xref: /openbmc/qemu/hw/net/e1000.c (revision 891f8dcd)
1 /*
2  * QEMU e1000 emulation
3  *
4  * Software developer's manual:
5  * http://download.intel.com/design/network/manuals/8254x_GBe_SDM.pdf
6  *
7  * Nir Peleg, Tutis Systems Ltd. for Qumranet Inc.
8  * Copyright (c) 2008 Qumranet
9  * Based on work done by:
10  * Copyright (c) 2007 Dan Aloni
11  * Copyright (c) 2004 Antony T Curtis
12  *
13  * This library is free software; you can redistribute it and/or
14  * modify it under the terms of the GNU Lesser General Public
15  * License as published by the Free Software Foundation; either
16  * version 2 of the License, or (at your option) any later version.
17  *
18  * This library is distributed in the hope that it will be useful,
19  * but WITHOUT ANY WARRANTY; without even the implied warranty of
20  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
21  * Lesser General Public License for more details.
22  *
23  * You should have received a copy of the GNU Lesser General Public
24  * License along with this library; if not, see <http://www.gnu.org/licenses/>.
25  */
26 
27 
28 #include "qemu/osdep.h"
29 #include "hw/hw.h"
30 #include "hw/pci/pci.h"
31 #include "net/net.h"
32 #include "net/checksum.h"
33 #include "hw/loader.h"
34 #include "sysemu/sysemu.h"
35 #include "sysemu/dma.h"
36 #include "qemu/iov.h"
37 #include "qemu/range.h"
38 
39 #include "e1000x_common.h"
40 
41 static const uint8_t bcast[] = {0xff, 0xff, 0xff, 0xff, 0xff, 0xff};
42 
43 #define E1000_DEBUG
44 
45 #ifdef E1000_DEBUG
46 enum {
47     DEBUG_GENERAL,      DEBUG_IO,       DEBUG_MMIO,     DEBUG_INTERRUPT,
48     DEBUG_RX,           DEBUG_TX,       DEBUG_MDIC,     DEBUG_EEPROM,
49     DEBUG_UNKNOWN,      DEBUG_TXSUM,    DEBUG_TXERR,    DEBUG_RXERR,
50     DEBUG_RXFILTER,     DEBUG_PHY,      DEBUG_NOTYET,
51 };
52 #define DBGBIT(x)    (1<<DEBUG_##x)
53 static int debugflags = DBGBIT(TXERR) | DBGBIT(GENERAL);
54 
55 #define DBGOUT(what, fmt, ...) do { \
56     if (debugflags & DBGBIT(what)) \
57         fprintf(stderr, "e1000: " fmt, ## __VA_ARGS__); \
58     } while (0)
59 #else
60 #define DBGOUT(what, fmt, ...) do {} while (0)
61 #endif
62 
63 #define IOPORT_SIZE       0x40
64 #define PNPMMIO_SIZE      0x20000
65 #define MIN_BUF_SIZE      60 /* Min. octets in an ethernet frame sans FCS */
66 
67 #define MAXIMUM_ETHERNET_HDR_LEN (14+4)
68 
69 /*
70  * HW models:
71  *  E1000_DEV_ID_82540EM works with Windows, Linux, and OS X <= 10.8
72  *  E1000_DEV_ID_82544GC_COPPER appears to work; not well tested
73  *  E1000_DEV_ID_82545EM_COPPER works with Linux and OS X >= 10.6
74  *  Others never tested
75  */
76 
77 typedef struct E1000State_st {
78     /*< private >*/
79     PCIDevice parent_obj;
80     /*< public >*/
81 
82     NICState *nic;
83     NICConf conf;
84     MemoryRegion mmio;
85     MemoryRegion io;
86 
87     uint32_t mac_reg[0x8000];
88     uint16_t phy_reg[0x20];
89     uint16_t eeprom_data[64];
90 
91     uint32_t rxbuf_size;
92     uint32_t rxbuf_min_shift;
93     struct e1000_tx {
94         unsigned char header[256];
95         unsigned char vlan_header[4];
96         /* Fields vlan and data must not be reordered or separated. */
97         unsigned char vlan[4];
98         unsigned char data[0x10000];
99         uint16_t size;
100         unsigned char vlan_needed;
101         e1000x_txd_props props;
102         uint16_t tso_frames;
103     } tx;
104 
105     struct {
106         uint32_t val_in;    /* shifted in from guest driver */
107         uint16_t bitnum_in;
108         uint16_t bitnum_out;
109         uint16_t reading;
110         uint32_t old_eecd;
111     } eecd_state;
112 
113     QEMUTimer *autoneg_timer;
114 
115     QEMUTimer *mit_timer;      /* Mitigation timer. */
116     bool mit_timer_on;         /* Mitigation timer is running. */
117     bool mit_irq_level;        /* Tracks interrupt pin level. */
118     uint32_t mit_ide;          /* Tracks E1000_TXD_CMD_IDE bit. */
119 
120 /* Compatibility flags for migration to/from qemu 1.3.0 and older */
121 #define E1000_FLAG_AUTONEG_BIT 0
122 #define E1000_FLAG_MIT_BIT 1
123 #define E1000_FLAG_MAC_BIT 2
124 #define E1000_FLAG_AUTONEG (1 << E1000_FLAG_AUTONEG_BIT)
125 #define E1000_FLAG_MIT (1 << E1000_FLAG_MIT_BIT)
126 #define E1000_FLAG_MAC (1 << E1000_FLAG_MAC_BIT)
127     uint32_t compat_flags;
128 } E1000State;
129 
130 #define chkflag(x)     (s->compat_flags & E1000_FLAG_##x)
131 
132 typedef struct E1000BaseClass {
133     PCIDeviceClass parent_class;
134     uint16_t phy_id2;
135 } E1000BaseClass;
136 
137 #define TYPE_E1000_BASE "e1000-base"
138 
139 #define E1000(obj) \
140     OBJECT_CHECK(E1000State, (obj), TYPE_E1000_BASE)
141 
142 #define E1000_DEVICE_CLASS(klass) \
143      OBJECT_CLASS_CHECK(E1000BaseClass, (klass), TYPE_E1000_BASE)
144 #define E1000_DEVICE_GET_CLASS(obj) \
145     OBJECT_GET_CLASS(E1000BaseClass, (obj), TYPE_E1000_BASE)
146 
147 static void
148 e1000_link_up(E1000State *s)
149 {
150     e1000x_update_regs_on_link_up(s->mac_reg, s->phy_reg);
151 
152     /* E1000_STATUS_LU is tested by e1000_can_receive() */
153     qemu_flush_queued_packets(qemu_get_queue(s->nic));
154 }
155 
156 static void
157 e1000_autoneg_done(E1000State *s)
158 {
159     e1000x_update_regs_on_autoneg_done(s->mac_reg, s->phy_reg);
160 
161     /* E1000_STATUS_LU is tested by e1000_can_receive() */
162     qemu_flush_queued_packets(qemu_get_queue(s->nic));
163 }
164 
165 static bool
166 have_autoneg(E1000State *s)
167 {
168     return chkflag(AUTONEG) && (s->phy_reg[PHY_CTRL] & MII_CR_AUTO_NEG_EN);
169 }
170 
171 static void
172 set_phy_ctrl(E1000State *s, int index, uint16_t val)
173 {
174     /* bits 0-5 reserved; MII_CR_[RESTART_AUTO_NEG,RESET] are self clearing */
175     s->phy_reg[PHY_CTRL] = val & ~(0x3f |
176                                    MII_CR_RESET |
177                                    MII_CR_RESTART_AUTO_NEG);
178 
179     /*
180      * QEMU 1.3 does not support link auto-negotiation emulation, so if we
181      * migrate during auto negotiation, after migration the link will be
182      * down.
183      */
184     if (have_autoneg(s) && (val & MII_CR_RESTART_AUTO_NEG)) {
185         e1000x_restart_autoneg(s->mac_reg, s->phy_reg, s->autoneg_timer);
186     }
187 }
188 
189 static void (*phyreg_writeops[])(E1000State *, int, uint16_t) = {
190     [PHY_CTRL] = set_phy_ctrl,
191 };
192 
193 enum { NPHYWRITEOPS = ARRAY_SIZE(phyreg_writeops) };
194 
195 enum { PHY_R = 1, PHY_W = 2, PHY_RW = PHY_R | PHY_W };
196 static const char phy_regcap[0x20] = {
197     [PHY_STATUS]      = PHY_R,     [M88E1000_EXT_PHY_SPEC_CTRL] = PHY_RW,
198     [PHY_ID1]         = PHY_R,     [M88E1000_PHY_SPEC_CTRL]     = PHY_RW,
199     [PHY_CTRL]        = PHY_RW,    [PHY_1000T_CTRL]             = PHY_RW,
200     [PHY_LP_ABILITY]  = PHY_R,     [PHY_1000T_STATUS]           = PHY_R,
201     [PHY_AUTONEG_ADV] = PHY_RW,    [M88E1000_RX_ERR_CNTR]       = PHY_R,
202     [PHY_ID2]         = PHY_R,     [M88E1000_PHY_SPEC_STATUS]   = PHY_R,
203     [PHY_AUTONEG_EXP] = PHY_R,
204 };
205 
206 /* PHY_ID2 documented in 8254x_GBe_SDM.pdf, pp. 250 */
207 static const uint16_t phy_reg_init[] = {
208     [PHY_CTRL]   = MII_CR_SPEED_SELECT_MSB |
209                    MII_CR_FULL_DUPLEX |
210                    MII_CR_AUTO_NEG_EN,
211 
212     [PHY_STATUS] = MII_SR_EXTENDED_CAPS |
213                    MII_SR_LINK_STATUS |   /* link initially up */
214                    MII_SR_AUTONEG_CAPS |
215                    /* MII_SR_AUTONEG_COMPLETE: initially NOT completed */
216                    MII_SR_PREAMBLE_SUPPRESS |
217                    MII_SR_EXTENDED_STATUS |
218                    MII_SR_10T_HD_CAPS |
219                    MII_SR_10T_FD_CAPS |
220                    MII_SR_100X_HD_CAPS |
221                    MII_SR_100X_FD_CAPS,
222 
223     [PHY_ID1] = 0x141,
224     /* [PHY_ID2] configured per DevId, from e1000_reset() */
225     [PHY_AUTONEG_ADV] = 0xde1,
226     [PHY_LP_ABILITY] = 0x1e0,
227     [PHY_1000T_CTRL] = 0x0e00,
228     [PHY_1000T_STATUS] = 0x3c00,
229     [M88E1000_PHY_SPEC_CTRL] = 0x360,
230     [M88E1000_PHY_SPEC_STATUS] = 0xac00,
231     [M88E1000_EXT_PHY_SPEC_CTRL] = 0x0d60,
232 };
233 
234 static const uint32_t mac_reg_init[] = {
235     [PBA]     = 0x00100030,
236     [LEDCTL]  = 0x602,
237     [CTRL]    = E1000_CTRL_SWDPIN2 | E1000_CTRL_SWDPIN0 |
238                 E1000_CTRL_SPD_1000 | E1000_CTRL_SLU,
239     [STATUS]  = 0x80000000 | E1000_STATUS_GIO_MASTER_ENABLE |
240                 E1000_STATUS_ASDV | E1000_STATUS_MTXCKOK |
241                 E1000_STATUS_SPEED_1000 | E1000_STATUS_FD |
242                 E1000_STATUS_LU,
243     [MANC]    = E1000_MANC_EN_MNG2HOST | E1000_MANC_RCV_TCO_EN |
244                 E1000_MANC_ARP_EN | E1000_MANC_0298_EN |
245                 E1000_MANC_RMCP_EN,
246 };
247 
248 /* Helper function, *curr == 0 means the value is not set */
249 static inline void
250 mit_update_delay(uint32_t *curr, uint32_t value)
251 {
252     if (value && (*curr == 0 || value < *curr)) {
253         *curr = value;
254     }
255 }
256 
257 static void
258 set_interrupt_cause(E1000State *s, int index, uint32_t val)
259 {
260     PCIDevice *d = PCI_DEVICE(s);
261     uint32_t pending_ints;
262     uint32_t mit_delay;
263 
264     s->mac_reg[ICR] = val;
265 
266     /*
267      * Make sure ICR and ICS registers have the same value.
268      * The spec says that the ICS register is write-only.  However in practice,
269      * on real hardware ICS is readable, and for reads it has the same value as
270      * ICR (except that ICS does not have the clear on read behaviour of ICR).
271      *
272      * The VxWorks PRO/1000 driver uses this behaviour.
273      */
274     s->mac_reg[ICS] = val;
275 
276     pending_ints = (s->mac_reg[IMS] & s->mac_reg[ICR]);
277     if (!s->mit_irq_level && pending_ints) {
278         /*
279          * Here we detect a potential raising edge. We postpone raising the
280          * interrupt line if we are inside the mitigation delay window
281          * (s->mit_timer_on == 1).
282          * We provide a partial implementation of interrupt mitigation,
283          * emulating only RADV, TADV and ITR (lower 16 bits, 1024ns units for
284          * RADV and TADV, 256ns units for ITR). RDTR is only used to enable
285          * RADV; relative timers based on TIDV and RDTR are not implemented.
286          */
287         if (s->mit_timer_on) {
288             return;
289         }
290         if (chkflag(MIT)) {
291             /* Compute the next mitigation delay according to pending
292              * interrupts and the current values of RADV (provided
293              * RDTR!=0), TADV and ITR.
294              * Then rearm the timer.
295              */
296             mit_delay = 0;
297             if (s->mit_ide &&
298                     (pending_ints & (E1000_ICR_TXQE | E1000_ICR_TXDW))) {
299                 mit_update_delay(&mit_delay, s->mac_reg[TADV] * 4);
300             }
301             if (s->mac_reg[RDTR] && (pending_ints & E1000_ICS_RXT0)) {
302                 mit_update_delay(&mit_delay, s->mac_reg[RADV] * 4);
303             }
304             mit_update_delay(&mit_delay, s->mac_reg[ITR]);
305 
306             /*
307              * According to e1000 SPEC, the Ethernet controller guarantees
308              * a maximum observable interrupt rate of 7813 interrupts/sec.
309              * Thus if mit_delay < 500 then the delay should be set to the
310              * minimum delay possible which is 500.
311              */
312             mit_delay = (mit_delay < 500) ? 500 : mit_delay;
313 
314             if (mit_delay) {
315                 s->mit_timer_on = 1;
316                 timer_mod(s->mit_timer, qemu_clock_get_ns(QEMU_CLOCK_VIRTUAL) +
317                           mit_delay * 256);
318             }
319             s->mit_ide = 0;
320         }
321     }
322 
323     s->mit_irq_level = (pending_ints != 0);
324     pci_set_irq(d, s->mit_irq_level);
325 }
326 
327 static void
328 e1000_mit_timer(void *opaque)
329 {
330     E1000State *s = opaque;
331 
332     s->mit_timer_on = 0;
333     /* Call set_interrupt_cause to update the irq level (if necessary). */
334     set_interrupt_cause(s, 0, s->mac_reg[ICR]);
335 }
336 
337 static void
338 set_ics(E1000State *s, int index, uint32_t val)
339 {
340     DBGOUT(INTERRUPT, "set_ics %x, ICR %x, IMR %x\n", val, s->mac_reg[ICR],
341         s->mac_reg[IMS]);
342     set_interrupt_cause(s, 0, val | s->mac_reg[ICR]);
343 }
344 
345 static void
346 e1000_autoneg_timer(void *opaque)
347 {
348     E1000State *s = opaque;
349     if (!qemu_get_queue(s->nic)->link_down) {
350         e1000_autoneg_done(s);
351         set_ics(s, 0, E1000_ICS_LSC); /* signal link status change to guest */
352     }
353 }
354 
355 static void e1000_reset(void *opaque)
356 {
357     E1000State *d = opaque;
358     E1000BaseClass *edc = E1000_DEVICE_GET_CLASS(d);
359     uint8_t *macaddr = d->conf.macaddr.a;
360 
361     timer_del(d->autoneg_timer);
362     timer_del(d->mit_timer);
363     d->mit_timer_on = 0;
364     d->mit_irq_level = 0;
365     d->mit_ide = 0;
366     memset(d->phy_reg, 0, sizeof d->phy_reg);
367     memmove(d->phy_reg, phy_reg_init, sizeof phy_reg_init);
368     d->phy_reg[PHY_ID2] = edc->phy_id2;
369     memset(d->mac_reg, 0, sizeof d->mac_reg);
370     memmove(d->mac_reg, mac_reg_init, sizeof mac_reg_init);
371     d->rxbuf_min_shift = 1;
372     memset(&d->tx, 0, sizeof d->tx);
373 
374     if (qemu_get_queue(d->nic)->link_down) {
375         e1000x_update_regs_on_link_down(d->mac_reg, d->phy_reg);
376     }
377 
378     e1000x_reset_mac_addr(d->nic, d->mac_reg, macaddr);
379 }
380 
381 static void
382 set_ctrl(E1000State *s, int index, uint32_t val)
383 {
384     /* RST is self clearing */
385     s->mac_reg[CTRL] = val & ~E1000_CTRL_RST;
386 }
387 
388 static void
389 set_rx_control(E1000State *s, int index, uint32_t val)
390 {
391     s->mac_reg[RCTL] = val;
392     s->rxbuf_size = e1000x_rxbufsize(val);
393     s->rxbuf_min_shift = ((val / E1000_RCTL_RDMTS_QUAT) & 3) + 1;
394     DBGOUT(RX, "RCTL: %d, mac_reg[RCTL] = 0x%x\n", s->mac_reg[RDT],
395            s->mac_reg[RCTL]);
396     qemu_flush_queued_packets(qemu_get_queue(s->nic));
397 }
398 
399 static void
400 set_mdic(E1000State *s, int index, uint32_t val)
401 {
402     uint32_t data = val & E1000_MDIC_DATA_MASK;
403     uint32_t addr = ((val & E1000_MDIC_REG_MASK) >> E1000_MDIC_REG_SHIFT);
404 
405     if ((val & E1000_MDIC_PHY_MASK) >> E1000_MDIC_PHY_SHIFT != 1) // phy #
406         val = s->mac_reg[MDIC] | E1000_MDIC_ERROR;
407     else if (val & E1000_MDIC_OP_READ) {
408         DBGOUT(MDIC, "MDIC read reg 0x%x\n", addr);
409         if (!(phy_regcap[addr] & PHY_R)) {
410             DBGOUT(MDIC, "MDIC read reg %x unhandled\n", addr);
411             val |= E1000_MDIC_ERROR;
412         } else
413             val = (val ^ data) | s->phy_reg[addr];
414     } else if (val & E1000_MDIC_OP_WRITE) {
415         DBGOUT(MDIC, "MDIC write reg 0x%x, value 0x%x\n", addr, data);
416         if (!(phy_regcap[addr] & PHY_W)) {
417             DBGOUT(MDIC, "MDIC write reg %x unhandled\n", addr);
418             val |= E1000_MDIC_ERROR;
419         } else {
420             if (addr < NPHYWRITEOPS && phyreg_writeops[addr]) {
421                 phyreg_writeops[addr](s, index, data);
422             } else {
423                 s->phy_reg[addr] = data;
424             }
425         }
426     }
427     s->mac_reg[MDIC] = val | E1000_MDIC_READY;
428 
429     if (val & E1000_MDIC_INT_EN) {
430         set_ics(s, 0, E1000_ICR_MDAC);
431     }
432 }
433 
434 static uint32_t
435 get_eecd(E1000State *s, int index)
436 {
437     uint32_t ret = E1000_EECD_PRES|E1000_EECD_GNT | s->eecd_state.old_eecd;
438 
439     DBGOUT(EEPROM, "reading eeprom bit %d (reading %d)\n",
440            s->eecd_state.bitnum_out, s->eecd_state.reading);
441     if (!s->eecd_state.reading ||
442         ((s->eeprom_data[(s->eecd_state.bitnum_out >> 4) & 0x3f] >>
443           ((s->eecd_state.bitnum_out & 0xf) ^ 0xf))) & 1)
444         ret |= E1000_EECD_DO;
445     return ret;
446 }
447 
448 static void
449 set_eecd(E1000State *s, int index, uint32_t val)
450 {
451     uint32_t oldval = s->eecd_state.old_eecd;
452 
453     s->eecd_state.old_eecd = val & (E1000_EECD_SK | E1000_EECD_CS |
454             E1000_EECD_DI|E1000_EECD_FWE_MASK|E1000_EECD_REQ);
455     if (!(E1000_EECD_CS & val)) {            /* CS inactive; nothing to do */
456         return;
457     }
458     if (E1000_EECD_CS & (val ^ oldval)) {    /* CS rise edge; reset state */
459         s->eecd_state.val_in = 0;
460         s->eecd_state.bitnum_in = 0;
461         s->eecd_state.bitnum_out = 0;
462         s->eecd_state.reading = 0;
463     }
464     if (!(E1000_EECD_SK & (val ^ oldval))) {    /* no clock edge */
465         return;
466     }
467     if (!(E1000_EECD_SK & val)) {               /* falling edge */
468         s->eecd_state.bitnum_out++;
469         return;
470     }
471     s->eecd_state.val_in <<= 1;
472     if (val & E1000_EECD_DI)
473         s->eecd_state.val_in |= 1;
474     if (++s->eecd_state.bitnum_in == 9 && !s->eecd_state.reading) {
475         s->eecd_state.bitnum_out = ((s->eecd_state.val_in & 0x3f)<<4)-1;
476         s->eecd_state.reading = (((s->eecd_state.val_in >> 6) & 7) ==
477             EEPROM_READ_OPCODE_MICROWIRE);
478     }
479     DBGOUT(EEPROM, "eeprom bitnum in %d out %d, reading %d\n",
480            s->eecd_state.bitnum_in, s->eecd_state.bitnum_out,
481            s->eecd_state.reading);
482 }
483 
484 static uint32_t
485 flash_eerd_read(E1000State *s, int x)
486 {
487     unsigned int index, r = s->mac_reg[EERD] & ~E1000_EEPROM_RW_REG_START;
488 
489     if ((s->mac_reg[EERD] & E1000_EEPROM_RW_REG_START) == 0)
490         return (s->mac_reg[EERD]);
491 
492     if ((index = r >> E1000_EEPROM_RW_ADDR_SHIFT) > EEPROM_CHECKSUM_REG)
493         return (E1000_EEPROM_RW_REG_DONE | r);
494 
495     return ((s->eeprom_data[index] << E1000_EEPROM_RW_REG_DATA) |
496            E1000_EEPROM_RW_REG_DONE | r);
497 }
498 
499 static void
500 putsum(uint8_t *data, uint32_t n, uint32_t sloc, uint32_t css, uint32_t cse)
501 {
502     uint32_t sum;
503 
504     if (cse && cse < n)
505         n = cse + 1;
506     if (sloc < n-1) {
507         sum = net_checksum_add(n-css, data+css);
508         stw_be_p(data + sloc, net_checksum_finish(sum));
509     }
510 }
511 
512 static inline void
513 inc_tx_bcast_or_mcast_count(E1000State *s, const unsigned char *arr)
514 {
515     if (!memcmp(arr, bcast, sizeof bcast)) {
516         e1000x_inc_reg_if_not_full(s->mac_reg, BPTC);
517     } else if (arr[0] & 1) {
518         e1000x_inc_reg_if_not_full(s->mac_reg, MPTC);
519     }
520 }
521 
522 static void
523 e1000_send_packet(E1000State *s, const uint8_t *buf, int size)
524 {
525     static const int PTCregs[6] = { PTC64, PTC127, PTC255, PTC511,
526                                     PTC1023, PTC1522 };
527 
528     NetClientState *nc = qemu_get_queue(s->nic);
529     if (s->phy_reg[PHY_CTRL] & MII_CR_LOOPBACK) {
530         nc->info->receive(nc, buf, size);
531     } else {
532         qemu_send_packet(nc, buf, size);
533     }
534     inc_tx_bcast_or_mcast_count(s, buf);
535     e1000x_increase_size_stats(s->mac_reg, PTCregs, size);
536 }
537 
538 static void
539 xmit_seg(E1000State *s)
540 {
541     uint16_t len, *sp;
542     unsigned int frames = s->tx.tso_frames, css, sofar;
543     struct e1000_tx *tp = &s->tx;
544 
545     if (tp->props.tse && tp->props.cptse) {
546         css = tp->props.ipcss;
547         DBGOUT(TXSUM, "frames %d size %d ipcss %d\n",
548                frames, tp->size, css);
549         if (tp->props.ip) {    /* IPv4 */
550             stw_be_p(tp->data+css+2, tp->size - css);
551             stw_be_p(tp->data+css+4,
552                      be16_to_cpup((uint16_t *)(tp->data+css+4))+frames);
553         } else {         /* IPv6 */
554             stw_be_p(tp->data+css+4, tp->size - css);
555         }
556         css = tp->props.tucss;
557         len = tp->size - css;
558         DBGOUT(TXSUM, "tcp %d tucss %d len %d\n", tp->props.tcp, css, len);
559         if (tp->props.tcp) {
560             sofar = frames * tp->props.mss;
561             stl_be_p(tp->data+css+4, ldl_be_p(tp->data+css+4)+sofar); /* seq */
562             if (tp->props.paylen - sofar > tp->props.mss) {
563                 tp->data[css + 13] &= ~9;    /* PSH, FIN */
564             } else if (frames) {
565                 e1000x_inc_reg_if_not_full(s->mac_reg, TSCTC);
566             }
567         } else    /* UDP */
568             stw_be_p(tp->data+css+4, len);
569         if (tp->props.sum_needed & E1000_TXD_POPTS_TXSM) {
570             unsigned int phsum;
571             // add pseudo-header length before checksum calculation
572             sp = (uint16_t *)(tp->data + tp->props.tucso);
573             phsum = be16_to_cpup(sp) + len;
574             phsum = (phsum >> 16) + (phsum & 0xffff);
575             stw_be_p(sp, phsum);
576         }
577         tp->tso_frames++;
578     }
579 
580     if (tp->props.sum_needed & E1000_TXD_POPTS_TXSM) {
581         putsum(tp->data, tp->size, tp->props.tucso,
582                tp->props.tucss, tp->props.tucse);
583     }
584     if (tp->props.sum_needed & E1000_TXD_POPTS_IXSM) {
585         putsum(tp->data, tp->size, tp->props.ipcso,
586                tp->props.ipcss, tp->props.ipcse);
587     }
588     if (tp->vlan_needed) {
589         memmove(tp->vlan, tp->data, 4);
590         memmove(tp->data, tp->data + 4, 8);
591         memcpy(tp->data + 8, tp->vlan_header, 4);
592         e1000_send_packet(s, tp->vlan, tp->size + 4);
593     } else {
594         e1000_send_packet(s, tp->data, tp->size);
595     }
596 
597     e1000x_inc_reg_if_not_full(s->mac_reg, TPT);
598     e1000x_grow_8reg_if_not_full(s->mac_reg, TOTL, s->tx.size);
599     s->mac_reg[GPTC] = s->mac_reg[TPT];
600     s->mac_reg[GOTCL] = s->mac_reg[TOTL];
601     s->mac_reg[GOTCH] = s->mac_reg[TOTH];
602 }
603 
604 static void
605 process_tx_desc(E1000State *s, struct e1000_tx_desc *dp)
606 {
607     PCIDevice *d = PCI_DEVICE(s);
608     uint32_t txd_lower = le32_to_cpu(dp->lower.data);
609     uint32_t dtype = txd_lower & (E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D);
610     unsigned int split_size = txd_lower & 0xffff, bytes, sz;
611     unsigned int msh = 0xfffff;
612     uint64_t addr;
613     struct e1000_context_desc *xp = (struct e1000_context_desc *)dp;
614     struct e1000_tx *tp = &s->tx;
615 
616     s->mit_ide |= (txd_lower & E1000_TXD_CMD_IDE);
617     if (dtype == E1000_TXD_CMD_DEXT) {    /* context descriptor */
618         e1000x_read_tx_ctx_descr(xp, &tp->props);
619         tp->tso_frames = 0;
620         if (tp->props.tucso == 0) {    /* this is probably wrong */
621             DBGOUT(TXSUM, "TCP/UDP: cso 0!\n");
622             tp->props.tucso = tp->props.tucss + (tp->props.tcp ? 16 : 6);
623         }
624         return;
625     } else if (dtype == (E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D)) {
626         // data descriptor
627         if (tp->size == 0) {
628             tp->props.sum_needed = le32_to_cpu(dp->upper.data) >> 8;
629         }
630         tp->props.cptse = (txd_lower & E1000_TXD_CMD_TSE) ? 1 : 0;
631     } else {
632         // legacy descriptor
633         tp->props.cptse = 0;
634     }
635 
636     if (e1000x_vlan_enabled(s->mac_reg) &&
637         e1000x_is_vlan_txd(txd_lower) &&
638         (tp->props.cptse || txd_lower & E1000_TXD_CMD_EOP)) {
639         tp->vlan_needed = 1;
640         stw_be_p(tp->vlan_header,
641                       le16_to_cpu(s->mac_reg[VET]));
642         stw_be_p(tp->vlan_header + 2,
643                       le16_to_cpu(dp->upper.fields.special));
644     }
645 
646     addr = le64_to_cpu(dp->buffer_addr);
647     if (tp->props.tse && tp->props.cptse) {
648         msh = tp->props.hdr_len + tp->props.mss;
649         do {
650             bytes = split_size;
651             if (tp->size + bytes > msh)
652                 bytes = msh - tp->size;
653 
654             bytes = MIN(sizeof(tp->data) - tp->size, bytes);
655             pci_dma_read(d, addr, tp->data + tp->size, bytes);
656             sz = tp->size + bytes;
657             if (sz >= tp->props.hdr_len && tp->size < tp->props.hdr_len) {
658                 memmove(tp->header, tp->data, tp->props.hdr_len);
659             }
660             tp->size = sz;
661             addr += bytes;
662             if (sz == msh) {
663                 xmit_seg(s);
664                 memmove(tp->data, tp->header, tp->props.hdr_len);
665                 tp->size = tp->props.hdr_len;
666             }
667             split_size -= bytes;
668         } while (bytes && split_size);
669     } else if (!tp->props.tse && tp->props.cptse) {
670         // context descriptor TSE is not set, while data descriptor TSE is set
671         DBGOUT(TXERR, "TCP segmentation error\n");
672     } else {
673         split_size = MIN(sizeof(tp->data) - tp->size, split_size);
674         pci_dma_read(d, addr, tp->data + tp->size, split_size);
675         tp->size += split_size;
676     }
677 
678     if (!(txd_lower & E1000_TXD_CMD_EOP))
679         return;
680     if (!(tp->props.tse && tp->props.cptse && tp->size < tp->props.hdr_len)) {
681         xmit_seg(s);
682     }
683     tp->tso_frames = 0;
684     tp->props.sum_needed = 0;
685     tp->vlan_needed = 0;
686     tp->size = 0;
687     tp->props.cptse = 0;
688 }
689 
690 static uint32_t
691 txdesc_writeback(E1000State *s, dma_addr_t base, struct e1000_tx_desc *dp)
692 {
693     PCIDevice *d = PCI_DEVICE(s);
694     uint32_t txd_upper, txd_lower = le32_to_cpu(dp->lower.data);
695 
696     if (!(txd_lower & (E1000_TXD_CMD_RS|E1000_TXD_CMD_RPS)))
697         return 0;
698     txd_upper = (le32_to_cpu(dp->upper.data) | E1000_TXD_STAT_DD) &
699                 ~(E1000_TXD_STAT_EC | E1000_TXD_STAT_LC | E1000_TXD_STAT_TU);
700     dp->upper.data = cpu_to_le32(txd_upper);
701     pci_dma_write(d, base + ((char *)&dp->upper - (char *)dp),
702                   &dp->upper, sizeof(dp->upper));
703     return E1000_ICR_TXDW;
704 }
705 
706 static uint64_t tx_desc_base(E1000State *s)
707 {
708     uint64_t bah = s->mac_reg[TDBAH];
709     uint64_t bal = s->mac_reg[TDBAL] & ~0xf;
710 
711     return (bah << 32) + bal;
712 }
713 
714 static void
715 start_xmit(E1000State *s)
716 {
717     PCIDevice *d = PCI_DEVICE(s);
718     dma_addr_t base;
719     struct e1000_tx_desc desc;
720     uint32_t tdh_start = s->mac_reg[TDH], cause = E1000_ICS_TXQE;
721 
722     if (!(s->mac_reg[TCTL] & E1000_TCTL_EN)) {
723         DBGOUT(TX, "tx disabled\n");
724         return;
725     }
726 
727     while (s->mac_reg[TDH] != s->mac_reg[TDT]) {
728         base = tx_desc_base(s) +
729                sizeof(struct e1000_tx_desc) * s->mac_reg[TDH];
730         pci_dma_read(d, base, &desc, sizeof(desc));
731 
732         DBGOUT(TX, "index %d: %p : %x %x\n", s->mac_reg[TDH],
733                (void *)(intptr_t)desc.buffer_addr, desc.lower.data,
734                desc.upper.data);
735 
736         process_tx_desc(s, &desc);
737         cause |= txdesc_writeback(s, base, &desc);
738 
739         if (++s->mac_reg[TDH] * sizeof(desc) >= s->mac_reg[TDLEN])
740             s->mac_reg[TDH] = 0;
741         /*
742          * the following could happen only if guest sw assigns
743          * bogus values to TDT/TDLEN.
744          * there's nothing too intelligent we could do about this.
745          */
746         if (s->mac_reg[TDH] == tdh_start ||
747             tdh_start >= s->mac_reg[TDLEN] / sizeof(desc)) {
748             DBGOUT(TXERR, "TDH wraparound @%x, TDT %x, TDLEN %x\n",
749                    tdh_start, s->mac_reg[TDT], s->mac_reg[TDLEN]);
750             break;
751         }
752     }
753     set_ics(s, 0, cause);
754 }
755 
756 static int
757 receive_filter(E1000State *s, const uint8_t *buf, int size)
758 {
759     uint32_t rctl = s->mac_reg[RCTL];
760     int isbcast = !memcmp(buf, bcast, sizeof bcast), ismcast = (buf[0] & 1);
761 
762     if (e1000x_is_vlan_packet(buf, le16_to_cpu(s->mac_reg[VET])) &&
763         e1000x_vlan_rx_filter_enabled(s->mac_reg)) {
764         uint16_t vid = be16_to_cpup((uint16_t *)(buf + 14));
765         uint32_t vfta = le32_to_cpup((uint32_t *)(s->mac_reg + VFTA) +
766                                      ((vid >> 5) & 0x7f));
767         if ((vfta & (1 << (vid & 0x1f))) == 0)
768             return 0;
769     }
770 
771     if (!isbcast && !ismcast && (rctl & E1000_RCTL_UPE)) { /* promiscuous ucast */
772         return 1;
773     }
774 
775     if (ismcast && (rctl & E1000_RCTL_MPE)) {          /* promiscuous mcast */
776         e1000x_inc_reg_if_not_full(s->mac_reg, MPRC);
777         return 1;
778     }
779 
780     if (isbcast && (rctl & E1000_RCTL_BAM)) {          /* broadcast enabled */
781         e1000x_inc_reg_if_not_full(s->mac_reg, BPRC);
782         return 1;
783     }
784 
785     return e1000x_rx_group_filter(s->mac_reg, buf);
786 }
787 
788 static void
789 e1000_set_link_status(NetClientState *nc)
790 {
791     E1000State *s = qemu_get_nic_opaque(nc);
792     uint32_t old_status = s->mac_reg[STATUS];
793 
794     if (nc->link_down) {
795         e1000x_update_regs_on_link_down(s->mac_reg, s->phy_reg);
796     } else {
797         if (have_autoneg(s) &&
798             !(s->phy_reg[PHY_STATUS] & MII_SR_AUTONEG_COMPLETE)) {
799             e1000x_restart_autoneg(s->mac_reg, s->phy_reg, s->autoneg_timer);
800         } else {
801             e1000_link_up(s);
802         }
803     }
804 
805     if (s->mac_reg[STATUS] != old_status)
806         set_ics(s, 0, E1000_ICR_LSC);
807 }
808 
809 static bool e1000_has_rxbufs(E1000State *s, size_t total_size)
810 {
811     int bufs;
812     /* Fast-path short packets */
813     if (total_size <= s->rxbuf_size) {
814         return s->mac_reg[RDH] != s->mac_reg[RDT];
815     }
816     if (s->mac_reg[RDH] < s->mac_reg[RDT]) {
817         bufs = s->mac_reg[RDT] - s->mac_reg[RDH];
818     } else if (s->mac_reg[RDH] > s->mac_reg[RDT]) {
819         bufs = s->mac_reg[RDLEN] /  sizeof(struct e1000_rx_desc) +
820             s->mac_reg[RDT] - s->mac_reg[RDH];
821     } else {
822         return false;
823     }
824     return total_size <= bufs * s->rxbuf_size;
825 }
826 
827 static int
828 e1000_can_receive(NetClientState *nc)
829 {
830     E1000State *s = qemu_get_nic_opaque(nc);
831 
832     return e1000x_rx_ready(&s->parent_obj, s->mac_reg) &&
833         e1000_has_rxbufs(s, 1);
834 }
835 
836 static uint64_t rx_desc_base(E1000State *s)
837 {
838     uint64_t bah = s->mac_reg[RDBAH];
839     uint64_t bal = s->mac_reg[RDBAL] & ~0xf;
840 
841     return (bah << 32) + bal;
842 }
843 
844 static ssize_t
845 e1000_receive_iov(NetClientState *nc, const struct iovec *iov, int iovcnt)
846 {
847     E1000State *s = qemu_get_nic_opaque(nc);
848     PCIDevice *d = PCI_DEVICE(s);
849     struct e1000_rx_desc desc;
850     dma_addr_t base;
851     unsigned int n, rdt;
852     uint32_t rdh_start;
853     uint16_t vlan_special = 0;
854     uint8_t vlan_status = 0;
855     uint8_t min_buf[MIN_BUF_SIZE];
856     struct iovec min_iov;
857     uint8_t *filter_buf = iov->iov_base;
858     size_t size = iov_size(iov, iovcnt);
859     size_t iov_ofs = 0;
860     size_t desc_offset;
861     size_t desc_size;
862     size_t total_size;
863 
864     if (!e1000x_hw_rx_enabled(s->mac_reg)) {
865         return -1;
866     }
867 
868     /* Pad to minimum Ethernet frame length */
869     if (size < sizeof(min_buf)) {
870         iov_to_buf(iov, iovcnt, 0, min_buf, size);
871         memset(&min_buf[size], 0, sizeof(min_buf) - size);
872         e1000x_inc_reg_if_not_full(s->mac_reg, RUC);
873         min_iov.iov_base = filter_buf = min_buf;
874         min_iov.iov_len = size = sizeof(min_buf);
875         iovcnt = 1;
876         iov = &min_iov;
877     } else if (iov->iov_len < MAXIMUM_ETHERNET_HDR_LEN) {
878         /* This is very unlikely, but may happen. */
879         iov_to_buf(iov, iovcnt, 0, min_buf, MAXIMUM_ETHERNET_HDR_LEN);
880         filter_buf = min_buf;
881     }
882 
883     /* Discard oversized packets if !LPE and !SBP. */
884     if (e1000x_is_oversized(s->mac_reg, size)) {
885         return size;
886     }
887 
888     if (!receive_filter(s, filter_buf, size)) {
889         return size;
890     }
891 
892     if (e1000x_vlan_enabled(s->mac_reg) &&
893         e1000x_is_vlan_packet(filter_buf, le16_to_cpu(s->mac_reg[VET]))) {
894         vlan_special = cpu_to_le16(be16_to_cpup((uint16_t *)(filter_buf
895                                                                 + 14)));
896         iov_ofs = 4;
897         if (filter_buf == iov->iov_base) {
898             memmove(filter_buf + 4, filter_buf, 12);
899         } else {
900             iov_from_buf(iov, iovcnt, 4, filter_buf, 12);
901             while (iov->iov_len <= iov_ofs) {
902                 iov_ofs -= iov->iov_len;
903                 iov++;
904             }
905         }
906         vlan_status = E1000_RXD_STAT_VP;
907         size -= 4;
908     }
909 
910     rdh_start = s->mac_reg[RDH];
911     desc_offset = 0;
912     total_size = size + e1000x_fcs_len(s->mac_reg);
913     if (!e1000_has_rxbufs(s, total_size)) {
914             set_ics(s, 0, E1000_ICS_RXO);
915             return -1;
916     }
917     do {
918         desc_size = total_size - desc_offset;
919         if (desc_size > s->rxbuf_size) {
920             desc_size = s->rxbuf_size;
921         }
922         base = rx_desc_base(s) + sizeof(desc) * s->mac_reg[RDH];
923         pci_dma_read(d, base, &desc, sizeof(desc));
924         desc.special = vlan_special;
925         desc.status |= (vlan_status | E1000_RXD_STAT_DD);
926         if (desc.buffer_addr) {
927             if (desc_offset < size) {
928                 size_t iov_copy;
929                 hwaddr ba = le64_to_cpu(desc.buffer_addr);
930                 size_t copy_size = size - desc_offset;
931                 if (copy_size > s->rxbuf_size) {
932                     copy_size = s->rxbuf_size;
933                 }
934                 do {
935                     iov_copy = MIN(copy_size, iov->iov_len - iov_ofs);
936                     pci_dma_write(d, ba, iov->iov_base + iov_ofs, iov_copy);
937                     copy_size -= iov_copy;
938                     ba += iov_copy;
939                     iov_ofs += iov_copy;
940                     if (iov_ofs == iov->iov_len) {
941                         iov++;
942                         iov_ofs = 0;
943                     }
944                 } while (copy_size);
945             }
946             desc_offset += desc_size;
947             desc.length = cpu_to_le16(desc_size);
948             if (desc_offset >= total_size) {
949                 desc.status |= E1000_RXD_STAT_EOP | E1000_RXD_STAT_IXSM;
950             } else {
951                 /* Guest zeroing out status is not a hardware requirement.
952                    Clear EOP in case guest didn't do it. */
953                 desc.status &= ~E1000_RXD_STAT_EOP;
954             }
955         } else { // as per intel docs; skip descriptors with null buf addr
956             DBGOUT(RX, "Null RX descriptor!!\n");
957         }
958         pci_dma_write(d, base, &desc, sizeof(desc));
959 
960         if (++s->mac_reg[RDH] * sizeof(desc) >= s->mac_reg[RDLEN])
961             s->mac_reg[RDH] = 0;
962         /* see comment in start_xmit; same here */
963         if (s->mac_reg[RDH] == rdh_start ||
964             rdh_start >= s->mac_reg[RDLEN] / sizeof(desc)) {
965             DBGOUT(RXERR, "RDH wraparound @%x, RDT %x, RDLEN %x\n",
966                    rdh_start, s->mac_reg[RDT], s->mac_reg[RDLEN]);
967             set_ics(s, 0, E1000_ICS_RXO);
968             return -1;
969         }
970     } while (desc_offset < total_size);
971 
972     e1000x_update_rx_total_stats(s->mac_reg, size, total_size);
973 
974     n = E1000_ICS_RXT0;
975     if ((rdt = s->mac_reg[RDT]) < s->mac_reg[RDH])
976         rdt += s->mac_reg[RDLEN] / sizeof(desc);
977     if (((rdt - s->mac_reg[RDH]) * sizeof(desc)) <= s->mac_reg[RDLEN] >>
978         s->rxbuf_min_shift)
979         n |= E1000_ICS_RXDMT0;
980 
981     set_ics(s, 0, n);
982 
983     return size;
984 }
985 
986 static ssize_t
987 e1000_receive(NetClientState *nc, const uint8_t *buf, size_t size)
988 {
989     const struct iovec iov = {
990         .iov_base = (uint8_t *)buf,
991         .iov_len = size
992     };
993 
994     return e1000_receive_iov(nc, &iov, 1);
995 }
996 
997 static uint32_t
998 mac_readreg(E1000State *s, int index)
999 {
1000     return s->mac_reg[index];
1001 }
1002 
1003 static uint32_t
1004 mac_low4_read(E1000State *s, int index)
1005 {
1006     return s->mac_reg[index] & 0xf;
1007 }
1008 
1009 static uint32_t
1010 mac_low11_read(E1000State *s, int index)
1011 {
1012     return s->mac_reg[index] & 0x7ff;
1013 }
1014 
1015 static uint32_t
1016 mac_low13_read(E1000State *s, int index)
1017 {
1018     return s->mac_reg[index] & 0x1fff;
1019 }
1020 
1021 static uint32_t
1022 mac_low16_read(E1000State *s, int index)
1023 {
1024     return s->mac_reg[index] & 0xffff;
1025 }
1026 
1027 static uint32_t
1028 mac_icr_read(E1000State *s, int index)
1029 {
1030     uint32_t ret = s->mac_reg[ICR];
1031 
1032     DBGOUT(INTERRUPT, "ICR read: %x\n", ret);
1033     set_interrupt_cause(s, 0, 0);
1034     return ret;
1035 }
1036 
1037 static uint32_t
1038 mac_read_clr4(E1000State *s, int index)
1039 {
1040     uint32_t ret = s->mac_reg[index];
1041 
1042     s->mac_reg[index] = 0;
1043     return ret;
1044 }
1045 
1046 static uint32_t
1047 mac_read_clr8(E1000State *s, int index)
1048 {
1049     uint32_t ret = s->mac_reg[index];
1050 
1051     s->mac_reg[index] = 0;
1052     s->mac_reg[index-1] = 0;
1053     return ret;
1054 }
1055 
1056 static void
1057 mac_writereg(E1000State *s, int index, uint32_t val)
1058 {
1059     uint32_t macaddr[2];
1060 
1061     s->mac_reg[index] = val;
1062 
1063     if (index == RA + 1) {
1064         macaddr[0] = cpu_to_le32(s->mac_reg[RA]);
1065         macaddr[1] = cpu_to_le32(s->mac_reg[RA + 1]);
1066         qemu_format_nic_info_str(qemu_get_queue(s->nic), (uint8_t *)macaddr);
1067     }
1068 }
1069 
1070 static void
1071 set_rdt(E1000State *s, int index, uint32_t val)
1072 {
1073     s->mac_reg[index] = val & 0xffff;
1074     if (e1000_has_rxbufs(s, 1)) {
1075         qemu_flush_queued_packets(qemu_get_queue(s->nic));
1076     }
1077 }
1078 
1079 static void
1080 set_16bit(E1000State *s, int index, uint32_t val)
1081 {
1082     s->mac_reg[index] = val & 0xffff;
1083 }
1084 
1085 static void
1086 set_dlen(E1000State *s, int index, uint32_t val)
1087 {
1088     s->mac_reg[index] = val & 0xfff80;
1089 }
1090 
1091 static void
1092 set_tctl(E1000State *s, int index, uint32_t val)
1093 {
1094     s->mac_reg[index] = val;
1095     s->mac_reg[TDT] &= 0xffff;
1096     start_xmit(s);
1097 }
1098 
1099 static void
1100 set_icr(E1000State *s, int index, uint32_t val)
1101 {
1102     DBGOUT(INTERRUPT, "set_icr %x\n", val);
1103     set_interrupt_cause(s, 0, s->mac_reg[ICR] & ~val);
1104 }
1105 
1106 static void
1107 set_imc(E1000State *s, int index, uint32_t val)
1108 {
1109     s->mac_reg[IMS] &= ~val;
1110     set_ics(s, 0, 0);
1111 }
1112 
1113 static void
1114 set_ims(E1000State *s, int index, uint32_t val)
1115 {
1116     s->mac_reg[IMS] |= val;
1117     set_ics(s, 0, 0);
1118 }
1119 
1120 #define getreg(x)    [x] = mac_readreg
1121 static uint32_t (*macreg_readops[])(E1000State *, int) = {
1122     getreg(PBA),      getreg(RCTL),     getreg(TDH),      getreg(TXDCTL),
1123     getreg(WUFC),     getreg(TDT),      getreg(CTRL),     getreg(LEDCTL),
1124     getreg(MANC),     getreg(MDIC),     getreg(SWSM),     getreg(STATUS),
1125     getreg(TORL),     getreg(TOTL),     getreg(IMS),      getreg(TCTL),
1126     getreg(RDH),      getreg(RDT),      getreg(VET),      getreg(ICS),
1127     getreg(TDBAL),    getreg(TDBAH),    getreg(RDBAH),    getreg(RDBAL),
1128     getreg(TDLEN),    getreg(RDLEN),    getreg(RDTR),     getreg(RADV),
1129     getreg(TADV),     getreg(ITR),      getreg(FCRUC),    getreg(IPAV),
1130     getreg(WUC),      getreg(WUS),      getreg(SCC),      getreg(ECOL),
1131     getreg(MCC),      getreg(LATECOL),  getreg(COLC),     getreg(DC),
1132     getreg(TNCRS),    getreg(SEC),      getreg(CEXTERR),  getreg(RLEC),
1133     getreg(XONRXC),   getreg(XONTXC),   getreg(XOFFRXC),  getreg(XOFFTXC),
1134     getreg(RFC),      getreg(RJC),      getreg(RNBC),     getreg(TSCTFC),
1135     getreg(MGTPRC),   getreg(MGTPDC),   getreg(MGTPTC),   getreg(GORCL),
1136     getreg(GOTCL),
1137 
1138     [TOTH]    = mac_read_clr8,      [TORH]    = mac_read_clr8,
1139     [GOTCH]   = mac_read_clr8,      [GORCH]   = mac_read_clr8,
1140     [PRC64]   = mac_read_clr4,      [PRC127]  = mac_read_clr4,
1141     [PRC255]  = mac_read_clr4,      [PRC511]  = mac_read_clr4,
1142     [PRC1023] = mac_read_clr4,      [PRC1522] = mac_read_clr4,
1143     [PTC64]   = mac_read_clr4,      [PTC127]  = mac_read_clr4,
1144     [PTC255]  = mac_read_clr4,      [PTC511]  = mac_read_clr4,
1145     [PTC1023] = mac_read_clr4,      [PTC1522] = mac_read_clr4,
1146     [GPRC]    = mac_read_clr4,      [GPTC]    = mac_read_clr4,
1147     [TPT]     = mac_read_clr4,      [TPR]     = mac_read_clr4,
1148     [RUC]     = mac_read_clr4,      [ROC]     = mac_read_clr4,
1149     [BPRC]    = mac_read_clr4,      [MPRC]    = mac_read_clr4,
1150     [TSCTC]   = mac_read_clr4,      [BPTC]    = mac_read_clr4,
1151     [MPTC]    = mac_read_clr4,
1152     [ICR]     = mac_icr_read,       [EECD]    = get_eecd,
1153     [EERD]    = flash_eerd_read,
1154     [RDFH]    = mac_low13_read,     [RDFT]    = mac_low13_read,
1155     [RDFHS]   = mac_low13_read,     [RDFTS]   = mac_low13_read,
1156     [RDFPC]   = mac_low13_read,
1157     [TDFH]    = mac_low11_read,     [TDFT]    = mac_low11_read,
1158     [TDFHS]   = mac_low13_read,     [TDFTS]   = mac_low13_read,
1159     [TDFPC]   = mac_low13_read,
1160     [AIT]     = mac_low16_read,
1161 
1162     [CRCERRS ... MPC]   = &mac_readreg,
1163     [IP6AT ... IP6AT+3] = &mac_readreg,    [IP4AT ... IP4AT+6] = &mac_readreg,
1164     [FFLT ... FFLT+6]   = &mac_low11_read,
1165     [RA ... RA+31]      = &mac_readreg,
1166     [WUPM ... WUPM+31]  = &mac_readreg,
1167     [MTA ... MTA+127]   = &mac_readreg,
1168     [VFTA ... VFTA+127] = &mac_readreg,
1169     [FFMT ... FFMT+254] = &mac_low4_read,
1170     [FFVT ... FFVT+254] = &mac_readreg,
1171     [PBM ... PBM+16383] = &mac_readreg,
1172 };
1173 enum { NREADOPS = ARRAY_SIZE(macreg_readops) };
1174 
1175 #define putreg(x)    [x] = mac_writereg
1176 static void (*macreg_writeops[])(E1000State *, int, uint32_t) = {
1177     putreg(PBA),      putreg(EERD),     putreg(SWSM),     putreg(WUFC),
1178     putreg(TDBAL),    putreg(TDBAH),    putreg(TXDCTL),   putreg(RDBAH),
1179     putreg(RDBAL),    putreg(LEDCTL),   putreg(VET),      putreg(FCRUC),
1180     putreg(TDFH),     putreg(TDFT),     putreg(TDFHS),    putreg(TDFTS),
1181     putreg(TDFPC),    putreg(RDFH),     putreg(RDFT),     putreg(RDFHS),
1182     putreg(RDFTS),    putreg(RDFPC),    putreg(IPAV),     putreg(WUC),
1183     putreg(WUS),      putreg(AIT),
1184 
1185     [TDLEN]  = set_dlen,   [RDLEN]  = set_dlen,       [TCTL] = set_tctl,
1186     [TDT]    = set_tctl,   [MDIC]   = set_mdic,       [ICS]  = set_ics,
1187     [TDH]    = set_16bit,  [RDH]    = set_16bit,      [RDT]  = set_rdt,
1188     [IMC]    = set_imc,    [IMS]    = set_ims,        [ICR]  = set_icr,
1189     [EECD]   = set_eecd,   [RCTL]   = set_rx_control, [CTRL] = set_ctrl,
1190     [RDTR]   = set_16bit,  [RADV]   = set_16bit,      [TADV] = set_16bit,
1191     [ITR]    = set_16bit,
1192 
1193     [IP6AT ... IP6AT+3] = &mac_writereg, [IP4AT ... IP4AT+6] = &mac_writereg,
1194     [FFLT ... FFLT+6]   = &mac_writereg,
1195     [RA ... RA+31]      = &mac_writereg,
1196     [WUPM ... WUPM+31]  = &mac_writereg,
1197     [MTA ... MTA+127]   = &mac_writereg,
1198     [VFTA ... VFTA+127] = &mac_writereg,
1199     [FFMT ... FFMT+254] = &mac_writereg, [FFVT ... FFVT+254] = &mac_writereg,
1200     [PBM ... PBM+16383] = &mac_writereg,
1201 };
1202 
1203 enum { NWRITEOPS = ARRAY_SIZE(macreg_writeops) };
1204 
1205 enum { MAC_ACCESS_PARTIAL = 1, MAC_ACCESS_FLAG_NEEDED = 2 };
1206 
1207 #define markflag(x)    ((E1000_FLAG_##x << 2) | MAC_ACCESS_FLAG_NEEDED)
1208 /* In the array below the meaning of the bits is: [f|f|f|f|f|f|n|p]
1209  * f - flag bits (up to 6 possible flags)
1210  * n - flag needed
1211  * p - partially implenented */
1212 static const uint8_t mac_reg_access[0x8000] = {
1213     [RDTR]    = markflag(MIT),    [TADV]    = markflag(MIT),
1214     [RADV]    = markflag(MIT),    [ITR]     = markflag(MIT),
1215 
1216     [IPAV]    = markflag(MAC),    [WUC]     = markflag(MAC),
1217     [IP6AT]   = markflag(MAC),    [IP4AT]   = markflag(MAC),
1218     [FFVT]    = markflag(MAC),    [WUPM]    = markflag(MAC),
1219     [ECOL]    = markflag(MAC),    [MCC]     = markflag(MAC),
1220     [DC]      = markflag(MAC),    [TNCRS]   = markflag(MAC),
1221     [RLEC]    = markflag(MAC),    [XONRXC]  = markflag(MAC),
1222     [XOFFTXC] = markflag(MAC),    [RFC]     = markflag(MAC),
1223     [TSCTFC]  = markflag(MAC),    [MGTPRC]  = markflag(MAC),
1224     [WUS]     = markflag(MAC),    [AIT]     = markflag(MAC),
1225     [FFLT]    = markflag(MAC),    [FFMT]    = markflag(MAC),
1226     [SCC]     = markflag(MAC),    [FCRUC]   = markflag(MAC),
1227     [LATECOL] = markflag(MAC),    [COLC]    = markflag(MAC),
1228     [SEC]     = markflag(MAC),    [CEXTERR] = markflag(MAC),
1229     [XONTXC]  = markflag(MAC),    [XOFFRXC] = markflag(MAC),
1230     [RJC]     = markflag(MAC),    [RNBC]    = markflag(MAC),
1231     [MGTPDC]  = markflag(MAC),    [MGTPTC]  = markflag(MAC),
1232     [RUC]     = markflag(MAC),    [ROC]     = markflag(MAC),
1233     [GORCL]   = markflag(MAC),    [GORCH]   = markflag(MAC),
1234     [GOTCL]   = markflag(MAC),    [GOTCH]   = markflag(MAC),
1235     [BPRC]    = markflag(MAC),    [MPRC]    = markflag(MAC),
1236     [TSCTC]   = markflag(MAC),    [PRC64]   = markflag(MAC),
1237     [PRC127]  = markflag(MAC),    [PRC255]  = markflag(MAC),
1238     [PRC511]  = markflag(MAC),    [PRC1023] = markflag(MAC),
1239     [PRC1522] = markflag(MAC),    [PTC64]   = markflag(MAC),
1240     [PTC127]  = markflag(MAC),    [PTC255]  = markflag(MAC),
1241     [PTC511]  = markflag(MAC),    [PTC1023] = markflag(MAC),
1242     [PTC1522] = markflag(MAC),    [MPTC]    = markflag(MAC),
1243     [BPTC]    = markflag(MAC),
1244 
1245     [TDFH]  = markflag(MAC) | MAC_ACCESS_PARTIAL,
1246     [TDFT]  = markflag(MAC) | MAC_ACCESS_PARTIAL,
1247     [TDFHS] = markflag(MAC) | MAC_ACCESS_PARTIAL,
1248     [TDFTS] = markflag(MAC) | MAC_ACCESS_PARTIAL,
1249     [TDFPC] = markflag(MAC) | MAC_ACCESS_PARTIAL,
1250     [RDFH]  = markflag(MAC) | MAC_ACCESS_PARTIAL,
1251     [RDFT]  = markflag(MAC) | MAC_ACCESS_PARTIAL,
1252     [RDFHS] = markflag(MAC) | MAC_ACCESS_PARTIAL,
1253     [RDFTS] = markflag(MAC) | MAC_ACCESS_PARTIAL,
1254     [RDFPC] = markflag(MAC) | MAC_ACCESS_PARTIAL,
1255     [PBM]   = markflag(MAC) | MAC_ACCESS_PARTIAL,
1256 };
1257 
1258 static void
1259 e1000_mmio_write(void *opaque, hwaddr addr, uint64_t val,
1260                  unsigned size)
1261 {
1262     E1000State *s = opaque;
1263     unsigned int index = (addr & 0x1ffff) >> 2;
1264 
1265     if (index < NWRITEOPS && macreg_writeops[index]) {
1266         if (!(mac_reg_access[index] & MAC_ACCESS_FLAG_NEEDED)
1267             || (s->compat_flags & (mac_reg_access[index] >> 2))) {
1268             if (mac_reg_access[index] & MAC_ACCESS_PARTIAL) {
1269                 DBGOUT(GENERAL, "Writing to register at offset: 0x%08x. "
1270                        "It is not fully implemented.\n", index<<2);
1271             }
1272             macreg_writeops[index](s, index, val);
1273         } else {    /* "flag needed" bit is set, but the flag is not active */
1274             DBGOUT(MMIO, "MMIO write attempt to disabled reg. addr=0x%08x\n",
1275                    index<<2);
1276         }
1277     } else if (index < NREADOPS && macreg_readops[index]) {
1278         DBGOUT(MMIO, "e1000_mmio_writel RO %x: 0x%04"PRIx64"\n",
1279                index<<2, val);
1280     } else {
1281         DBGOUT(UNKNOWN, "MMIO unknown write addr=0x%08x,val=0x%08"PRIx64"\n",
1282                index<<2, val);
1283     }
1284 }
1285 
1286 static uint64_t
1287 e1000_mmio_read(void *opaque, hwaddr addr, unsigned size)
1288 {
1289     E1000State *s = opaque;
1290     unsigned int index = (addr & 0x1ffff) >> 2;
1291 
1292     if (index < NREADOPS && macreg_readops[index]) {
1293         if (!(mac_reg_access[index] & MAC_ACCESS_FLAG_NEEDED)
1294             || (s->compat_flags & (mac_reg_access[index] >> 2))) {
1295             if (mac_reg_access[index] & MAC_ACCESS_PARTIAL) {
1296                 DBGOUT(GENERAL, "Reading register at offset: 0x%08x. "
1297                        "It is not fully implemented.\n", index<<2);
1298             }
1299             return macreg_readops[index](s, index);
1300         } else {    /* "flag needed" bit is set, but the flag is not active */
1301             DBGOUT(MMIO, "MMIO read attempt of disabled reg. addr=0x%08x\n",
1302                    index<<2);
1303         }
1304     } else {
1305         DBGOUT(UNKNOWN, "MMIO unknown read addr=0x%08x\n", index<<2);
1306     }
1307     return 0;
1308 }
1309 
1310 static const MemoryRegionOps e1000_mmio_ops = {
1311     .read = e1000_mmio_read,
1312     .write = e1000_mmio_write,
1313     .endianness = DEVICE_LITTLE_ENDIAN,
1314     .impl = {
1315         .min_access_size = 4,
1316         .max_access_size = 4,
1317     },
1318 };
1319 
1320 static uint64_t e1000_io_read(void *opaque, hwaddr addr,
1321                               unsigned size)
1322 {
1323     E1000State *s = opaque;
1324 
1325     (void)s;
1326     return 0;
1327 }
1328 
1329 static void e1000_io_write(void *opaque, hwaddr addr,
1330                            uint64_t val, unsigned size)
1331 {
1332     E1000State *s = opaque;
1333 
1334     (void)s;
1335 }
1336 
1337 static const MemoryRegionOps e1000_io_ops = {
1338     .read = e1000_io_read,
1339     .write = e1000_io_write,
1340     .endianness = DEVICE_LITTLE_ENDIAN,
1341 };
1342 
1343 static bool is_version_1(void *opaque, int version_id)
1344 {
1345     return version_id == 1;
1346 }
1347 
1348 static void e1000_pre_save(void *opaque)
1349 {
1350     E1000State *s = opaque;
1351     NetClientState *nc = qemu_get_queue(s->nic);
1352 
1353     /* If the mitigation timer is active, emulate a timeout now. */
1354     if (s->mit_timer_on) {
1355         e1000_mit_timer(s);
1356     }
1357 
1358     /*
1359      * If link is down and auto-negotiation is supported and ongoing,
1360      * complete auto-negotiation immediately. This allows us to look
1361      * at MII_SR_AUTONEG_COMPLETE to infer link status on load.
1362      */
1363     if (nc->link_down && have_autoneg(s)) {
1364         s->phy_reg[PHY_STATUS] |= MII_SR_AUTONEG_COMPLETE;
1365     }
1366 }
1367 
1368 static int e1000_post_load(void *opaque, int version_id)
1369 {
1370     E1000State *s = opaque;
1371     NetClientState *nc = qemu_get_queue(s->nic);
1372 
1373     if (!chkflag(MIT)) {
1374         s->mac_reg[ITR] = s->mac_reg[RDTR] = s->mac_reg[RADV] =
1375             s->mac_reg[TADV] = 0;
1376         s->mit_irq_level = false;
1377     }
1378     s->mit_ide = 0;
1379     s->mit_timer_on = false;
1380 
1381     /* nc.link_down can't be migrated, so infer link_down according
1382      * to link status bit in mac_reg[STATUS].
1383      * Alternatively, restart link negotiation if it was in progress. */
1384     nc->link_down = (s->mac_reg[STATUS] & E1000_STATUS_LU) == 0;
1385 
1386     if (have_autoneg(s) &&
1387         !(s->phy_reg[PHY_STATUS] & MII_SR_AUTONEG_COMPLETE)) {
1388         nc->link_down = false;
1389         timer_mod(s->autoneg_timer,
1390                   qemu_clock_get_ms(QEMU_CLOCK_VIRTUAL) + 500);
1391     }
1392 
1393     return 0;
1394 }
1395 
1396 static bool e1000_mit_state_needed(void *opaque)
1397 {
1398     E1000State *s = opaque;
1399 
1400     return chkflag(MIT);
1401 }
1402 
1403 static bool e1000_full_mac_needed(void *opaque)
1404 {
1405     E1000State *s = opaque;
1406 
1407     return chkflag(MAC);
1408 }
1409 
1410 static const VMStateDescription vmstate_e1000_mit_state = {
1411     .name = "e1000/mit_state",
1412     .version_id = 1,
1413     .minimum_version_id = 1,
1414     .needed = e1000_mit_state_needed,
1415     .fields = (VMStateField[]) {
1416         VMSTATE_UINT32(mac_reg[RDTR], E1000State),
1417         VMSTATE_UINT32(mac_reg[RADV], E1000State),
1418         VMSTATE_UINT32(mac_reg[TADV], E1000State),
1419         VMSTATE_UINT32(mac_reg[ITR], E1000State),
1420         VMSTATE_BOOL(mit_irq_level, E1000State),
1421         VMSTATE_END_OF_LIST()
1422     }
1423 };
1424 
1425 static const VMStateDescription vmstate_e1000_full_mac_state = {
1426     .name = "e1000/full_mac_state",
1427     .version_id = 1,
1428     .minimum_version_id = 1,
1429     .needed = e1000_full_mac_needed,
1430     .fields = (VMStateField[]) {
1431         VMSTATE_UINT32_ARRAY(mac_reg, E1000State, 0x8000),
1432         VMSTATE_END_OF_LIST()
1433     }
1434 };
1435 
1436 static const VMStateDescription vmstate_e1000 = {
1437     .name = "e1000",
1438     .version_id = 2,
1439     .minimum_version_id = 1,
1440     .pre_save = e1000_pre_save,
1441     .post_load = e1000_post_load,
1442     .fields = (VMStateField[]) {
1443         VMSTATE_PCI_DEVICE(parent_obj, E1000State),
1444         VMSTATE_UNUSED_TEST(is_version_1, 4), /* was instance id */
1445         VMSTATE_UNUSED(4), /* Was mmio_base.  */
1446         VMSTATE_UINT32(rxbuf_size, E1000State),
1447         VMSTATE_UINT32(rxbuf_min_shift, E1000State),
1448         VMSTATE_UINT32(eecd_state.val_in, E1000State),
1449         VMSTATE_UINT16(eecd_state.bitnum_in, E1000State),
1450         VMSTATE_UINT16(eecd_state.bitnum_out, E1000State),
1451         VMSTATE_UINT16(eecd_state.reading, E1000State),
1452         VMSTATE_UINT32(eecd_state.old_eecd, E1000State),
1453         VMSTATE_UINT8(tx.props.ipcss, E1000State),
1454         VMSTATE_UINT8(tx.props.ipcso, E1000State),
1455         VMSTATE_UINT16(tx.props.ipcse, E1000State),
1456         VMSTATE_UINT8(tx.props.tucss, E1000State),
1457         VMSTATE_UINT8(tx.props.tucso, E1000State),
1458         VMSTATE_UINT16(tx.props.tucse, E1000State),
1459         VMSTATE_UINT32(tx.props.paylen, E1000State),
1460         VMSTATE_UINT8(tx.props.hdr_len, E1000State),
1461         VMSTATE_UINT16(tx.props.mss, E1000State),
1462         VMSTATE_UINT16(tx.size, E1000State),
1463         VMSTATE_UINT16(tx.tso_frames, E1000State),
1464         VMSTATE_UINT8(tx.props.sum_needed, E1000State),
1465         VMSTATE_INT8(tx.props.ip, E1000State),
1466         VMSTATE_INT8(tx.props.tcp, E1000State),
1467         VMSTATE_BUFFER(tx.header, E1000State),
1468         VMSTATE_BUFFER(tx.data, E1000State),
1469         VMSTATE_UINT16_ARRAY(eeprom_data, E1000State, 64),
1470         VMSTATE_UINT16_ARRAY(phy_reg, E1000State, 0x20),
1471         VMSTATE_UINT32(mac_reg[CTRL], E1000State),
1472         VMSTATE_UINT32(mac_reg[EECD], E1000State),
1473         VMSTATE_UINT32(mac_reg[EERD], E1000State),
1474         VMSTATE_UINT32(mac_reg[GPRC], E1000State),
1475         VMSTATE_UINT32(mac_reg[GPTC], E1000State),
1476         VMSTATE_UINT32(mac_reg[ICR], E1000State),
1477         VMSTATE_UINT32(mac_reg[ICS], E1000State),
1478         VMSTATE_UINT32(mac_reg[IMC], E1000State),
1479         VMSTATE_UINT32(mac_reg[IMS], E1000State),
1480         VMSTATE_UINT32(mac_reg[LEDCTL], E1000State),
1481         VMSTATE_UINT32(mac_reg[MANC], E1000State),
1482         VMSTATE_UINT32(mac_reg[MDIC], E1000State),
1483         VMSTATE_UINT32(mac_reg[MPC], E1000State),
1484         VMSTATE_UINT32(mac_reg[PBA], E1000State),
1485         VMSTATE_UINT32(mac_reg[RCTL], E1000State),
1486         VMSTATE_UINT32(mac_reg[RDBAH], E1000State),
1487         VMSTATE_UINT32(mac_reg[RDBAL], E1000State),
1488         VMSTATE_UINT32(mac_reg[RDH], E1000State),
1489         VMSTATE_UINT32(mac_reg[RDLEN], E1000State),
1490         VMSTATE_UINT32(mac_reg[RDT], E1000State),
1491         VMSTATE_UINT32(mac_reg[STATUS], E1000State),
1492         VMSTATE_UINT32(mac_reg[SWSM], E1000State),
1493         VMSTATE_UINT32(mac_reg[TCTL], E1000State),
1494         VMSTATE_UINT32(mac_reg[TDBAH], E1000State),
1495         VMSTATE_UINT32(mac_reg[TDBAL], E1000State),
1496         VMSTATE_UINT32(mac_reg[TDH], E1000State),
1497         VMSTATE_UINT32(mac_reg[TDLEN], E1000State),
1498         VMSTATE_UINT32(mac_reg[TDT], E1000State),
1499         VMSTATE_UINT32(mac_reg[TORH], E1000State),
1500         VMSTATE_UINT32(mac_reg[TORL], E1000State),
1501         VMSTATE_UINT32(mac_reg[TOTH], E1000State),
1502         VMSTATE_UINT32(mac_reg[TOTL], E1000State),
1503         VMSTATE_UINT32(mac_reg[TPR], E1000State),
1504         VMSTATE_UINT32(mac_reg[TPT], E1000State),
1505         VMSTATE_UINT32(mac_reg[TXDCTL], E1000State),
1506         VMSTATE_UINT32(mac_reg[WUFC], E1000State),
1507         VMSTATE_UINT32(mac_reg[VET], E1000State),
1508         VMSTATE_UINT32_SUB_ARRAY(mac_reg, E1000State, RA, 32),
1509         VMSTATE_UINT32_SUB_ARRAY(mac_reg, E1000State, MTA, 128),
1510         VMSTATE_UINT32_SUB_ARRAY(mac_reg, E1000State, VFTA, 128),
1511         VMSTATE_END_OF_LIST()
1512     },
1513     .subsections = (const VMStateDescription*[]) {
1514         &vmstate_e1000_mit_state,
1515         &vmstate_e1000_full_mac_state,
1516         NULL
1517     }
1518 };
1519 
1520 /*
1521  * EEPROM contents documented in Tables 5-2 and 5-3, pp. 98-102.
1522  * Note: A valid DevId will be inserted during pci_e1000_init().
1523  */
1524 static const uint16_t e1000_eeprom_template[64] = {
1525     0x0000, 0x0000, 0x0000, 0x0000,      0xffff, 0x0000,      0x0000, 0x0000,
1526     0x3000, 0x1000, 0x6403, 0 /*DevId*/, 0x8086, 0 /*DevId*/, 0x8086, 0x3040,
1527     0x0008, 0x2000, 0x7e14, 0x0048,      0x1000, 0x00d8,      0x0000, 0x2700,
1528     0x6cc9, 0x3150, 0x0722, 0x040b,      0x0984, 0x0000,      0xc000, 0x0706,
1529     0x1008, 0x0000, 0x0f04, 0x7fff,      0x4d01, 0xffff,      0xffff, 0xffff,
1530     0xffff, 0xffff, 0xffff, 0xffff,      0xffff, 0xffff,      0xffff, 0xffff,
1531     0x0100, 0x4000, 0x121c, 0xffff,      0xffff, 0xffff,      0xffff, 0xffff,
1532     0xffff, 0xffff, 0xffff, 0xffff,      0xffff, 0xffff,      0xffff, 0x0000,
1533 };
1534 
1535 /* PCI interface */
1536 
1537 static void
1538 e1000_mmio_setup(E1000State *d)
1539 {
1540     int i;
1541     const uint32_t excluded_regs[] = {
1542         E1000_MDIC, E1000_ICR, E1000_ICS, E1000_IMS,
1543         E1000_IMC, E1000_TCTL, E1000_TDT, PNPMMIO_SIZE
1544     };
1545 
1546     memory_region_init_io(&d->mmio, OBJECT(d), &e1000_mmio_ops, d,
1547                           "e1000-mmio", PNPMMIO_SIZE);
1548     memory_region_add_coalescing(&d->mmio, 0, excluded_regs[0]);
1549     for (i = 0; excluded_regs[i] != PNPMMIO_SIZE; i++)
1550         memory_region_add_coalescing(&d->mmio, excluded_regs[i] + 4,
1551                                      excluded_regs[i+1] - excluded_regs[i] - 4);
1552     memory_region_init_io(&d->io, OBJECT(d), &e1000_io_ops, d, "e1000-io", IOPORT_SIZE);
1553 }
1554 
1555 static void
1556 pci_e1000_uninit(PCIDevice *dev)
1557 {
1558     E1000State *d = E1000(dev);
1559 
1560     timer_del(d->autoneg_timer);
1561     timer_free(d->autoneg_timer);
1562     timer_del(d->mit_timer);
1563     timer_free(d->mit_timer);
1564     qemu_del_nic(d->nic);
1565 }
1566 
1567 static NetClientInfo net_e1000_info = {
1568     .type = NET_CLIENT_OPTIONS_KIND_NIC,
1569     .size = sizeof(NICState),
1570     .can_receive = e1000_can_receive,
1571     .receive = e1000_receive,
1572     .receive_iov = e1000_receive_iov,
1573     .link_status_changed = e1000_set_link_status,
1574 };
1575 
1576 static void e1000_write_config(PCIDevice *pci_dev, uint32_t address,
1577                                 uint32_t val, int len)
1578 {
1579     E1000State *s = E1000(pci_dev);
1580 
1581     pci_default_write_config(pci_dev, address, val, len);
1582 
1583     if (range_covers_byte(address, len, PCI_COMMAND) &&
1584         (pci_dev->config[PCI_COMMAND] & PCI_COMMAND_MASTER)) {
1585         qemu_flush_queued_packets(qemu_get_queue(s->nic));
1586     }
1587 }
1588 
1589 static void pci_e1000_realize(PCIDevice *pci_dev, Error **errp)
1590 {
1591     DeviceState *dev = DEVICE(pci_dev);
1592     E1000State *d = E1000(pci_dev);
1593     uint8_t *pci_conf;
1594     uint8_t *macaddr;
1595 
1596     pci_dev->config_write = e1000_write_config;
1597 
1598     pci_conf = pci_dev->config;
1599 
1600     /* TODO: RST# value should be 0, PCI spec 6.2.4 */
1601     pci_conf[PCI_CACHE_LINE_SIZE] = 0x10;
1602 
1603     pci_conf[PCI_INTERRUPT_PIN] = 1; /* interrupt pin A */
1604 
1605     e1000_mmio_setup(d);
1606 
1607     pci_register_bar(pci_dev, 0, PCI_BASE_ADDRESS_SPACE_MEMORY, &d->mmio);
1608 
1609     pci_register_bar(pci_dev, 1, PCI_BASE_ADDRESS_SPACE_IO, &d->io);
1610 
1611     qemu_macaddr_default_if_unset(&d->conf.macaddr);
1612     macaddr = d->conf.macaddr.a;
1613 
1614     e1000x_core_prepare_eeprom(d->eeprom_data,
1615                                e1000_eeprom_template,
1616                                sizeof(e1000_eeprom_template),
1617                                PCI_DEVICE_GET_CLASS(pci_dev)->device_id,
1618                                macaddr);
1619 
1620     d->nic = qemu_new_nic(&net_e1000_info, &d->conf,
1621                           object_get_typename(OBJECT(d)), dev->id, d);
1622 
1623     qemu_format_nic_info_str(qemu_get_queue(d->nic), macaddr);
1624 
1625     d->autoneg_timer = timer_new_ms(QEMU_CLOCK_VIRTUAL, e1000_autoneg_timer, d);
1626     d->mit_timer = timer_new_ns(QEMU_CLOCK_VIRTUAL, e1000_mit_timer, d);
1627 }
1628 
1629 static void qdev_e1000_reset(DeviceState *dev)
1630 {
1631     E1000State *d = E1000(dev);
1632     e1000_reset(d);
1633 }
1634 
1635 static Property e1000_properties[] = {
1636     DEFINE_NIC_PROPERTIES(E1000State, conf),
1637     DEFINE_PROP_BIT("autonegotiation", E1000State,
1638                     compat_flags, E1000_FLAG_AUTONEG_BIT, true),
1639     DEFINE_PROP_BIT("mitigation", E1000State,
1640                     compat_flags, E1000_FLAG_MIT_BIT, true),
1641     DEFINE_PROP_BIT("extra_mac_registers", E1000State,
1642                     compat_flags, E1000_FLAG_MAC_BIT, true),
1643     DEFINE_PROP_END_OF_LIST(),
1644 };
1645 
1646 typedef struct E1000Info {
1647     const char *name;
1648     uint16_t   device_id;
1649     uint8_t    revision;
1650     uint16_t   phy_id2;
1651 } E1000Info;
1652 
1653 static void e1000_class_init(ObjectClass *klass, void *data)
1654 {
1655     DeviceClass *dc = DEVICE_CLASS(klass);
1656     PCIDeviceClass *k = PCI_DEVICE_CLASS(klass);
1657     E1000BaseClass *e = E1000_DEVICE_CLASS(klass);
1658     const E1000Info *info = data;
1659 
1660     k->realize = pci_e1000_realize;
1661     k->exit = pci_e1000_uninit;
1662     k->romfile = "efi-e1000.rom";
1663     k->vendor_id = PCI_VENDOR_ID_INTEL;
1664     k->device_id = info->device_id;
1665     k->revision = info->revision;
1666     e->phy_id2 = info->phy_id2;
1667     k->class_id = PCI_CLASS_NETWORK_ETHERNET;
1668     set_bit(DEVICE_CATEGORY_NETWORK, dc->categories);
1669     dc->desc = "Intel Gigabit Ethernet";
1670     dc->reset = qdev_e1000_reset;
1671     dc->vmsd = &vmstate_e1000;
1672     dc->props = e1000_properties;
1673 }
1674 
1675 static void e1000_instance_init(Object *obj)
1676 {
1677     E1000State *n = E1000(obj);
1678     device_add_bootindex_property(obj, &n->conf.bootindex,
1679                                   "bootindex", "/ethernet-phy@0",
1680                                   DEVICE(n), NULL);
1681 }
1682 
1683 static const TypeInfo e1000_base_info = {
1684     .name          = TYPE_E1000_BASE,
1685     .parent        = TYPE_PCI_DEVICE,
1686     .instance_size = sizeof(E1000State),
1687     .instance_init = e1000_instance_init,
1688     .class_size    = sizeof(E1000BaseClass),
1689     .abstract      = true,
1690 };
1691 
1692 static const E1000Info e1000_devices[] = {
1693     {
1694         .name      = "e1000",
1695         .device_id = E1000_DEV_ID_82540EM,
1696         .revision  = 0x03,
1697         .phy_id2   = E1000_PHY_ID2_8254xx_DEFAULT,
1698     },
1699     {
1700         .name      = "e1000-82544gc",
1701         .device_id = E1000_DEV_ID_82544GC_COPPER,
1702         .revision  = 0x03,
1703         .phy_id2   = E1000_PHY_ID2_82544x,
1704     },
1705     {
1706         .name      = "e1000-82545em",
1707         .device_id = E1000_DEV_ID_82545EM_COPPER,
1708         .revision  = 0x03,
1709         .phy_id2   = E1000_PHY_ID2_8254xx_DEFAULT,
1710     },
1711 };
1712 
1713 static void e1000_register_types(void)
1714 {
1715     int i;
1716 
1717     type_register_static(&e1000_base_info);
1718     for (i = 0; i < ARRAY_SIZE(e1000_devices); i++) {
1719         const E1000Info *info = &e1000_devices[i];
1720         TypeInfo type_info = {};
1721 
1722         type_info.name = info->name;
1723         type_info.parent = TYPE_E1000_BASE;
1724         type_info.class_data = (void *)info;
1725         type_info.class_init = e1000_class_init;
1726         type_info.instance_init = e1000_instance_init;
1727 
1728         type_register(&type_info);
1729     }
1730 }
1731 
1732 type_init(e1000_register_types)
1733