG, "cb[%d]->status = 0x%04X\n",
(int)(((void*)cb - (void*)nic->cbs)/sizeof(struct cb)),
cb->status);
if(likely(cb->skb != NULL)) {
dev->stats.tx_packets++;
dev->stats.tx_bytes += cb->skb->len;
pci_unmap_single(nic->pdev,
le32_to_cpu(cb->u.tcb.tbd.buf_addr),
le16_to_cpu(cb->u.tcb.tbd.size),
PCI_DMA_TODEVICE);
dev_kfree_skb_any(cb->skb);
cb->skb = NULL;
tx_cleaned = 1;
}
cb->status = 0;
nic->cbs_avail++;
}
spin_unlock(&nic->cb_lock);
/* Recover from running out of Tx resources in xmit_frame */
if(unlikely(tx_cleaned && netif_queue_stopped(nic->netdev)))
netif_wake_queue(nic->netdev);
return tx_cleaned;
}
static void e100_clean_cbs(struct nic *nic)
{
if(nic->cbs) {
while(nic->cbs_avail != nic->params.cbs.count) {
struct cb *cb = nic->cb_to_clean;
if(cb->skb) {
pci_unmap_single(nic->pdev,
le32_to_cpu(cb->u.tcb.tbd.buf_addr),
le16_to_cpu(cb->u.tcb.tbd.size),
PCI_DMA_TODEVICE);
dev_kfree_skb(cb->skb);
}
nic->cb_to_clean = nic->cb_to_clean->next;
nic->cbs_avail++;
}
pci_free_consistent(nic->pdev,
sizeof(struct cb) * nic->params.cbs.count,
nic->cbs, nic->cbs_dma_addr);
nic->cbs = NULL;
nic->cbs_avail = 0;
}
nic->cuc_cmd = cuc_start;
nic->cb_to_use = nic->cb_to_send = nic->cb_to_clean =
nic->cbs;
}
static int e100_alloc_cbs(struct nic *nic)
{
struct cb *cb;
unsigned int i, count = nic->params.cbs.count;
nic->cuc_cmd = cuc_start;
nic->cb_to_use = nic->cb_to_send = nic->cb_to_clean = NULL;
nic->cbs_avail = 0;
nic->cbs = pci_alloc_consistent(nic->pdev,
sizeof(struct cb) * count, &nic->cbs_dma_addr);
if(!nic->cbs)
return -ENOMEM;
for(cb = nic->cbs, i = 0; i < count; cb++, i++) {
cb->next = (i + 1 < count) ? cb + 1 : nic->cbs;
cb->prev = (i == 0) ? nic->cbs + count - 1 : cb - 1;
cb->dma_addr = nic->cbs_dma_addr + i * sizeof(struct cb);
cb->link = cpu_to_le32(nic->cbs_dma_addr +
((i+1) % count) * sizeof(struct cb));
cb->skb = NULL;
}
nic->cb_to_use = nic->cb_to_send = nic->cb_to_clean = nic->cbs;
nic->cbs_avail = count;
return 0;
}
static inline void e100_start_receiver(struct nic *nic, struct rx *rx)
{
if(!nic->rxs) return;
if(RU_SUSPENDED != nic->ru_running) return;
/* handle init time starts */
if(!rx) rx = nic->rxs;
/* (Re)start RU if suspended or idle and RFA is non-NULL */
if(rx->skb) {
e100_exec_cmd(nic, ruc_start, rx->dma_addr);
nic->ru_running = RU_RUNNING;
}
}
#define RFD_BUF_LEN (sizeof(struct rfd) + VLAN_ETH_FRAME_LEN)
static int e100_rx_alloc_skb(struct nic *nic, struct rx *rx)
{
if(!(rx->skb = netdev_alloc_skb(nic->netdev, RFD_BUF_LEN + NET_IP_ALIGN)))
return -ENOMEM;
/* Align, init, and map the RFD. */
skb_reserve(rx->skb, NET_IP_ALIGN);
skb_copy_to_linear_data(rx->skb, &nic->blank_rfd, sizeof(struct rfd));
rx->dma_addr = pci_map_single(nic->pdev, rx->skb->data,
RFD_BUF_LEN, PCI_DMA_BIDIRECTIONAL);
dma-mapping: add the device argument to dma_mapping_error() Add per-device dma_mapping_ops support for CONFIG_X86_64 as POWER architecture does: This enables us to cleanly fix the Calgary IOMMU issue that some devices are not behind the IOMMU (http://lkml.org/lkml/2008/5/8/423). I think that per-device dma_mapping_ops support would be also helpful for KVM people to support PCI passthrough but Andi thinks that this makes it difficult to support the PCI passthrough (see the above thread). So I CC'ed this to KVM camp. Comments are appreciated. A pointer to dma_mapping_ops to struct dev_archdata is added. If the pointer is non NULL, DMA operations in asm/dma-mapping.h use it. If it's NULL, the system-wide dma_ops pointer is used as before. If it's useful for KVM people, I plan to implement a mechanism to register a hook called when a new pci (or dma capable) device is created (it works with hot plugging). It enables IOMMUs to set up an appropriate dma_mapping_ops per device. The major obstacle is that dma_mapping_error doesn't take a pointer to the device unlike other DMA operations. So x86 can't have dma_mapping_ops per device. Note all the POWER IOMMUs use the same dma_mapping_error function so this is not a problem for POWER but x86 IOMMUs use different dma_mapping_error functions. The first patch adds the device argument to dma_mapping_error. The patch is trivial but large since it touches lots of drivers and dma-mapping.h in all the architecture. This patch: dma_mapping_error() doesn't take a pointer to the device unlike other DMA operations. So we can't have dma_mapping_ops per device. Note that POWER already has dma_mapping_ops per device but all the POWER IOMMUs use the same dma_mapping_error function. x86 IOMMUs use device argument. [akpm@linux-foundation.org: fix sge] [akpm@linux-foundation.org: fix svc_rdma] [akpm@linux-foundation.org: build fix] [akpm@linux-foundation.org: fix bnx2x] [akpm@linux-foundation.org: fix s2io] [akpm@linux-foundation.org: fix pasemi_mac] [akpm@linux-foundation.org: fix sdhci] [akpm@linux-foundation.org: build fix] [akpm@linux-foundation.org: fix sparc] [akpm@linux-foundation.org: fix ibmvscsi] Signed-off-by: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp> Cc: Muli Ben-Yehuda <muli@il.ibm.com> Cc: Andi Kleen <andi@firstfloor.org> Cc: Thomas Gleixner <tglx@linutronix.de> Cc: Ingo Molnar <mingo@elte.hu> Cc: Avi Kivity <avi@qumranet.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-07-26 10:44:49 +08:00
if (pci_dma_mapping_error(nic->pdev, rx->dma_addr)) {
dev_kfree_skb_any(rx->skb);
rx->skb = NULL;
rx->dma_addr = 0;
return -ENOMEM;
}
/* Link the RFD to end of RFA by linking previous RFD to
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
* this one. We are safe to touch the previous RFD because
* it is protected by the before last buffer's el bit being set */
if (rx->prev->skb) {
struct rfd *prev_rfd = (struct rfd *)rx->prev->skb->data;
put_unaligned_le32(rx->dma_addr, &prev_rfd->link);
pci_dma_sync_single_for_device(nic->pdev, rx->prev->dma_addr,
sizeof(struct rfd), PCI_DMA_TODEVICE);
}
return 0;
}
static int e100_rx_indicate(struct nic *nic, struct rx *rx,
unsigned int *work_done, unsigned int work_to_do)
{
struct net_device *dev = nic->netdev;
struct sk_buff *skb = rx->skb;
struct rfd *rfd = (struct rfd *)skb->data;
u16 rfd_status, actual_size;
if(unlikely(work_done && *work_done >= work_to_do))
return -EAGAIN;
/* Need to sync before taking a peek at cb_complete bit */
pci_dma_sync_single_for_cpu(nic->pdev, rx->dma_addr,
sizeof(struct rfd), PCI_DMA_FROMDEVICE);
rfd_status = le16_to_cpu(rfd->status);
DPRINTK(RX_STATUS, DEBUG, "status=0x%04X\n", rfd_status);
/* If data isn't ready, nothing to indicate */
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
if (unlikely(!(rfd_status & cb_complete))) {
/* If the next buffer has the el bit, but we think the receiver
* is still running, check to see if it really stopped while
* we had interrupts off.
* This allows for a fast restart without re-enabling
* interrupts */
if ((le16_to_cpu(rfd->command) & cb_el) &&
(RU_RUNNING == nic->ru_running))
if (ioread8(&nic->csr->scb.status) & rus_no_res)
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
nic->ru_running = RU_SUSPENDED;
return -ENODATA;
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
}
/* Get actual data size */
actual_size = le16_to_cpu(rfd->actual_size) & 0x3FFF;
if(unlikely(actual_size > RFD_BUF_LEN - sizeof(struct rfd)))
actual_size = RFD_BUF_LEN - sizeof(struct rfd);
/* Get data */
pci_unmap_single(nic->pdev, rx->dma_addr,
RFD_BUF_LEN, PCI_DMA_FROMDEVICE);
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
/* If this buffer has the el bit, but we think the receiver
* is still running, check to see if it really stopped while
* we had interrupts off.
* This allows for a fast restart without re-enabling interrupts.
* This can happen when the RU sees the size change but also sees
* the el bit set. */
if ((le16_to_cpu(rfd->command) & cb_el) &&
(RU_RUNNING == nic->ru_running)) {
if (ioread8(&nic->csr->scb.status) & rus_no_res)
nic->ru_running = RU_SUSPENDED;
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
}
/* Pull off the RFD and put the actual data (minus eth hdr) */
skb_reserve(skb, sizeof(struct rfd));
skb_put(skb, actual_size);
skb->protocol = eth_type_trans(skb, nic->netdev);
if(unlikely(!(rfd_status & cb_ok))) {
/* Don't indicate if hardware indicates errors */
dev_kfree_skb_any(skb);
} else if(actual_size > ETH_DATA_LEN + VLAN_ETH_HLEN) {
/* Don't indicate oversized frames */
nic->rx_over_length_errors++;
dev_kfree_skb_any(skb);
} else {
dev->stats.rx_packets++;
dev->stats.rx_bytes += actual_size;
nic->netdev->last_rx = jiffies;
netif_receive_skb(skb);
if(work_done)
(*work_done)++;
}
rx->skb = NULL;
return 0;
}
static void e100_rx_clean(struct nic *nic, unsigned int *work_done,
unsigned int work_to_do)
{
struct rx *rx;
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
int restart_required = 0, err = 0;
struct rx *old_before_last_rx, *new_before_last_rx;
struct rfd *old_before_last_rfd, *new_before_last_rfd;
/* Indicate newly arrived packets */
for(rx = nic->rx_to_clean; rx->skb; rx = nic->rx_to_clean = rx->next) {
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
err = e100_rx_indicate(nic, rx, work_done, work_to_do);
/* Hit quota or no more to clean */
if (-EAGAIN == err || -ENODATA == err)
break;
}
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
/* On EAGAIN, hit quota so have more work to do, restart once
* cleanup is complete.
* Else, are we already rnr? then pay attention!!! this ensures that
* the state machine progression never allows a start with a
* partially cleaned list, avoiding a race between hardware
* and rx_to_clean when in NAPI mode */
if (-EAGAIN != err && RU_SUSPENDED == nic->ru_running)
restart_required = 1;
old_before_last_rx = nic->rx_to_use->prev->prev;
old_before_last_rfd = (struct rfd *)old_before_last_rx->skb->data;
/* Alloc new skbs to refill list */
for(rx = nic->rx_to_use; !rx->skb; rx = nic->rx_to_use = rx->next) {
if(unlikely(e100_rx_alloc_skb(nic, rx)))
break; /* Better luck next time (see watchdog) */
}
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
new_before_last_rx = nic->rx_to_use->prev->prev;
if (new_before_last_rx != old_before_last_rx) {
/* Set the el-bit on the buffer that is before the last buffer.
* This lets us update the next pointer on the last buffer
* without worrying about hardware touching it.
* We set the size to 0 to prevent hardware from touching this
* buffer.
* When the hardware hits the before last buffer with el-bit
* and size of 0, it will RNR interrupt, the RUS will go into
* the No Resources state. It will not complete nor write to
* this buffer. */
new_before_last_rfd =
(struct rfd *)new_before_last_rx->skb->data;
new_before_last_rfd->size = 0;
new_before_last_rfd->command |= cpu_to_le16(cb_el);
pci_dma_sync_single_for_device(nic->pdev,
new_before_last_rx->dma_addr, sizeof(struct rfd),
PCI_DMA_TODEVICE);
/* Now that we have a new stopping point, we can clear the old
* stopping point. We must sync twice to get the proper
* ordering on the hardware side of things. */
old_before_last_rfd->command &= ~cpu_to_le16(cb_el);
pci_dma_sync_single_for_device(nic->pdev,
old_before_last_rx->dma_addr, sizeof(struct rfd),
PCI_DMA_TODEVICE);
old_before_last_rfd->size = cpu_to_le16(VLAN_ETH_FRAME_LEN);
pci_dma_sync_single_for_device(nic->pdev,
old_before_last_rx->dma_addr, sizeof(struct rfd),
PCI_DMA_TODEVICE);
}
if(restart_required) {
// ack the rnr?
iowrite8(stat_ack_rnr, &nic->csr->scb.stat_ack);
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
e100_start_receiver(nic, nic->rx_to_clean);
if(work_done)
(*work_done)++;
}
}
static void e100_rx_clean_list(struct nic *nic)
{
struct rx *rx;
unsigned int i, count = nic->params.rfds.count;
nic->ru_running = RU_UNINITIALIZED;
if(nic->rxs) {
for(rx = nic->rxs, i = 0; i < count; rx++, i++) {
if(rx->skb) {
pci_unmap_single(nic->pdev, rx->dma_addr,
RFD_BUF_LEN, PCI_DMA_FROMDEVICE);
dev_kfree_skb(rx->skb);
}
}
kfree(nic->rxs);
nic->rxs = NULL;
}
nic->rx_to_use = nic->rx_to_clean = NULL;
}
static int e100_rx_alloc_list(struct nic *nic)
{
struct rx *rx;
unsigned int i, count = nic->params.rfds.count;
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
struct rfd *before_last;
nic->rx_to_use = nic->rx_to_clean = NULL;
nic->ru_running = RU_UNINITIALIZED;
if(!(nic->rxs = kcalloc(count, sizeof(struct rx), GFP_ATOMIC)))
return -ENOMEM;
for(rx = nic->rxs, i = 0; i < count; rx++, i++) {
rx->next = (i + 1 < count) ? rx + 1 : nic->rxs;
rx->prev = (i == 0) ? nic->rxs + count - 1 : rx - 1;
if(e100_rx_alloc_skb(nic, rx)) {
e100_rx_clean_list(nic);
return -ENOMEM;
}
}
Fix e100 on systems that have cache incoherent DMA On the systems that have cache incoherent DMA, including ARM, there is a race condition between software allocating a new receive buffer and hardware writing into a buffer. The two race on touching the last Receive Frame Descriptor (RFD). It has its el-bit set and its next link equal to 0. When hardware encounters this buffer it attempts to write data to it and then update Status Word bits and Actual Count in the RFD. At the same time software may try to clear the el-bit and set the link address to a new buffer. Since the entire RFD is once cache-line, the two write operations can collide. This can lead to the receive unit stalling or interpreting random memory as its receive area. The fix is to set the el-bit on and the size to 0 on the next to last buffer in the chain. When the hardware encounters this buffer it stops and does not write to it at all. The hardware issues an RNR interrupt with the receive unit in the No Resources state. Software can write to the tail of the list because it knows hardware will stop on the previous descriptor that was marked as the end of list. Once it has a new next to last buffer prepared, it can clear the el-bit and set the size on the previous one. The race on this buffer is safe since the link already points to a valid next buffer and the software can handle the race setting the size (assuming aligned 16 bit writes are atomic with respect to the DMA read). If the hardware sees the el-bit cleared without the size set, it will move on to the next buffer and skip this one. If it sees the size set but the el-bit still set, it will complete that buffer and then RNR interrupt and wait. Signed-off-by: David Acker <dacker@roinet.com> Signed-off-by: Auke Kok <auke-jan.h.kok@intel.com> Signed-off-by: Jeff Garzik <jeff@garzik.org>
2007-11-09 02:17:41 +08:00
/* Set the el-bit on the buffer that is before the last buffer.
* This lets us update the next pointer on the last buffer without
* worrying about hardware touching it.
* We set the size to 0 to prevent hardware from touching this buffer.
* When the hardware hits the before last buffer with el-bit and size
* of 0, it will RNR interrupt, the RU will go into the No Resources
* state. It will not complete nor write to this buffer. */
rx = nic->rxs->prev->prev;
before_last = (struct rfd *)rx->skb->data;
before_last->command |= cpu_to_le16(cb_el);
before_last->size = 0;
pci_dma_sync_single_for_device(nic->pdev, rx->dma_addr,
sizeof(struct rfd), PCI_DMA_TODEVICE);
nic->rx_to_use = nic->rx_to_clean = nic->rxs;
nic->ru_running = RU_SUSPENDED;
return 0;
}
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 21:55:46 +08:00
static irqreturn_t e100_intr(int irq, void *dev_id)
{
struct net_device *netdev = dev_id;
struct nic *nic = netdev_priv(netdev);
u8 stat_ack = ioread8(&nic->csr->scb.stat_ack);
DPRINTK(INTR, DEBUG, "stat_ack = 0x%02X\n", stat_ack);
if(stat_ack == stat_ack_not_ours || /* Not our interrupt */
stat_ack == stat_ack_not_present) /* Hardware is ejected */
return IRQ_NONE;
/* Ack interrupt(s) */
iowrite8(stat_ack, &nic->csr->scb.stat_ack);
/* We hit Receive No Resource (RNR); restart RU after cleaning */
if(stat_ack & stat_ack_rnr)
nic->ru_running = RU_SUSPENDED;
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
if(likely(netif_rx_schedule_prep(netdev, &nic->napi))) {
e100_disable_irq(nic);
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
__netif_rx_schedule(netdev, &nic->napi);
}
return IRQ_HANDLED;
}
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
static int e100_poll(struct napi_struct *napi, int budget)
{
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
struct nic *nic = container_of(napi, struct nic, napi);
struct net_device *netdev = nic->netdev;
unsigned int work_done = 0;
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
e100_rx_clean(nic, &work_done, budget);
e100_tx_clean(nic);
/* If budget not fully consumed, exit the polling mode */
if (work_done < budget) {
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
netif_rx_complete(netdev, napi);
e100_enable_irq(nic);
}
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
return work_done;
}
#ifdef CONFIG_NET_POLL_CONTROLLER
static void e100_netpoll(struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
e100_disable_irq(nic);
IRQ: Maintain regs pointer globally rather than passing to IRQ handlers Maintain a per-CPU global "struct pt_regs *" variable which can be used instead of passing regs around manually through all ~1800 interrupt handlers in the Linux kernel. The regs pointer is used in few places, but it potentially costs both stack space and code to pass it around. On the FRV arch, removing the regs parameter from all the genirq function results in a 20% speed up of the IRQ exit path (ie: from leaving timer_interrupt() to leaving do_IRQ()). Where appropriate, an arch may override the generic storage facility and do something different with the variable. On FRV, for instance, the address is maintained in GR28 at all times inside the kernel as part of general exception handling. Having looked over the code, it appears that the parameter may be handed down through up to twenty or so layers of functions. Consider a USB character device attached to a USB hub, attached to a USB controller that posts its interrupts through a cascaded auxiliary interrupt controller. A character device driver may want to pass regs to the sysrq handler through the input layer which adds another few layers of parameter passing. I've build this code with allyesconfig for x86_64 and i386. I've runtested the main part of the code on FRV and i386, though I can't test most of the drivers. I've also done partial conversion for powerpc and MIPS - these at least compile with minimal configurations. This will affect all archs. Mostly the changes should be relatively easy. Take do_IRQ(), store the regs pointer at the beginning, saving the old one: struct pt_regs *old_regs = set_irq_regs(regs); And put the old one back at the end: set_irq_regs(old_regs); Don't pass regs through to generic_handle_irq() or __do_IRQ(). In timer_interrupt(), this sort of change will be necessary: - update_process_times(user_mode(regs)); - profile_tick(CPU_PROFILING, regs); + update_process_times(user_mode(get_irq_regs())); + profile_tick(CPU_PROFILING); I'd like to move update_process_times()'s use of get_irq_regs() into itself, except that i386, alone of the archs, uses something other than user_mode(). Some notes on the interrupt handling in the drivers: (*) input_dev() is now gone entirely. The regs pointer is no longer stored in the input_dev struct. (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking. It does something different depending on whether it's been supplied with a regs pointer or not. (*) Various IRQ handler function pointers have been moved to type irq_handler_t. Signed-Off-By: David Howells <dhowells@redhat.com> (cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)
2006-10-05 21:55:46 +08:00
e100_intr(nic->pdev->irq, netdev);
e100_tx_clean(nic);
e100_enable_irq(nic);
}
#endif
static int e100_set_mac_address(struct net_device *netdev, void *p)
{
struct nic *nic = netdev_priv(netdev);
struct sockaddr *addr = p;
if (!is_valid_ether_addr(addr->sa_data))
return -EADDRNOTAVAIL;
memcpy(netdev->dev_addr, addr->sa_data, netdev->addr_len);
e100_exec_cb(nic, NULL, e100_setup_iaaddr);
return 0;
}
static int e100_change_mtu(struct net_device *netdev, int new_mtu)
{
if(new_mtu < ETH_ZLEN || new_mtu > ETH_DATA_LEN)
return -EINVAL;
netdev->mtu = new_mtu;
return 0;
}
static int e100_asf(struct nic *nic)
{
/* ASF can be enabled from eeprom */
return((nic->pdev->device >= 0x1050) && (nic->pdev->device <= 0x1057) &&
(nic->eeprom[eeprom_config_asf] & eeprom_asf) &&
!(nic->eeprom[eeprom_config_asf] & eeprom_gcl) &&
((nic->eeprom[eeprom_smbus_addr] & 0xFF) != 0xFE));
}
static int e100_up(struct nic *nic)
{
int err;
if((err = e100_rx_alloc_list(nic)))
return err;
if((err = e100_alloc_cbs(nic)))
goto err_rx_clean_list;
if((err = e100_hw_init(nic)))
goto err_clean_cbs;
e100_set_multicast_list(nic->netdev);
e100_start_receiver(nic, NULL);
mod_timer(&nic->watchdog, jiffies);
if((err = request_irq(nic->pdev->irq, e100_intr, IRQF_SHARED,
nic->netdev->name, nic->netdev)))
goto err_no_irq;
netif_wake_queue(nic->netdev);
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
napi_enable(&nic->napi);
/* enable ints _after_ enabling poll, preventing a race between
* disable ints+schedule */
e100_enable_irq(nic);
return 0;
err_no_irq:
del_timer_sync(&nic->watchdog);
err_clean_cbs:
e100_clean_cbs(nic);
err_rx_clean_list:
e100_rx_clean_list(nic);
return err;
}
static void e100_down(struct nic *nic)
{
/* wait here for poll to complete */
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
napi_disable(&nic->napi);
netif_stop_queue(nic->netdev);
e100_hw_reset(nic);
free_irq(nic->pdev->irq, nic->netdev);
del_timer_sync(&nic->watchdog);
netif_carrier_off(nic->netdev);
e100_clean_cbs(nic);
e100_rx_clean_list(nic);
}
static void e100_tx_timeout(struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
/* Reset outside of interrupt context, to avoid request_irq
* in interrupt context */
schedule_work(&nic->tx_timeout_task);
}
static void e100_tx_timeout_task(struct work_struct *work)
{
struct nic *nic = container_of(work, struct nic, tx_timeout_task);
struct net_device *netdev = nic->netdev;
DPRINTK(TX_ERR, DEBUG, "scb.status=0x%02X\n",
ioread8(&nic->csr->scb.status));
e100_down(netdev_priv(netdev));
e100_up(netdev_priv(netdev));
}
static int e100_loopback_test(struct nic *nic, enum loopback loopback_mode)
{
int err;
struct sk_buff *skb;
/* Use driver resources to perform internal MAC or PHY
* loopback test. A single packet is prepared and transmitted
* in loopback mode, and the test passes if the received
* packet compares byte-for-byte to the transmitted packet. */
if((err = e100_rx_alloc_list(nic)))
return err;
if((err = e100_alloc_cbs(nic)))
goto err_clean_rx;
/* ICH PHY loopback is broken so do MAC loopback instead */
if(nic->flags & ich && loopback_mode == lb_phy)
loopback_mode = lb_mac;
nic->loopback = loopback_mode;
if((err = e100_hw_init(nic)))
goto err_loopback_none;
if(loopback_mode == lb_phy)
mdio_write(nic->netdev, nic->mii.phy_id, MII_BMCR,
BMCR_LOOPBACK);
e100_start_receiver(nic, NULL);
if(!(skb = netdev_alloc_skb(nic->netdev, ETH_DATA_LEN))) {
err = -ENOMEM;
goto err_loopback_none;
}
skb_put(skb, ETH_DATA_LEN);
memset(skb->data, 0xFF, ETH_DATA_LEN);
e100_xmit_frame(skb, nic->netdev);
msleep(10);
pci_dma_sync_single_for_cpu(nic->pdev, nic->rx_to_clean->dma_addr,
RFD_BUF_LEN, PCI_DMA_FROMDEVICE);
if(memcmp(nic->rx_to_clean->skb->data + sizeof(struct rfd),
skb->data, ETH_DATA_LEN))
err = -EAGAIN;
err_loopback_none:
mdio_write(nic->netdev, nic->mii.phy_id, MII_BMCR, 0);
nic->loopback = lb_none;
e100_clean_cbs(nic);
e100_hw_reset(nic);
err_clean_rx:
e100_rx_clean_list(nic);
return err;
}
#define MII_LED_CONTROL 0x1B
static void e100_blink_led(unsigned long data)
{
struct nic *nic = (struct nic *)data;
enum led_state {
led_on = 0x01,
led_off = 0x04,
led_on_559 = 0x05,
led_on_557 = 0x07,
};
nic->leds = (nic->leds & led_on) ? led_off :
(nic->mac < mac_82559_D101M) ? led_on_557 : led_on_559;
mdio_write(nic->netdev, nic->mii.phy_id, MII_LED_CONTROL, nic->leds);
mod_timer(&nic->blink_timer, jiffies + HZ / 4);
}
static int e100_get_settings(struct net_device *netdev, struct ethtool_cmd *cmd)
{
struct nic *nic = netdev_priv(netdev);
return mii_ethtool_gset(&nic->mii, cmd);
}
static int e100_set_settings(struct net_device *netdev, struct ethtool_cmd *cmd)
{
struct nic *nic = netdev_priv(netdev);
int err;
mdio_write(netdev, nic->mii.phy_id, MII_BMCR, BMCR_RESET);
err = mii_ethtool_sset(&nic->mii, cmd);
e100_exec_cb(nic, NULL, e100_configure);
return err;
}
static void e100_get_drvinfo(struct net_device *netdev,
struct ethtool_drvinfo *info)
{
struct nic *nic = netdev_priv(netdev);
strcpy(info->driver, DRV_NAME);
strcpy(info->version, DRV_VERSION);
strcpy(info->fw_version, "N/A");
strcpy(info->bus_info, pci_name(nic->pdev));
}
#define E100_PHY_REGS 0x1C
static int e100_get_regs_len(struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
return 1 + E100_PHY_REGS + sizeof(nic->mem->dump_buf);
}
static void e100_get_regs(struct net_device *netdev,
struct ethtool_regs *regs, void *p)
{
struct nic *nic = netdev_priv(netdev);
u32 *buff = p;
int i;
regs->version = (1 << 24) | nic->pdev->revision;
buff[0] = ioread8(&nic->csr->scb.cmd_hi) << 24 |
ioread8(&nic->csr->scb.cmd_lo) << 16 |
ioread16(&nic->csr->scb.status);
for(i = E100_PHY_REGS; i >= 0; i--)
buff[1 + E100_PHY_REGS - i] =
mdio_read(netdev, nic->mii.phy_id, i);
memset(nic->mem->dump_buf, 0, sizeof(nic->mem->dump_buf));
e100_exec_cb(nic, NULL, e100_dump);
msleep(10);
memcpy(&buff[2 + E100_PHY_REGS], nic->mem->dump_buf,
sizeof(nic->mem->dump_buf));
}
static void e100_get_wol(struct net_device *netdev, struct ethtool_wolinfo *wol)
{
struct nic *nic = netdev_priv(netdev);
wol->supported = (nic->mac >= mac_82558_D101_A4) ? WAKE_MAGIC : 0;
wol->wolopts = (nic->flags & wol_magic) ? WAKE_MAGIC : 0;
}
static int e100_set_wol(struct net_device *netdev, struct ethtool_wolinfo *wol)
{
struct nic *nic = netdev_priv(netdev);
if(wol->wolopts != WAKE_MAGIC && wol->wolopts != 0)
return -EOPNOTSUPP;
if(wol->wolopts)
nic->flags |= wol_magic;
else
nic->flags &= ~wol_magic;
e100_exec_cb(nic, NULL, e100_configure);
return 0;
}
static u32 e100_get_msglevel(struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
return nic->msg_enable;
}
static void e100_set_msglevel(struct net_device *netdev, u32 value)
{
struct nic *nic = netdev_priv(netdev);
nic->msg_enable = value;
}
static int e100_nway_reset(struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
return mii_nway_restart(&nic->mii);
}
static u32 e100_get_link(struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
return mii_link_ok(&nic->mii);
}
static int e100_get_eeprom_len(struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
return nic->eeprom_wc << 1;
}
#define E100_EEPROM_MAGIC 0x1234
static int e100_get_eeprom(struct net_device *netdev,
struct ethtool_eeprom *eeprom, u8 *bytes)
{
struct nic *nic = netdev_priv(netdev);
eeprom->magic = E100_EEPROM_MAGIC;
memcpy(bytes, &((u8 *)nic->eeprom)[eeprom->offset], eeprom->len);
return 0;
}
static int e100_set_eeprom(struct net_device *netdev,
struct ethtool_eeprom *eeprom, u8 *bytes)
{
struct nic *nic = netdev_priv(netdev);
if(eeprom->magic != E100_EEPROM_MAGIC)
return -EINVAL;
memcpy(&((u8 *)nic->eeprom)[eeprom->offset], bytes, eeprom->len);
return e100_eeprom_save(nic, eeprom->offset >> 1,
(eeprom->len >> 1) + 1);
}
static void e100_get_ringparam(struct net_device *netdev,
struct ethtool_ringparam *ring)
{
struct nic *nic = netdev_priv(netdev);
struct param_range *rfds = &nic->params.rfds;
struct param_range *cbs = &nic->params.cbs;
ring->rx_max_pending = rfds->max;
ring->tx_max_pending = cbs->max;
ring->rx_mini_max_pending = 0;
ring->rx_jumbo_max_pending = 0;
ring->rx_pending = rfds->count;
ring->tx_pending = cbs->count;
ring->rx_mini_pending = 0;
ring->rx_jumbo_pending = 0;
}
static int e100_set_ringparam(struct net_device *netdev,
struct ethtool_ringparam *ring)
{
struct nic *nic = netdev_priv(netdev);
struct param_range *rfds = &nic->params.rfds;
struct param_range *cbs = &nic->params.cbs;
if ((ring->rx_mini_pending) || (ring->rx_jumbo_pending))
return -EINVAL;
if(netif_running(netdev))
e100_down(nic);
rfds->count = max(ring->rx_pending, rfds->min);
rfds->count = min(rfds->count, rfds->max);
cbs->count = max(ring->tx_pending, cbs->min);
cbs->count = min(cbs->count, cbs->max);
DPRINTK(DRV, INFO, "Ring Param settings: rx: %d, tx %d\n",
rfds->count, cbs->count);
if(netif_running(netdev))
e100_up(nic);
return 0;
}
static const char e100_gstrings_test[][ETH_GSTRING_LEN] = {
"Link test (on/offline)",
"Eeprom test (on/offline)",
"Self test (offline)",
"Mac loopback (offline)",
"Phy loopback (offline)",
};
#define E100_TEST_LEN ARRAY_SIZE(e100_gstrings_test)
static void e100_diag_test(struct net_device *netdev,
struct ethtool_test *test, u64 *data)
{
struct ethtool_cmd cmd;
struct nic *nic = netdev_priv(netdev);
int i, err;
memset(data, 0, E100_TEST_LEN * sizeof(u64));
data[0] = !mii_link_ok(&nic->mii);
data[1] = e100_eeprom_load(nic);
if(test->flags & ETH_TEST_FL_OFFLINE) {
/* save speed, duplex & autoneg settings */
err = mii_ethtool_gset(&nic->mii, &cmd);
if(netif_running(netdev))
e100_down(nic);
data[2] = e100_self_test(nic);
data[3] = e100_loopback_test(nic, lb_mac);
data[4] = e100_loopback_test(nic, lb_phy);
/* restore speed, duplex & autoneg settings */
err = mii_ethtool_sset(&nic->mii, &cmd);
if(netif_running(netdev))
e100_up(nic);
}
for(i = 0; i < E100_TEST_LEN; i++)
test->flags |= data[i] ? ETH_TEST_FL_FAILED : 0;
msleep_interruptible(4 * 1000);
}
static int e100_phys_id(struct net_device *netdev, u32 data)
{
struct nic *nic = netdev_priv(netdev);
if(!data || data > (u32)(MAX_SCHEDULE_TIMEOUT / HZ))
data = (u32)(MAX_SCHEDULE_TIMEOUT / HZ);
mod_timer(&nic->blink_timer, jiffies);
msleep_interruptible(data * 1000);
del_timer_sync(&nic->blink_timer);
mdio_write(netdev, nic->mii.phy_id, MII_LED_CONTROL, 0);
return 0;
}
static const char e100_gstrings_stats[][ETH_GSTRING_LEN] = {
"rx_packets", "tx_packets", "rx_bytes", "tx_bytes", "rx_errors",
"tx_errors", "rx_dropped", "tx_dropped", "multicast", "collisions",
"rx_length_errors", "rx_over_errors", "rx_crc_errors",
"rx_frame_errors", "rx_fifo_errors", "rx_missed_errors",
"tx_aborted_errors", "tx_carrier_errors", "tx_fifo_errors",
"tx_heartbeat_errors", "tx_window_errors",
/* device-specific stats */
"tx_deferred", "tx_single_collisions", "tx_multi_collisions",
"tx_flow_control_pause", "rx_flow_control_pause",
"rx_flow_control_unsupported", "tx_tco_packets", "rx_tco_packets",
};
#define E100_NET_STATS_LEN 21
#define E100_STATS_LEN ARRAY_SIZE(e100_gstrings_stats)
static int e100_get_sset_count(struct net_device *netdev, int sset)
{
switch (sset) {
case ETH_SS_TEST:
return E100_TEST_LEN;
case ETH_SS_STATS:
return E100_STATS_LEN;
default:
return -EOPNOTSUPP;
}
}
static void e100_get_ethtool_stats(struct net_device *netdev,
struct ethtool_stats *stats, u64 *data)
{
struct nic *nic = netdev_priv(netdev);
int i;
for(i = 0; i < E100_NET_STATS_LEN; i++)
data[i] = ((unsigned long *)&netdev->stats)[i];
data[i++] = nic->tx_deferred;
data[i++] = nic->tx_single_collisions;
data[i++] = nic->tx_multiple_collisions;
data[i++] = nic->tx_fc_pause;
data[i++] = nic->rx_fc_pause;
data[i++] = nic->rx_fc_unsupported;
data[i++] = nic->tx_tco_frames;
data[i++] = nic->rx_tco_frames;
}
static void e100_get_strings(struct net_device *netdev, u32 stringset, u8 *data)
{
switch(stringset) {
case ETH_SS_TEST:
memcpy(data, *e100_gstrings_test, sizeof(e100_gstrings_test));
break;
case ETH_SS_STATS:
memcpy(data, *e100_gstrings_stats, sizeof(e100_gstrings_stats));
break;
}
}
static const struct ethtool_ops e100_ethtool_ops = {
.get_settings = e100_get_settings,
.set_settings = e100_set_settings,
.get_drvinfo = e100_get_drvinfo,
.get_regs_len = e100_get_regs_len,
.get_regs = e100_get_regs,
.get_wol = e100_get_wol,
.set_wol = e100_set_wol,
.get_msglevel = e100_get_msglevel,
.set_msglevel = e100_set_msglevel,
.nway_reset = e100_nway_reset,
.get_link = e100_get_link,
.get_eeprom_len = e100_get_eeprom_len,
.get_eeprom = e100_get_eeprom,
.set_eeprom = e100_set_eeprom,
.get_ringparam = e100_get_ringparam,
.set_ringparam = e100_set_ringparam,
.self_test = e100_diag_test,
.get_strings = e100_get_strings,
.phys_id = e100_phys_id,
.get_ethtool_stats = e100_get_ethtool_stats,
.get_sset_count = e100_get_sset_count,
};
static int e100_do_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
{
struct nic *nic = netdev_priv(netdev);
return generic_mii_ioctl(&nic->mii, if_mii(ifr), cmd, NULL);
}
static int e100_alloc(struct nic *nic)
{
nic->mem = pci_alloc_consistent(nic->pdev, sizeof(struct mem),
&nic->dma_addr);
return nic->mem ? 0 : -ENOMEM;
}
static void e100_free(struct nic *nic)
{
if(nic->mem) {
pci_free_consistent(nic->pdev, sizeof(struct mem),
nic->mem, nic->dma_addr);
nic->mem = NULL;
}
}
static int e100_open(struct net_device *netdev)
{
struct nic *nic = netdev_priv(netdev);
int err = 0;
netif_carrier_off(netdev);
if((err = e100_up(nic)))
DPRINTK(IFUP, ERR, "Cannot open interface, aborting.\n");
return err;
}
static int e100_close(struct net_device *netdev)
{
e100_down(netdev_priv(netdev));
return 0;
}
static int __devinit e100_probe(struct pci_dev *pdev,
const struct pci_device_id *ent)
{
struct net_device *netdev;
struct nic *nic;
int err;
DECLARE_MAC_BUF(mac);
if(!(netdev = alloc_etherdev(sizeof(struct nic)))) {
if(((1 << debug) - 1) & NETIF_MSG_PROBE)
printk(KERN_ERR PFX "Etherdev alloc failed, abort.\n");
return -ENOMEM;
}
netdev->open = e100_open;
netdev->stop = e100_close;
netdev->hard_start_xmit = e100_xmit_frame;
netdev->set_multicast_list = e100_set_multicast_list;
netdev->set_mac_address = e100_set_mac_address;
netdev->change_mtu = e100_change_mtu;
netdev->do_ioctl = e100_do_ioctl;
SET_ETHTOOL_OPS(netdev, &e100_ethtool_ops);
netdev->tx_timeout = e100_tx_timeout;
netdev->watchdog_timeo = E100_WATCHDOG_PERIOD;
#ifdef CONFIG_NET_POLL_CONTROLLER
netdev->poll_controller = e100_netpoll;
#endif
strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
nic = netdev_priv(netdev);
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
netif_napi_add(netdev, &nic->napi, e100_poll, E100_NAPI_WEIGHT);
nic->netdev = netdev;
nic->pdev = pdev;
nic->msg_enable = (1 << debug) - 1;
pci_set_drvdata(pdev, netdev);
if((err = pci_enable_device(pdev))) {
DPRINTK(PROBE, ERR, "Cannot enable PCI device, aborting.\n");
goto err_out_free_dev;
}
if(!(pci_resource_flags(pdev, 0) & IORESOURCE_MEM)) {
DPRINTK(PROBE, ERR, "Cannot find proper PCI device "
"base address, aborting.\n");
err = -ENODEV;
goto err_out_disable_pdev;
}
if((err = pci_request_regions(pdev, DRV_NAME))) {
DPRINTK(PROBE, ERR, "Cannot obtain PCI resources, aborting.\n");
goto err_out_disable_pdev;
}
if((err = pci_set_dma_mask(pdev, DMA_32BIT_MASK))) {
DPRINTK(PROBE, ERR, "No usable DMA configuration, aborting.\n");
goto err_out_free_res;
}
SET_NETDEV_DEV(netdev, &pdev->dev);
if (use_io)
DPRINTK(PROBE, INFO, "using i/o access mode\n");
nic->csr = pci_iomap(pdev, (use_io ? 1 : 0), sizeof(struct csr));
if(!nic->csr) {
DPRINTK(PROBE, ERR, "Cannot map device registers, aborting.\n");
err = -ENOMEM;
goto err_out_free_res;
}
if(ent->driver_data)
nic->flags |= ich;
else
nic->flags &= ~ich;
e100_get_defaults(nic);
/* locks must be initialized before calling hw_reset */
spin_lock_init(&nic->cb_lock);
spin_lock_init(&nic->cmd_lock);
spin_lock_init(&nic->mdio_lock);
/* Reset the device before pci_set_master() in case device is in some
* funky state and has an interrupt pending - hint: we don't have the
* interrupt handler registered yet. */
e100_hw_reset(nic);
pci_set_master(pdev);
init_timer(&nic->watchdog);
nic->watchdog.function = e100_watchdog;
nic->watchdog.data = (unsigned long)nic;
init_timer(&nic->blink_timer);
nic->blink_timer.function = e100_blink_led;
nic->blink_timer.data = (unsigned long)nic;
INIT_WORK(&nic->tx_timeout_task, e100_tx_timeout_task);
if((err = e100_alloc(nic))) {
DPRINTK(PROBE, ERR, "Cannot alloc driver memory, aborting.\n");
goto err_out_iounmap;
}
if((err = e100_eeprom_load(nic)))
goto err_out_free;
e100_phy_init(nic);
memcpy(netdev->dev_addr, nic->eeprom, ETH_ALEN);
memcpy(netdev->perm_addr, nic->eeprom, ETH_ALEN);
if (!is_valid_ether_addr(netdev->perm_addr)) {
if (!eeprom_bad_csum_allow) {
DPRINTK(PROBE, ERR, "Invalid MAC address from "
"EEPROM, aborting.\n");
err = -EAGAIN;
goto err_out_free;
} else {
DPRINTK(PROBE, ERR, "Invalid MAC address from EEPROM, "
"you MUST configure one.\n");
}
}
/* Wol magic packet can be enabled from eeprom */
if((nic->mac >= mac_82558_D101_A4) &&
(nic->eeprom[eeprom_id] & eeprom_id_wol))
nic->flags |= wol_magic;
/* ack any pending wake events, disable PME */
err = pci_enable_wake(pdev, 0, 0);
if (err)
DPRINTK(PROBE, ERR, "Error clearing wake event\n");
strcpy(netdev->name, "eth%d");
if((err = register_netdev(netdev))) {
DPRINTK(PROBE, ERR, "Cannot register net device, aborting.\n");
goto err_out_free;
}
DPRINTK(PROBE, INFO, "addr 0x%llx, irq %d, MAC addr %s\n",
(unsigned long long)pci_resource_start(pdev, use_io ? 1 : 0),
pdev->irq, print_mac(mac, netdev->dev_addr));
return 0;
err_out_free:
e100_free(nic);
err_out_iounmap:
pci_iounmap(pdev, nic->csr);
err_out_free_res:
pci_release_regions(pdev);
err_out_disable_pdev:
pci_disable_device(pdev);
err_out_free_dev:
pci_set_drvdata(pdev, NULL);
free_netdev(netdev);
return err;
}
static void __devexit e100_remove(struct pci_dev *pdev)
{
struct net_device *netdev = pci_get_drvdata(pdev);
if(netdev) {
struct nic *nic = netdev_priv(netdev);
unregister_netdev(netdev);
e100_free(nic);
pci_iounmap(pdev, nic->csr);
free_netdev(netdev);
pci_release_regions(pdev);
pci_disable_device(pdev);
pci_set_drvdata(pdev, NULL);
}
}
static int e100_suspend(struct pci_dev *pdev, pm_message_t state)
{
struct net_device *netdev = pci_get_drvdata(pdev);
struct nic *nic = netdev_priv(netdev);
if (netif_running(netdev))
e100_down(nic);
netif_device_detach(netdev);
pci_save_state(pdev);
if ((nic->flags & wol_magic) | e100_asf(nic)) {
pci_enable_wake(pdev, PCI_D3hot, 1);
pci_enable_wake(pdev, PCI_D3cold, 1);
} else {
pci_enable_wake(pdev, PCI_D3hot, 0);
pci_enable_wake(pdev, PCI_D3cold, 0);
}
pci_disable_device(pdev);
pci_set_power_state(pdev, PCI_D3hot);
return 0;
}
#ifdef CONFIG_PM
static int e100_resume(struct pci_dev *pdev)
{
struct net_device *netdev = pci_get_drvdata(pdev);
struct nic *nic = netdev_priv(netdev);
pci_set_power_state(pdev, PCI_D0);
pci_restore_state(pdev);
/* ack any pending wake events, disable PME */
pci_enable_wake(pdev, 0, 0);
netif_device_attach(netdev);
if (netif_running(netdev))
e100_up(nic);
return 0;
}
#endif /* CONFIG_PM */
static void e100_shutdown(struct pci_dev *pdev)
{
e100_suspend(pdev, PMSG_SUSPEND);
}
/* ------------------ PCI Error Recovery infrastructure -------------- */
/**
* e100_io_error_detected - called when PCI error is detected.
* @pdev: Pointer to PCI device
* @state: The current pci connection state
*/
static pci_ers_result_t e100_io_error_detected(struct pci_dev *pdev, pci_channel_state_t state)
{
struct net_device *netdev = pci_get_drvdata(pdev);
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
struct nic *nic = netdev_priv(netdev);
/* Similar to calling e100_down(), but avoids adapter I/O. */
netdev->stop(netdev);
/* Detach; put netif into a state similar to hotplug unplug. */
[NET]: Make NAPI polling independent of struct net_device objects. Several devices have multiple independant RX queues per net device, and some have a single interrupt doorbell for several queues. In either case, it's easier to support layouts like that if the structure representing the poll is independant from the net device itself. The signature of the ->poll() call back goes from: int foo_poll(struct net_device *dev, int *budget) to int foo_poll(struct napi_struct *napi, int budget) The caller is returned the number of RX packets processed (or the number of "NAPI credits" consumed if you want to get abstract). The callee no longer messes around bumping dev->quota, *budget, etc. because that is all handled in the caller upon return. The napi_struct is to be embedded in the device driver private data structures. Furthermore, it is the driver's responsibility to disable all NAPI instances in it's ->stop() device close handler. Since the napi_struct is privatized into the driver's private data structures, only the driver knows how to get at all of the napi_struct instances it may have per-device. With lots of help and suggestions from Rusty Russell, Roland Dreier, Michael Chan, Jeff Garzik, and Jamal Hadi Salim. Bug fixes from Thomas Graf, Roland Dreier, Peter Zijlstra, Joseph Fannin, Scott Wood, Hans J. Koch, and Michael Chan. [ Ported to current tree and all drivers converted. Integrated Stephen's follow-on kerneldoc additions, and restored poll_list handling to the old style to fix mutual exclusion issues. -DaveM ] Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
2007-10-04 07:41:36 +08:00
napi_enable(&nic->napi);
netif_device_detach(netdev);
pci_disable_device(pdev);
/* Request a slot reset. */
return PCI_ERS_RESULT_NEED_RESET;
}
/**
* e100_io_slot_reset - called after the pci bus has been reset.
* @pdev: Pointer to PCI device
*
* Restart the card from scratch.
*/
static pci_ers_result_t e100_io_slot_reset(struct pci_dev *pdev)
{
struct net_device *netdev = pci_get_drvdata(pdev);
struct nic *nic = netdev_priv(netdev);
if (pci_enable_device(pdev)) {
printk(KERN_ERR "e100: Cannot re-enable PCI device after reset.\n");
return PCI_ERS_RESULT_DISCONNECT;
}
pci_set_master(pdev);
/* Only one device per card can do a reset */
if (0 != PCI_FUNC(pdev->devfn))
return PCI_ERS_RESULT_RECOVERED;
e100_hw_reset(nic);
e100_phy_init(nic);
return PCI_ERS_RESULT_RECOVERED;
}
/**
* e100_io_resume - resume normal operations
* @pdev: Pointer to PCI device
*
* Resume normal operations after an error recovery
* sequence has been completed.
*/
static void e100_io_resume(struct pci_dev *pdev)
{
struct net_device *netdev = pci_get_drvdata(pdev);
struct nic *nic = netdev_priv(netdev);
/* ack any pending wake events, disable PME */
pci_enable_wake(pdev, 0, 0);
netif_device_attach(netdev);
if (netif_running(netdev)) {
e100_open(netdev);
mod_timer(&nic->watchdog, jiffies);
}
}
static struct pci_error_handlers e100_err_handler = {
.error_detected = e100_io_error_detected,
.slot_reset = e100_io_slot_reset,
.resume = e100_io_resume,
};
static struct pci_driver e100_driver = {
.name = DRV_NAME,
.id_table = e100_id_table,
.probe = e100_probe,
.remove = __devexit_p(e100_remove),
#ifdef CONFIG_PM
/* Power Management hooks */
.suspend = e100_suspend,
.resume = e100_resume,
#endif
.shutdown = e100_shutdown,
.err_handler = &e100_err_handler,
};
static int __init e100_init_module(void)
{
if(((1 << debug) - 1) & NETIF_MSG_DRV) {
printk(KERN_INFO PFX "%s, %s\n", DRV_DESCRIPTION, DRV_VERSION);
printk(KERN_INFO PFX "%s\n", DRV_COPYRIGHT);
}
return pci_register_driver(&e100_driver);
}
static void __exit e100_cleanup_module(void)
{
pci_unregister_driver(&e100_driver);
}
module_init(e100_init_module);
module_exit(e100_cleanup_module);