The driver provides the recv() and send() functions which will
be required by SLOF's libnet code for receiving and sending
packets.

Signed-off-by: Thomas Huth <thuth@redhat.com>
---
 pc-bios/s390-ccw/netboot.mak  |   2 +-
 pc-bios/s390-ccw/virtio-net.c | 130 ++++++++++++++++++++++++++++++++++++++++++
 pc-bios/s390-ccw/virtio.c     |   9 ++-
 pc-bios/s390-ccw/virtio.h     |  13 ++++-
 4 files changed, 149 insertions(+), 5 deletions(-)
 create mode 100644 pc-bios/s390-ccw/virtio-net.c

diff --git a/pc-bios/s390-ccw/netboot.mak b/pc-bios/s390-ccw/netboot.mak
index 1eaa0d8..dd9119a 100644
--- a/pc-bios/s390-ccw/netboot.mak
+++ b/pc-bios/s390-ccw/netboot.mak
@@ -1,7 +1,7 @@
 
 SLOF_DIR := $(SRC_PATH)/roms/SLOF
 
-NETOBJS := start.o sclp.o virtio.o netmain.o sbrk.o libc.a
+NETOBJS := start.o sclp.o virtio.o virtio-net.o netmain.o sbrk.o libc.a
 
 LIBC_INC := -I$(SLOF_DIR)/lib/libc/include
 LIBNET_INC := -I$(SLOF_DIR)/lib/libnet
diff --git a/pc-bios/s390-ccw/virtio-net.c b/pc-bios/s390-ccw/virtio-net.c
new file mode 100644
index 0000000..290b018
--- /dev/null
+++ b/pc-bios/s390-ccw/virtio-net.c
@@ -0,0 +1,130 @@
+/*
+ * Virtio-net driver for the s390-ccw firmware
+ *
+ * Copyright 2017 Thomas Huth, Red Hat Inc.
+ *
+ * This code is free software; you can redistribute it and/or modify it
+ * under the terms of the GNU General Public License as published by the
+ * Free Software Foundation; either version 2 of the License, or (at your
+ * option) any later version.
+ */
+
+#include <stdint.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+#include <sys/socket.h>
+#include <ethernet.h>
+#include "s390-ccw.h"
+#include "virtio.h"
+
+#ifndef DEBUG_VIRTIO_NET
+#define DEBUG_VIRTIO_NET 0
+#endif
+
+#define VQ_RX 0         /* Receive queue */
+#define VQ_TX 1         /* Transmit queue */
+
+struct VirtioNetHdr {
+    uint8_t flags;
+    uint8_t gso_type;
+    uint16_t hdr_len;
+    uint16_t gso_size;
+    uint16_t csum_start;
+    uint16_t csum_offset;
+    /*uint16_t num_buffers;*/ /* Only with VIRTIO_NET_F_MRG_RXBUF or VIRTIO1 */
+};
+typedef struct VirtioNetHdr VirtioNetHdr;
+
+static uint16_t rx_last_idx;  /* Last index in receive queue "used" ring */
+
+int virtio_net_init(void *mac_addr)
+{
+    VDev *vdev = virtio_get_device();
+    VRing *rxvq = &vdev->vrings[VQ_RX];
+    void *buf;
+    int i;
+
+    virtio_setup_ccw(vdev);
+
+    memcpy(mac_addr, vdev->config.net.mac, ETH_ALEN);
+
+    for (i = 0; i < 64; i++) {
+        buf = malloc(ETH_MTU_SIZE + sizeof(VirtioNetHdr));
+        IPL_assert(buf != NULL, "Can not allocate memory for receive buffers");
+        vring_send_buf(rxvq, buf, ETH_MTU_SIZE + sizeof(VirtioNetHdr),
+                       VRING_DESC_F_WRITE);
+    }
+    vring_notify(rxvq);
+
+    return 0;
+}
+
+int send(int fd, const void *buf, int len, int flags)
+{
+    VirtioNetHdr tx_hdr;
+    VDev *vdev = virtio_get_device();
+    VRing *txvq = &vdev->vrings[VQ_TX];
+
+    /* Set up header - we do not use anything special, so simply clear it */
+    memset(&tx_hdr, 0, sizeof(tx_hdr));
+
+    vring_send_buf(txvq, &tx_hdr, sizeof(tx_hdr), VRING_DESC_F_NEXT);
+    vring_send_buf(txvq, (void *)buf, len, VRING_HIDDEN_IS_CHAIN);
+    while (!vr_poll(txvq)) {
+        yield();
+    }
+    if (drain_irqs(txvq->schid)) {
+        puts("send: drain irqs failed");
+        return -1;
+    }
+
+    return len;
+}
+
+int recv(int fd, void *buf, int maxlen, int flags)
+{
+    VDev *vdev = virtio_get_device();
+    VRing *rxvq = &vdev->vrings[VQ_RX];
+    int len, id;
+    uint8_t *pkt;
+
+    if (rx_last_idx == rxvq->used->idx) {
+        return 0;
+    }
+
+    len = rxvq->used->ring[rx_last_idx % rxvq->num].len - sizeof(VirtioNetHdr);
+    if (len > maxlen) {
+        puts("virtio-net: Receive buffer too small");
+        len = maxlen;
+    }
+    id = rxvq->used->ring[rx_last_idx % rxvq->num].id % rxvq->num;
+    pkt = (uint8_t *)(rxvq->desc[id].addr + sizeof(VirtioNetHdr));
+
+#if DEBUG_VIRTIO_NET   /* Dump packet */
+    int i;
+    printf("\nbuf %p: len=%i\n", (void *)rxvq->desc[id].addr, len);
+    for (i = 0; i < 64; i++) {
+        printf(" %02x", pkt[i]);
+        if ((i % 16) == 15) {
+            printf("\n");
+        }
+    }
+    printf("\n");
+#endif
+
+    /* Copy data to destination buffer */
+    memcpy(buf, pkt, len);
+
+    /* Mark buffer as available to the host again */
+    rxvq->avail->ring[rxvq->avail->idx % rxvq->num] = id;
+    rxvq->avail->idx = rxvq->avail->idx + 1;
+    vring_notify(rxvq);
+
+    /* Move index to next entry */
+    rx_last_idx = rx_last_idx + 1;
+
+    return len;
+}
diff --git a/pc-bios/s390-ccw/virtio.c b/pc-bios/s390-ccw/virtio.c
index 09ab291..729d926 100644
--- a/pc-bios/s390-ccw/virtio.c
+++ b/pc-bios/s390-ccw/virtio.c
@@ -149,7 +149,7 @@ static void vring_init(VRing *vr, VqInfo *info)
     debug_print_addr("init vr", vr);
 }
 
-static bool vring_notify(VRing *vr)
+bool vring_notify(VRing *vr)
 {
     vr->cookie = virtio_notify(vr->schid, vr->id, vr->cookie);
     return vr->cookie >= 0;
@@ -188,7 +188,7 @@ ulong get_second(void)
     return (get_clock() >> 12) / 1000000;
 }
 
-static int vr_poll(VRing *vr)
+int vr_poll(VRing *vr)
 {
     if (vr->used->idx == vr->used_idx) {
         vring_notify(vr);
@@ -261,6 +261,11 @@ void virtio_setup_ccw(VDev *vdev)
     run_ccw(vdev, CCW_CMD_VDEV_RESET, NULL, 0);
 
     switch (vdev->senseid.cu_model) {
+    case VIRTIO_ID_NET:
+        vdev->nr_vqs = 3;
+        vdev->cmd_vr_idx = 0;
+        cfg_size = sizeof(vdev->config.net);
+        break;
     case VIRTIO_ID_BLOCK:
         vdev->nr_vqs = 1;
         vdev->cmd_vr_idx = 0;
diff --git a/pc-bios/s390-ccw/virtio.h b/pc-bios/s390-ccw/virtio.h
index d743919..d6af022 100644
--- a/pc-bios/s390-ccw/virtio.h
+++ b/pc-bios/s390-ccw/virtio.h
@@ -11,8 +11,6 @@
 #ifndef VIRTIO_H
 #define VIRTIO_H
 
-#include "s390-ccw.h"
-
 /* Status byte for guest to report progress, and synchronize features. */
 /* We have seen device and processed generic fields (VIRTIO_CONFIG_F_VIRTIO) */
 #define VIRTIO_CONFIG_S_ACKNOWLEDGE     1
@@ -254,6 +252,12 @@ struct ScsiDevice {
 };
 typedef struct ScsiDevice ScsiDevice;
 
+struct VirtioNetConfig {
+    uint8_t mac[6];
+    uint16_t status;
+};
+typedef struct VirtioNetConfig VirtioNetConfig;
+
 struct VDev {
     int nr_vqs;
     VRing *vrings;
@@ -266,6 +270,7 @@ struct VDev {
     union {
         VirtioBlkConfig blk;
         VirtioScsiConfig scsi;
+        VirtioNetConfig net;
     } config;
     ScsiDevice *scsi_device;
     bool is_cdrom;
@@ -291,10 +296,14 @@ struct VirtioCmd {
 };
 typedef struct VirtioCmd VirtioCmd;
 
+bool vring_notify(VRing *vr);
 int drain_irqs(SubChannelId schid);
 void vring_send_buf(VRing *vr, void *p, int len, int flags);
+int vr_poll(VRing *vr);
 int vring_wait_reply(void);
 int virtio_run(VDev *vdev, int vqid, VirtioCmd *cmd);
 void virtio_setup_ccw(VDev *vdev);
 
+int virtio_net_init(void *mac_addr);
+
 #endif /* VIRTIO_H */
-- 
1.8.3.1

On Fri,  7 Jul 2017 12:27:03 +0200
Thomas Huth <thuth@redhat.com> wrote:

> The driver provides the recv() and send() functions which will
> be required by SLOF's libnet code for receiving and sending
> packets.

No thorough review yet, only some things I noticed.

> 
> Signed-off-by: Thomas Huth <thuth@redhat.com>
> ---
>  pc-bios/s390-ccw/netboot.mak  |   2 +-
>  pc-bios/s390-ccw/virtio-net.c | 130 ++++++++++++++++++++++++++++++++++++++++++
>  pc-bios/s390-ccw/virtio.c     |   9 ++-
>  pc-bios/s390-ccw/virtio.h     |  13 ++++-
>  4 files changed, 149 insertions(+), 5 deletions(-)
>  create mode 100644 pc-bios/s390-ccw/virtio-net.c
> 

> diff --git a/pc-bios/s390-ccw/virtio-net.c b/pc-bios/s390-ccw/virtio-net.c
> new file mode 100644
> index 0000000..290b018
> --- /dev/null
> +++ b/pc-bios/s390-ccw/virtio-net.c
> @@ -0,0 +1,130 @@
> +/*
> + * Virtio-net driver for the s390-ccw firmware
> + *
> + * Copyright 2017 Thomas Huth, Red Hat Inc.
> + *
> + * This code is free software; you can redistribute it and/or modify it
> + * under the terms of the GNU General Public License as published by the
> + * Free Software Foundation; either version 2 of the License, or (at your
> + * option) any later version.
> + */
> +
> +#include <stdint.h>
> +#include <stdbool.h>
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <string.h>
> +#include <unistd.h>
> +#include <sys/socket.h>
> +#include <ethernet.h>
> +#include "s390-ccw.h"
> +#include "virtio.h"
> +
> +#ifndef DEBUG_VIRTIO_NET
> +#define DEBUG_VIRTIO_NET 0
> +#endif
> +
> +#define VQ_RX 0         /* Receive queue */
> +#define VQ_TX 1         /* Transmit queue */
> +
> +struct VirtioNetHdr {
> +    uint8_t flags;
> +    uint8_t gso_type;
> +    uint16_t hdr_len;
> +    uint16_t gso_size;
> +    uint16_t csum_start;
> +    uint16_t csum_offset;
> +    /*uint16_t num_buffers;*/ /* Only with VIRTIO_NET_F_MRG_RXBUF or VIRTIO1 */
> +};
> +typedef struct VirtioNetHdr VirtioNetHdr;
> +
> +static uint16_t rx_last_idx;  /* Last index in receive queue "used" ring */
> +
> +int virtio_net_init(void *mac_addr)
> +{
> +    VDev *vdev = virtio_get_device();
> +    VRing *rxvq = &vdev->vrings[VQ_RX];
> +    void *buf;
> +    int i;
> +
> +    virtio_setup_ccw(vdev);
> +
> +    memcpy(mac_addr, vdev->config.net.mac, ETH_ALEN);

The mac address in the config space is only valid if the mac feature
flag has been negotiated (although the field always exists).

> +
> +    for (i = 0; i < 64; i++) {
> +        buf = malloc(ETH_MTU_SIZE + sizeof(VirtioNetHdr));
> +        IPL_assert(buf != NULL, "Can not allocate memory for receive buffers");
> +        vring_send_buf(rxvq, buf, ETH_MTU_SIZE + sizeof(VirtioNetHdr),
> +                       VRING_DESC_F_WRITE);
> +    }
> +    vring_notify(rxvq);
> +
> +    return 0;
> +}
> +
> +int send(int fd, const void *buf, int len, int flags)
> +{
> +    VirtioNetHdr tx_hdr;
> +    VDev *vdev = virtio_get_device();
> +    VRing *txvq = &vdev->vrings[VQ_TX];
> +
> +    /* Set up header - we do not use anything special, so simply clear it */
> +    memset(&tx_hdr, 0, sizeof(tx_hdr));
> +
> +    vring_send_buf(txvq, &tx_hdr, sizeof(tx_hdr), VRING_DESC_F_NEXT);
> +    vring_send_buf(txvq, (void *)buf, len, VRING_HIDDEN_IS_CHAIN);
> +    while (!vr_poll(txvq)) {
> +        yield();
> +    }
> +    if (drain_irqs(txvq->schid)) {
> +        puts("send: drain irqs failed");
> +        return -1;

Does the caller try again later? drain_irqs() errors may be transient.

> +    }
> +
> +    return len;
> +}
> +
> +int recv(int fd, void *buf, int maxlen, int flags)
> +{
> +    VDev *vdev = virtio_get_device();
> +    VRing *rxvq = &vdev->vrings[VQ_RX];
> +    int len, id;
> +    uint8_t *pkt;
> +
> +    if (rx_last_idx == rxvq->used->idx) {
> +        return 0;
> +    }
> +
> +    len = rxvq->used->ring[rx_last_idx % rxvq->num].len - sizeof(VirtioNetHdr);
> +    if (len > maxlen) {
> +        puts("virtio-net: Receive buffer too small");
> +        len = maxlen;
> +    }
> +    id = rxvq->used->ring[rx_last_idx % rxvq->num].id % rxvq->num;
> +    pkt = (uint8_t *)(rxvq->desc[id].addr + sizeof(VirtioNetHdr));
> +
> +#if DEBUG_VIRTIO_NET   /* Dump packet */
> +    int i;
> +    printf("\nbuf %p: len=%i\n", (void *)rxvq->desc[id].addr, len);
> +    for (i = 0; i < 64; i++) {
> +        printf(" %02x", pkt[i]);
> +        if ((i % 16) == 15) {
> +            printf("\n");
> +        }
> +    }
> +    printf("\n");
> +#endif
> +
> +    /* Copy data to destination buffer */
> +    memcpy(buf, pkt, len);
> +
> +    /* Mark buffer as available to the host again */
> +    rxvq->avail->ring[rxvq->avail->idx % rxvq->num] = id;
> +    rxvq->avail->idx = rxvq->avail->idx + 1;
> +    vring_notify(rxvq);
> +
> +    /* Move index to next entry */
> +    rx_last_idx = rx_last_idx + 1;
> +
> +    return len;
> +}
> diff --git a/pc-bios/s390-ccw/virtio.c b/pc-bios/s390-ccw/virtio.c
> index 09ab291..729d926 100644
> --- a/pc-bios/s390-ccw/virtio.c
> +++ b/pc-bios/s390-ccw/virtio.c
> @@ -149,7 +149,7 @@ static void vring_init(VRing *vr, VqInfo *info)
>      debug_print_addr("init vr", vr);
>  }
>  
> -static bool vring_notify(VRing *vr)
> +bool vring_notify(VRing *vr)
>  {
>      vr->cookie = virtio_notify(vr->schid, vr->id, vr->cookie);
>      return vr->cookie >= 0;
> @@ -188,7 +188,7 @@ ulong get_second(void)
>      return (get_clock() >> 12) / 1000000;
>  }
>  
> -static int vr_poll(VRing *vr)
> +int vr_poll(VRing *vr)
>  {
>      if (vr->used->idx == vr->used_idx) {
>          vring_notify(vr);
> @@ -261,6 +261,11 @@ void virtio_setup_ccw(VDev *vdev)
>      run_ccw(vdev, CCW_CMD_VDEV_RESET, NULL, 0);
>  
>      switch (vdev->senseid.cu_model) {
> +    case VIRTIO_ID_NET:
> +        vdev->nr_vqs = 3;

I think this should be 2, as you don't negotiate for the control vq.

> +        vdev->cmd_vr_idx = 0;
> +        cfg_size = sizeof(vdev->config.net);
> +        break;
>      case VIRTIO_ID_BLOCK:
>          vdev->nr_vqs = 1;
>          vdev->cmd_vr_idx = 0;
> diff --git a/pc-bios/s390-ccw/virtio.h b/pc-bios/s390-ccw/virtio.h
> index d743919..d6af022 100644
> --- a/pc-bios/s390-ccw/virtio.h
> +++ b/pc-bios/s390-ccw/virtio.h
> @@ -11,8 +11,6 @@
>  #ifndef VIRTIO_H
>  #define VIRTIO_H
>  
> -#include "s390-ccw.h"
> -
>  /* Status byte for guest to report progress, and synchronize features. */
>  /* We have seen device and processed generic fields (VIRTIO_CONFIG_F_VIRTIO) */
>  #define VIRTIO_CONFIG_S_ACKNOWLEDGE     1
> @@ -254,6 +252,12 @@ struct ScsiDevice {
>  };
>  typedef struct ScsiDevice ScsiDevice;
>  
> +struct VirtioNetConfig {
> +    uint8_t mac[6];
> +    uint16_t status;

I don't think status exists if you don't negotiate the respective
feature bit.

> +};
> +typedef struct VirtioNetConfig VirtioNetConfig;
> +
>  struct VDev {
>      int nr_vqs;
>      VRing *vrings;
> @@ -266,6 +270,7 @@ struct VDev {
>      union {
>          VirtioBlkConfig blk;
>          VirtioScsiConfig scsi;
> +        VirtioNetConfig net;
>      } config;
>      ScsiDevice *scsi_device;
>      bool is_cdrom;
> @@ -291,10 +296,14 @@ struct VirtioCmd {
>  };
>  typedef struct VirtioCmd VirtioCmd;
>  
> +bool vring_notify(VRing *vr);
>  int drain_irqs(SubChannelId schid);
>  void vring_send_buf(VRing *vr, void *p, int len, int flags);
> +int vr_poll(VRing *vr);
>  int vring_wait_reply(void);
>  int virtio_run(VDev *vdev, int vqid, VirtioCmd *cmd);
>  void virtio_setup_ccw(VDev *vdev);
>  
> +int virtio_net_init(void *mac_addr);
> +
>  #endif /* VIRTIO_H */

On 07.07.2017 14:37, Cornelia Huck wrote:
> On Fri,  7 Jul 2017 12:27:03 +0200
> Thomas Huth <thuth@redhat.com> wrote:
> 
>> The driver provides the recv() and send() functions which will
>> be required by SLOF's libnet code for receiving and sending
>> packets.
[...]
>> +int virtio_net_init(void *mac_addr)
>> +{
>> +    VDev *vdev = virtio_get_device();
>> +    VRing *rxvq = &vdev->vrings[VQ_RX];
>> +    void *buf;
>> +    int i;
>> +
>> +    virtio_setup_ccw(vdev);
>> +
>> +    memcpy(mac_addr, vdev->config.net.mac, ETH_ALEN);
> 
> The mac address in the config space is only valid if the mac feature
> flag has been negotiated (although the field always exists).

QEMU seems to always set it ... but yes, let's better negotiate
VIRTIO_NET_F_MAC here.

>> +
>> +    for (i = 0; i < 64; i++) {
>> +        buf = malloc(ETH_MTU_SIZE + sizeof(VirtioNetHdr));
>> +        IPL_assert(buf != NULL, "Can not allocate memory for receive buffers");
>> +        vring_send_buf(rxvq, buf, ETH_MTU_SIZE + sizeof(VirtioNetHdr),
>> +                       VRING_DESC_F_WRITE);
>> +    }
>> +    vring_notify(rxvq);
>> +
>> +    return 0;
>> +}
>> +
>> +int send(int fd, const void *buf, int len, int flags)
>> +{
>> +    VirtioNetHdr tx_hdr;
>> +    VDev *vdev = virtio_get_device();
>> +    VRing *txvq = &vdev->vrings[VQ_TX];
>> +
>> +    /* Set up header - we do not use anything special, so simply clear it */
>> +    memset(&tx_hdr, 0, sizeof(tx_hdr));
>> +
>> +    vring_send_buf(txvq, &tx_hdr, sizeof(tx_hdr), VRING_DESC_F_NEXT);
>> +    vring_send_buf(txvq, (void *)buf, len, VRING_HIDDEN_IS_CHAIN);
>> +    while (!vr_poll(txvq)) {
>> +        yield();
>> +    }
>> +    if (drain_irqs(txvq->schid)) {
>> +        puts("send: drain irqs failed");
>> +        return -1;
> 
> Does the caller try again later? drain_irqs() errors may be transient.

Yes, the TFTP code retries to send packets after a certain amount of
time (since the packets could also be lost in the network somewhere).

>> diff --git a/pc-bios/s390-ccw/virtio.c b/pc-bios/s390-ccw/virtio.c
>> index 09ab291..729d926 100644
>> --- a/pc-bios/s390-ccw/virtio.c
>> +++ b/pc-bios/s390-ccw/virtio.c
>> @@ -149,7 +149,7 @@ static void vring_init(VRing *vr, VqInfo *info)
>>      debug_print_addr("init vr", vr);
>>  }
>>  
>> -static bool vring_notify(VRing *vr)
>> +bool vring_notify(VRing *vr)
>>  {
>>      vr->cookie = virtio_notify(vr->schid, vr->id, vr->cookie);
>>      return vr->cookie >= 0;
>> @@ -188,7 +188,7 @@ ulong get_second(void)
>>      return (get_clock() >> 12) / 1000000;
>>  }
>>  
>> -static int vr_poll(VRing *vr)
>> +int vr_poll(VRing *vr)
>>  {
>>      if (vr->used->idx == vr->used_idx) {
>>          vring_notify(vr);
>> @@ -261,6 +261,11 @@ void virtio_setup_ccw(VDev *vdev)
>>      run_ccw(vdev, CCW_CMD_VDEV_RESET, NULL, 0);
>>  
>>      switch (vdev->senseid.cu_model) {
>> +    case VIRTIO_ID_NET:
>> +        vdev->nr_vqs = 3;
> 
> I think this should be 2, as you don't negotiate for the control vq.

OK.

>> +        vdev->cmd_vr_idx = 0;
>> +        cfg_size = sizeof(vdev->config.net);
>> +        break;
>>      case VIRTIO_ID_BLOCK:
>>          vdev->nr_vqs = 1;
>>          vdev->cmd_vr_idx = 0;
>> diff --git a/pc-bios/s390-ccw/virtio.h b/pc-bios/s390-ccw/virtio.h
>> index d743919..d6af022 100644
>> --- a/pc-bios/s390-ccw/virtio.h
>> +++ b/pc-bios/s390-ccw/virtio.h
>> @@ -11,8 +11,6 @@
>>  #ifndef VIRTIO_H
>>  #define VIRTIO_H
>>  
>> -#include "s390-ccw.h"
>> -
>>  /* Status byte for guest to report progress, and synchronize features. */
>>  /* We have seen device and processed generic fields (VIRTIO_CONFIG_F_VIRTIO) */
>>  #define VIRTIO_CONFIG_S_ACKNOWLEDGE     1
>> @@ -254,6 +252,12 @@ struct ScsiDevice {
>>  };
>>  typedef struct ScsiDevice ScsiDevice;
>>  
>> +struct VirtioNetConfig {
>> +    uint8_t mac[6];
>> +    uint16_t status;
> 
> I don't think status exists if you don't negotiate the respective
> feature bit.

OK.

>> +};
>> +typedef struct VirtioNetConfig VirtioNetConfig;
>> +
>>  struct VDev {
>>      int nr_vqs;
>>      VRing *vrings;
>> @@ -266,6 +270,7 @@ struct VDev {
>>      union {
>>          VirtioBlkConfig blk;
>>          VirtioScsiConfig scsi;
>> +        VirtioNetConfig net;
>>      } config;
>>      ScsiDevice *scsi_device;
>>      bool is_cdrom;
>> @@ -291,10 +296,14 @@ struct VirtioCmd {
>>  };
>>  typedef struct VirtioCmd VirtioCmd;
>>  
>> +bool vring_notify(VRing *vr);
>>  int drain_irqs(SubChannelId schid);
>>  void vring_send_buf(VRing *vr, void *p, int len, int flags);
>> +int vr_poll(VRing *vr);
>>  int vring_wait_reply(void);
>>  int virtio_run(VDev *vdev, int vqid, VirtioCmd *cmd);
>>  void virtio_setup_ccw(VDev *vdev);
>>  
>> +int virtio_net_init(void *mac_addr);
>> +
>>  #endif /* VIRTIO_H */
> 

Thanks for the review!

 Thomas

On Mon, 10 Jul 2017 13:23:07 +0200
Thomas Huth <thuth@redhat.com> wrote:

> On 07.07.2017 14:37, Cornelia Huck wrote:
> > On Fri,  7 Jul 2017 12:27:03 +0200
> > Thomas Huth <thuth@redhat.com> wrote:
> >   
> >> The driver provides the recv() and send() functions which will
> >> be required by SLOF's libnet code for receiving and sending
> >> packets.  
> [...]
> >> +int virtio_net_init(void *mac_addr)
> >> +{
> >> +    VDev *vdev = virtio_get_device();
> >> +    VRing *rxvq = &vdev->vrings[VQ_RX];
> >> +    void *buf;
> >> +    int i;
> >> +
> >> +    virtio_setup_ccw(vdev);
> >> +
> >> +    memcpy(mac_addr, vdev->config.net.mac, ETH_ALEN);  
> > 
> > The mac address in the config space is only valid if the mac feature
> > flag has been negotiated (although the field always exists).  
> 
> QEMU seems to always set it ... but yes, let's better negotiate
> VIRTIO_NET_F_MAC here.

Ah, feature negotiation in the bios, nice :)

> 
> >> +
> >> +    for (i = 0; i < 64; i++) {
> >> +        buf = malloc(ETH_MTU_SIZE + sizeof(VirtioNetHdr));
> >> +        IPL_assert(buf != NULL, "Can not allocate memory for receive buffers");
> >> +        vring_send_buf(rxvq, buf, ETH_MTU_SIZE + sizeof(VirtioNetHdr),
> >> +                       VRING_DESC_F_WRITE);
> >> +    }
> >> +    vring_notify(rxvq);
> >> +
> >> +    return 0;
> >> +}
> >> +
> >> +int send(int fd, const void *buf, int len, int flags)
> >> +{
> >> +    VirtioNetHdr tx_hdr;
> >> +    VDev *vdev = virtio_get_device();
> >> +    VRing *txvq = &vdev->vrings[VQ_TX];
> >> +
> >> +    /* Set up header - we do not use anything special, so simply clear it */
> >> +    memset(&tx_hdr, 0, sizeof(tx_hdr));
> >> +
> >> +    vring_send_buf(txvq, &tx_hdr, sizeof(tx_hdr), VRING_DESC_F_NEXT);
> >> +    vring_send_buf(txvq, (void *)buf, len, VRING_HIDDEN_IS_CHAIN);
> >> +    while (!vr_poll(txvq)) {
> >> +        yield();
> >> +    }
> >> +    if (drain_irqs(txvq->schid)) {
> >> +        puts("send: drain irqs failed");
> >> +        return -1;  
> > 
> > Does the caller try again later? drain_irqs() errors may be transient.  
> 
> Yes, the TFTP code retries to send packets after a certain amount of
> time (since the packets could also be lost in the network somewhere).

Sounds good. (Not that I expect many errors in drain_irqs()...)