Skip to contentSkip to conversation

VMXNET3 paravirtual NIC device implementation

PATCH Applied Dmitry Fleytman 4 participants 12 messages Mar 9, 2013 – Mar 28, 2013 Networking

lore
Share & tools
lore
#0 lore raw

v14 cover VMXNET3 paravirtual NIC device implementation

This set of patches implements VMWare VMXNET3 paravirtual NIC device. The device supports of all the device features including offload capabilties, VLANs and etc.
The device is tested on different OSes:

    Fedora 15
    Ubuntu 10.4
    Centos 6.2
    Windows 2008R2
    Windows 2008 64bit
    Windows 2008 32bit
    Windows 2003 64bit
    Windows 2003 32bit

Changes in V14:
QOM-related fixes

Changes in V13:
Licensing of vmxnet3.h changed to match original VMWare's header file license

Changes in V12:

QOM-related fixes

Changes in V11:
Rebased to master HEAD to fix multiqueue compilation issues

Changes in V10:

Issues reported by Stefan Hajnoczi fixed.

Changes in V9:

Issues reported by Stefan Hajnoczi fixed.

Changes in V8:

Issues reported by Stefan Hajnoczi reviewed and mostly fixed:

Review

Quoted 5 lines
(quoted diff) +5 -0
+        }
+        curr_src_off += src[i].iov_len;
+    }
+    return j;
+}

The existing iov_copy() function provides equivalent functionality. I don't think iov_rebuild() is needed.

[DF] Done. Thanks, missed it.

Review

Quoted 6 lines
(quoted diff) +6 -0
+            size -= len;
+        }
+        iovec_off += iov[i].iov_len;
+    }
+    return res;
+}

Rename this net_checksum_add_iov() and place it in net/checksum.c, then the new dependency on net from block can be dropped.

[DF] Done.

Quoted 5 lines

+vmw_shmem_read(hwaddr addr, void *buf, int len)

 {
     VMW_SHPRN("SHMEM r: %" PRIx64 ", len: %d to %p", addr, len, buf);
     cpu_physical_memory_read(addr, buf, len);

}

All changes to this file should be squashed with the previous patch.

[DF] Done

Review

Quoted 13 lines
(quoted diff) +9 -0
+#ifdef VMXNET_DEBUG_SHMEM_ACCESS
+#define VMW_SHPRN(fmt, ...)                                                  
 \
+    do {                                                                     
 \
+        printf("[%s][SH][%s]: " fmt "\n", VMXNET_DEVICE_NAME, __func__,      
 \
+            ## __VA_ARGS__);                                                 
 \
+    } while (0)
+#else
+#define VMW_SHPRN(fmt, ...) do {} while (0)
+#endif

Please use QEMU tracing. It eliminates all this boilerplate and conditional compilation. Tracing can be enabled/disabled at runtime and works with SystemTap/DTrace. See docs/tracing.txt.

[DF] We'd like to stick with compile time logic in this case becase of 2 reasons:
[DF] 1. These printouts are intended for reverse engineering/development only and there is
[DF] no need to enable them at run time
[DF] 2. There is a big number of printouts, all driver-device communication is traced,
[DF] they hit performance even on strongest x86 in case of run-time logic

Review

Quoted 5 lines
(quoted diff) +5 -0
+struct eth_header {
+    uint8_t  h_dest[ETH_ALEN];   /* destination eth addr */
+    uint8_t  h_source[ETH_ALEN]; /* source ether addr    */
+    uint16_t h_proto;            /* packet type ID field */
+};

Looks like it's copy-pasted stuff from /usr/include/linux/if_*.h, /usr/include/netinet/*.h, and friends. If the system-wide headers are included names will collide for some of the macros at least.

Did you check if the slirp/ definitions can be reused?

[DF] Yes, you are right. This is copy-pasted from different places. [DF] Slips definishing do not fully cover our needs.

I'd rather we import network header definitions once in a generic place into the source tree. That way vmxnet and other components don't need to redefine these structs.

[DF] Exaclty! Our intention is to create generic header with network definitions and make everyone use it. [DF] We can move our header to some shared place if you want, however I'd do it in parallel with cleanup [DF] of similar definitions in existing code and this is a big change that os out of scope of these patches.

Quoted 2 lines

+ *===========================================================================*/

Is this huge comment box a sign that the code should be split into a foo_tx.c and an foo_rx.c file?

[DF] As for me this file is not that big to be splitted (<800 lines), however I'll do this if you insist :)

Quoted 1 line

+size_t vmxnet_tx_pkt_send(VmxnetTxPktH pkt, NetClientState *vc)

'vc' is an old name that was used for VLANClientState. The struct has since been renamed to NetClientState and the rest of QEMU uses 'nc' instead of 'vc'.

[DF] Fixed. Thanks.

Review

Quoted 2 lines
(quoted diff) +2 -0
+/* tx module context handle */
+typedef void *VmxnetTxPktH;

Forward-declaring the struct is nicer:

typedef struct VmxnetTxPkt VmxnetTxPkt;

The definition of VmxnetTxPkt is still hidden from the caller but you avoid the void* and casting. In vmxnet_pkt.c define using:

struct VmxnetTxPkt {
    ...
};

[DF] Agree, fixed. Thanks.

hw/vmxnet_utils.h

Quoted 4 lines
hw/vmxnet_utils.h +0 -0
index 7fd9a01..fac3b7b 100644

Please squash these fixes into the previous patch.

[DF] Oops, my bad. Fixed.

hw/vmxnet_utils.h

Quoted 3 lines
hw/vmxnet_utils.h +3 -0
+        uint8_t msix_used;
+        /* Whether MSI support was installed successfully */
+        uint8_t msi_used;

These two fields should be bool.

hw/vmxnet_utils.h

Quoted 2 lines
hw/vmxnet_utils.h +2 -0
+        /* Whether automatic interrupts masking enabled */
+        uint8_t auto_int_masking;

bool

[DF] Done.

hw/vmxnet_utils.h

Quoted 9 lines
hw/vmxnet_utils.h +9 -0
+static inline void vmxnet3_flush_shmem_changes(void)
+{
+    /*
+     * Flush shared memory changes
+     * Needed before sending interrupt to guest to ensure
+     * it gets consistent memory state
+     */
+    smp_wmb();
+}

It's useful to document why a memory barrier is being used in each instance. Therefore hiding smp_wmb() inside a wrapper function isn't great.

[DF] Fixed.

Also, it's suspicious that smb_wmb() is used but no other barriers are used. What about a read memory barrier when accessing shared memory written by the guest?

[DF] VMWARE interface build a in safe way - you always read out data buffer (packet descriptor) with memcopy, [DF] and then check its last byte to see whether it was fully filled by driver.
[DF] In this case device doesn't need read barriers. Drivers need write barriers of course to secure [DF] proper order of writes.

Changes in V7:

Issues reported by Michael S. Tsirkin reviewed and mostly fixed:

File vmware_utils.h:

...

hw/vmxnet_utils.h

Quoted 7 lines
hw/vmxnet_utils.h +7 -0
+static inline void
+vmw_shmem_st64(target_phys_addr_t addr, uint64_t value)
+{
+    VMW_SHPRN("SHMEM store64: %" PRIx64 " (value %" PRIx64 ")", addr, value);
+    stq_le_phys(addr, value);
+}
+

Pls remove these wrappers. These are just memory stores. Our codebase is too large as it is without every driver wrapping all standard calls.

[DF] Idea behind this macros is to have printout in each of them [DF] Printouts are really needed when reverse-engineering windows guest drivers [DF] Since there is no windows drivers code this is the only way to understand [DF] sequences they use

Quoted 1 line

+/* MACROS for simplification of operations on array-style registers */

UPPERCASE ABUSE

[DF] Fixed

hw/vmxnet_utils.h

Quoted 2 lines
hw/vmxnet_utils.h +2 -0
+#define IS_MULTIREG_ADDR(addr, base, cnt, regsize)                 \
+    (((addr + 1) > (base)) && ((addr) < (base) + (cnt) * (regsize)))

Same as range_covers_byte(base, cnt * regsize, addr)?

[DF] Fixed, thanks

hw/vmxnet_utils.h

Quoted 4 lines
hw/vmxnet_utils.h +4 -0
+
+#define MULTIREG_IDX_BY_ADDR(addr, base, regsize)                  \
+    (((addr) - (base)) / (regsize))
+

Above two macros is all that's left. No objection but it does not say what they do - want to add minimal documentation?
And please prefix with VMWARE_ or something.

[DF] Prefix added, macros documented

File vmxnet_utils.h (and related with the same problems):

...

hw/vmxnet_utils.h

Quoted 6 lines

+ * See the COPYING file in the top-level directory.
+ *
+ */

hw/vmxnet_utils.h +3 -0
+
+#ifndef _VMXNET_UTILS_H_
+#define _VMXNET_UTILS_H_

Please do not start identifiers with _ followed by an uppercase lattters.

[DF] Fixed

...

hw/vmxnet_utils.h

Quoted 7 lines
hw/vmxnet_utils.h +7 -0
+
+struct eth_header {
+    uint8_t  h_dest[ETH_ALEN];   /* destination eth addr */
+    uint8_t  h_source[ETH_ALEN]; /* source ether addr    */
+    uint16_t h_proto;            /* packet type ID field */
+};
+

And fix struct definitions to
1. follow qemu coding style
[DF] It's not clear what's wrong with the coding style. Please, elaborate.
2. start with vmxnet
[DF] Since this is generic network definitions with no device specifics
[DF] I'm not sure it makes sense to add vmxnet prefix.
[DF] I'd say it makes sense to put them into some generic header files under "net" directory instead [DF] and clean up other devices that have their local definitions of the same structures. [DF] Please, advise.

I also don't really understand why are these functions split out - vmxnet is the only user, no?

[DF] Currently vmxnet3 is the only user, hovewer we have vmxnet2 implementation as well (not submitted yet) [DF] and it uses the same header

File vmxnet_pkt.h:

...

hw/vmxnet_utils.h

Quoted 3 lines
hw/vmxnet_utils.h +3 -0
+
+/* tx module context handle */
+typedef void *VmxnetTxPktH;

This gets you zero type safety and makes debugging impossible. Just use pointers like everyone does.

[DF] Handle type is added to hide VmxnetTxPkt structure into .c header [DF] for better encapsulation.
[DF] Should we drop it anyway?

Files vmxnet3.*:

...

hw/vmxnet_utils.h

Quoted 3 lines
hw/vmxnet_utils.h +3 -0
+
+    if (zero_region) {
+        vmw_shmem_set(pa, 0, size*cell_size);

spaces around *

[DF] Fixed

...

Quoted 7 lines

+#define vmxnet3_ring_dump(macro, ring_name, ridx, r) \
+ macro("%s#%d: base %" PRIx64 " size %lu cell_size %lu gen %d next %lu",
\
+ (ring_name), (ridx),
\
+ (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next)

make macros upper case

[DF] Fixed

...

hw/vmxnet_utils.h

Quoted 5 lines
hw/vmxnet_utils.h +5 -0
+static inline void vmxnet3_flush_shmem_changes(void)
+{
+    /*
+     * Flush shared memory changes
+     * Needed before transferring control to guest

what does 'transferring control to guest' mean?
[DF] Changed to
[DF] /*

[DF]     * Flush shared memory changes
[DF]     * Needed before sending interrupt to guest to ensure
[DF]     * it gets consistent memory state
[DF]     */

...

hw/vmxnet_utils.h

Quoted 3 lines
hw/vmxnet_utils.h +3 -0
+     */
+    smp_wmb();
+}

Don't use wrappers like this. They just hide bugs. For example it's not helpful before an interrupt in the function below.

[DF] I guess you are talking about vmxnet3_complete_packet() [DF] Strictly speaking barrier is a must because we change shared memory in [DF] vmxnet3_complete_packet()

[DF] And the wrapper is a good thing because its name explains its effect [DF] in a formal way as opposed to comments

...

hw/vmxnet_utils.h

Quoted 2 lines
hw/vmxnet_utils.h +2 -0
+    switch (status) {
+    case VMXNET3_PKT_STATUS_OK: {

don't put {} around cases: they align incorrectly
if it's too big move to a function.

[DF] Fixed

...

hw/vmxnet_utils.h

Quoted 5 lines
hw/vmxnet_utils.h +5 -0
+static bool
+vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
+{
+    size_t bytes_sent = 0;
+    bool res = true;

why = true? don't initialize just because.

[DF] Fixed

...

Quoted 6 lines

+/*
+ * VMWARE headers we got from Linux kernel do not fully comply QEMU coding
+ * standards in sense of types and defines used.
+ * Since we didn't want to change VMWARE code, following set of typedefs
+ * and defines needed to compile these headers with QEMU introduced.
+ */

No need for this now.
You can export headers and put them under linux-headers.

[DF] Not sure it is possible because the header as-is is not stand-alone and won't compile [DF] without changes. We extracted definitions we use from their header and dropped unused [DF] and kernel-specific stuff.
[DF} Please, advise.

...

Quoted 1 line

+ if (VMXNET3_OM_TSO == s->offload_mode) {

Don't do Yoda style like this

[DF] "Yoda" style removed everywhere

Changes in V6:
Fixed most of problems pointed out by Michael S. Tsirkin
The only issue still open is creation of shared place
with generic network structures and functions. Currently
all generic network code introduced by VMXNET3 resides in vmxnet_utils.c/h files. It could be moved to some shared location however we believe it is a matter of separate refactoring as there are a lot of copy-pasted definitions in almost every device and code cleanup efforts requred in order to create truly shared codebase.

Implemented suggestions by Anthony Liguori

Fixed incorrect checksum caclulation for some packets in SW offloads mode

Changes in V5:
MSI-X save/load implemented in the device instead of pci bus as suggested by Michael S. Tsirkin

Patches regrouped as suggested by Paolo Bonzini

Changes in V4:
Fixed a few problems uncovered by NETIO test suit Assertion on failure to initialize MSI/MSI-X replaced with warning message and fallback to Legacy/MSI respectively

Various coding style adjustments and patch split-up as suggested by Anthony
Liguori

Live migration support added

Changes in V3:
Fixed crash when net device that is used as network fronted has no virtio HDR support.
Task offloads emulation for cases when net device that is used as network fronted has no virtio HDR support.

Changes in V2:
License text changed accoring to community suggestions Standard license header from GPLv2+ - licensed QEMU files used

Dmitry Fleytman (5):
Checksum-related utility functions
net: iovec checksum calculator
Common definitions for VMWARE devices
Packet abstraction for VMWARE network devices
VMXNET3 device implementation

 default-configs/pci.mak |    1 +
 hw/Makefile.objs        |    2 +
 hw/pci/pci.h            |    1 +
 hw/vmware_utils.h       |  143 +++
 hw/vmxnet3.c            | 2461 +++++++++++++++++++++++++++++++++++++++++++++++
 hw/vmxnet3.h            |  760 +++++++++++++++
 hw/vmxnet_debug.h       |  115 +++
 hw/vmxnet_rx_pkt.c      |  187 ++++
 hw/vmxnet_rx_pkt.h      |  174 ++++
 hw/vmxnet_tx_pkt.c      |  567 +++++++++++
 hw/vmxnet_tx_pkt.h      |  148 +++
 include/net/checksum.h  |   26 +-
 include/net/eth.h       |  347 +++++++
 net/Makefile.objs       |    1 +
 net/checksum.c          |   42 +-
 net/eth.c               |  217 +++++
 16 files changed, 5185 insertions(+), 7 deletions(-)
 create mode 100644 hw/vmware_utils.h
 create mode 100644 hw/vmxnet3.c
 create mode 100644 hw/vmxnet3.h
 create mode 100644 hw/vmxnet_debug.h
 create mode 100644 hw/vmxnet_rx_pkt.c
 create mode 100644 hw/vmxnet_rx_pkt.h
 create mode 100644 hw/vmxnet_tx_pkt.c
 create mode 100644 hw/vmxnet_tx_pkt.h
 create mode 100644 include/net/eth.h
 create mode 100644 net/eth.c
Signature · 1 line
1.8.1.2
#1 lore raw

PATCH v14 1/5 Checksum-related utility functions

net_checksum_add_cont()
checksum calculation for scattered data with odd chunk sizes

net_raw_checksum()
checksum calculation for a buffer

---
 include/net/checksum.h | 14 +++++++++++++-
 net/checksum.c         | 13 +++++++------
 2 files changed, 20 insertions(+), 7 deletions(-)
include/net/checksum.h +13 -1
index 1f05298..3e7b93d 100644
@@ -20,10 +20,22 @@
 
 #include <stdint.h>
 
-uint32_t net_checksum_add(int len, uint8_t *buf);
+uint32_t net_checksum_add_cont(int len, uint8_t *buf, int seq);
 uint16_t net_checksum_finish(uint32_t sum);
 uint16_t net_checksum_tcpudp(uint16_t length, uint16_t proto,
                              uint8_t *addrs, uint8_t *buf);
 void net_checksum_calculate(uint8_t *data, int length);
 
+static inline uint32_t
+net_checksum_add(int len, uint8_t *buf)
+{
+    return net_checksum_add_cont(len, buf, 0);
+}
+
+static inline uint16_t
+net_raw_checksum(uint8_t *data, int length)
+{
+    return net_checksum_finish(net_checksum_add(length, data));
+}
+
 #endif /* QEMU_NET_CHECKSUM_H */
net/checksum.c +7 -6
index 9919b2e..4fa5563 100644
@@ -20,16 +20,17 @@
 #define PROTO_TCP  6
 #define PROTO_UDP 17
 
-uint32_t net_checksum_add(int len, uint8_t *buf)
+uint32_t net_checksum_add_cont(int len, uint8_t *buf, int seq)
 {
     uint32_t sum = 0;
     int i;
 
-    for (i = 0; i < len; i++) {
-	if (i & 1)
-	    sum += (uint32_t)buf[i];
-	else
-	    sum += (uint32_t)buf[i] << 8;
+    for (i = seq; i < seq + len; i++) {
+        if (i & 1) {
+            sum += (uint32_t)buf[i - seq];
+        } else {
+            sum += (uint32_t)buf[i - seq] << 8;
+        }
     }
     return sum;
 }
Signature · 1 line
1.8.1.2
#2 lore raw

PATCH v14 2/5 net: iovec checksum calculator

In reply to Dmitry Fleytman
---
 include/net/checksum.h | 12 ++++++++++++
 net/checksum.c         | 29 +++++++++++++++++++++++++++++
 2 files changed, 41 insertions(+)
include/net/checksum.h +12 -0
index 3e7b93d..80203fb 100644
@@ -38,4 +38,16 @@ net_raw_checksum(uint8_t *data, int length)
     return net_checksum_finish(net_checksum_add(length, data));
 }
 
+/**
+ * net_checksum_add_iov: scatter-gather vector checksumming
+ *
+ * @iov: input scatter-gather array
+ * @iov_cnt: number of array elements
+ * @iov_off: starting iov offset for checksumming
+ * @size: length of data to be checksummed
+ */
+uint32_t net_checksum_add_iov(const struct iovec *iov,
+                              const unsigned int iov_cnt,
+                              uint32_t iov_off, uint32_t size);
+
 #endif /* QEMU_NET_CHECKSUM_H */
net/checksum.c +29 -0
index 4fa5563..14c0855 100644
@@ -15,6 +15,7 @@
  *  along with this program; if not, see <http://www.gnu.org/licenses/>.
  */
 
+#include "qemu-common.h"
 #include "net/checksum.h"
 
 #define PROTO_TCP  6
@@ -84,3 +85,31 @@ void net_checksum_calculate(uint8_t *data, int length)
     data[14+hlen+csum_offset]   = csum >> 8;
     data[14+hlen+csum_offset+1] = csum & 0xff;
 }
+
+uint32_t
+net_checksum_add_iov(const struct iovec *iov, const unsigned int iov_cnt,
+                     uint32_t iov_off, uint32_t size)
+{
+    size_t iovec_off, buf_off;
+    unsigned int i;
+    uint32_t res = 0;
+    uint32_t seq = 0;
+
+    iovec_off = 0;
+    buf_off = 0;
+    for (i = 0; i < iov_cnt && size; i++) {
+        if (iov_off < (iovec_off + iov[i].iov_len)) {
+            size_t len = MIN((iovec_off + iov[i].iov_len) - iov_off , size);
+            void *chunk_buf = iov[i].iov_base + (iov_off - iovec_off);
+
+            res += net_checksum_add_cont(len, chunk_buf, seq);
+            seq += len;
+
+            buf_off += len;
+            iov_off += len;
+            size -= len;
+        }
+        iovec_off += iov[i].iov_len;
+    }
+    return res;
+}
Signature · 1 line
1.8.1.2
#3 lore raw

PATCH v14 3/5 Common definitions for VMWARE devices

In reply to Dmitry Fleytman
---
 hw/vmware_utils.h | 143 ++++++++++++++++++++++
 hw/vmxnet_debug.h | 115 ++++++++++++++++++
 include/net/eth.h | 347 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
 net/Makefile.objs |   1 +
 net/eth.c         | 217 ++++++++++++++++++++++++++++++++++
 5 files changed, 823 insertions(+)
 create mode 100644 hw/vmware_utils.h
 create mode 100644 hw/vmxnet_debug.h
 create mode 100644 include/net/eth.h
 create mode 100644 net/eth.c
hw/vmware_utils.h new file +143 -0
new file mode 100644
index 0000000..5307e2c
@@ -0,0 +1,143 @@
+/*
+ * QEMU VMWARE paravirtual devices - auxiliary code
+ *
+ * Copyright (c) 2012 Ravello Systems LTD (http://ravellosystems.com)
+ *
+ * Developed by Daynix Computing LTD (http://www.daynix.com)
+ *
+ * Authors:
+ * Dmitry Fleytman <[email protected]>
+ * Yan Vugenfirer <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2 or later.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef VMWARE_UTILS_H
+#define VMWARE_UTILS_H
+
+#include "qemu/range.h"
+
+#ifndef VMW_SHPRN
+#define VMW_SHPRN(fmt, ...) do {} while (0)
+#endif
+
+/*
+ * Shared memory access functions with byte swap support
+ * Each function contains printout for reverse-engineering needs
+ *
+ */
+static inline void
+vmw_shmem_read(hwaddr addr, void *buf, int len)
+{
+    VMW_SHPRN("SHMEM r: %" PRIx64 ", len: %d to %p", addr, len, buf);
+    cpu_physical_memory_read(addr, buf, len);
+}
+
+static inline void
+vmw_shmem_write(hwaddr addr, void *buf, int len)
+{
+    VMW_SHPRN("SHMEM w: %" PRIx64 ", len: %d to %p", addr, len, buf);
+    cpu_physical_memory_write(addr, buf, len);
+}
+
+static inline void
+vmw_shmem_rw(hwaddr addr, void *buf, int len, int is_write)
+{
+    VMW_SHPRN("SHMEM r/w: %" PRIx64 ", len: %d (to %p), is write: %d",
+              addr, len, buf, is_write);
+
+    cpu_physical_memory_rw(addr, buf, len, is_write);
+}
+
+static inline void
+vmw_shmem_set(hwaddr addr, uint8 val, int len)
+{
+    int i;
+    VMW_SHPRN("SHMEM set: %" PRIx64 ", len: %d (value 0x%X)", addr, len, val);
+
+    for (i = 0; i < len; i++) {
+        cpu_physical_memory_write(addr + i, &val, 1);
+    }
+}
+
+static inline uint32_t
+vmw_shmem_ld8(hwaddr addr)
+{
+    uint8_t res = ldub_phys(addr);
+    VMW_SHPRN("SHMEM load8: %" PRIx64 " (value 0x%X)", addr, res);
+    return res;
+}
+
+static inline void
+vmw_shmem_st8(hwaddr addr, uint8_t value)
+{
+    VMW_SHPRN("SHMEM store8: %" PRIx64 " (value 0x%X)", addr, value);
+    stb_phys(addr, value);
+}
+
+static inline uint32_t
+vmw_shmem_ld16(hwaddr addr)
+{
+    uint16_t res = lduw_le_phys(addr);
+    VMW_SHPRN("SHMEM load16: %" PRIx64 " (value 0x%X)", addr, res);
+    return res;
+}
+
+static inline void
+vmw_shmem_st16(hwaddr addr, uint16_t value)
+{
+    VMW_SHPRN("SHMEM store16: %" PRIx64 " (value 0x%X)", addr, value);
+    stw_le_phys(addr, value);
+}
+
+static inline uint32_t
+vmw_shmem_ld32(hwaddr addr)
+{
+    uint32_t res = ldl_le_phys(addr);
+    VMW_SHPRN("SHMEM load32: %" PRIx64 " (value 0x%X)", addr, res);
+    return res;
+}
+
+static inline void
+vmw_shmem_st32(hwaddr addr, uint32_t value)
+{
+    VMW_SHPRN("SHMEM store32: %" PRIx64 " (value 0x%X)", addr, value);
+    stl_le_phys(addr, value);
+}
+
+static inline uint64_t
+vmw_shmem_ld64(hwaddr addr)
+{
+    uint64_t res = ldq_le_phys(addr);
+    VMW_SHPRN("SHMEM load64: %" PRIx64 " (value %" PRIx64 ")", addr, res);
+    return res;
+}
+
+static inline void
+vmw_shmem_st64(hwaddr addr, uint64_t value)
+{
+    VMW_SHPRN("SHMEM store64: %" PRIx64 " (value %" PRIx64 ")", addr, value);
+    stq_le_phys(addr, value);
+}
+
+/* Macros for simplification of operations on array-style registers */
+
+/*
+ * Whether <addr> lies inside of array-style register defined by <base>,
+ * number of elements (<cnt>) and element size (<regsize>)
+ *
+*/
+#define VMW_IS_MULTIREG_ADDR(addr, base, cnt, regsize)                 \
+    range_covers_byte(base, cnt * regsize, addr)
+
+/*
+ * Returns index of given register (<addr>) in array-style register defined by
+ * <base> and element size (<regsize>)
+ *
+*/
+#define VMW_MULTIREG_IDX_BY_ADDR(addr, base, regsize)                  \
+    (((addr) - (base)) / (regsize))
+
+#endif
hw/vmxnet_debug.h new file +115 -0

115 lines not on this page · raw message

include/net/eth.h new file +347 -0

347 lines not on this page · raw message

net/Makefile.objs +1 -0
index a08cd14..4854a14 100644
@@ -1,6 +1,7 @@
 common-obj-y = net.o queue.o checksum.o util.o hub.o
 common-obj-y += socket.o
 common-obj-y += dump.o
+common-obj-y += eth.o
 common-obj-$(CONFIG_POSIX) += tap.o
 common-obj-$(CONFIG_LINUX) += tap-linux.o
 common-obj-$(CONFIG_WIN32) += tap-win32.o
net/eth.c new file +217 -0

217 lines not on this page · raw message

Signature · 1 line
1.8.1.2
#4 lore raw

PATCH v14 4/5 Packet abstraction for VMWARE network devices

In reply to Dmitry Fleytman
---
 hw/Makefile.objs   |   1 +
 hw/vmxnet_rx_pkt.c | 187 ++++++++++++++++++
 hw/vmxnet_rx_pkt.h | 174 ++++++++++++++++
 hw/vmxnet_tx_pkt.c | 567 +++++++++++++++++++++++++++++++++++++++++++++++++++++
 hw/vmxnet_tx_pkt.h | 148 ++++++++++++++
 5 files changed, 1077 insertions(+)
 create mode 100644 hw/vmxnet_rx_pkt.c
 create mode 100644 hw/vmxnet_rx_pkt.h
 create mode 100644 hw/vmxnet_tx_pkt.c
 create mode 100644 hw/vmxnet_tx_pkt.h
hw/Makefile.objs +1 -0
index 40ebe46..14922cb 100644
@@ -119,6 +119,7 @@ common-obj-$(CONFIG_PCNET_PCI) += pcnet-pci.o
 common-obj-$(CONFIG_PCNET_COMMON) += pcnet.o
 common-obj-$(CONFIG_E1000_PCI) += e1000.o
 common-obj-$(CONFIG_RTL8139_PCI) += rtl8139.o
+common-obj-$(CONFIG_VMXNET3_PCI) += vmxnet_tx_pkt.o vmxnet_rx_pkt.o
 
 common-obj-$(CONFIG_SMC91C111) += smc91c111.o
 common-obj-$(CONFIG_LAN9118) += lan9118.o
hw/vmxnet_rx_pkt.c new file +187 -0

187 lines not on this page · raw message

hw/vmxnet_rx_pkt.h new file +174 -0

174 lines not on this page · raw message

hw/vmxnet_tx_pkt.c new file +567 -0

567 lines not on this page · raw message

hw/vmxnet_tx_pkt.h new file +148 -0

148 lines not on this page · raw message

Signature · 1 line
1.8.1.2
#5 lore raw

PATCH v14 5/5 VMXNET3 device implementation

In reply to Dmitry Fleytman
---
 default-configs/pci.mak |    1 +
 hw/Makefile.objs        |    1 +
 hw/pci/pci.h            |    1 +
 hw/vmxnet3.c            | 2461 +++++++++++++++++++++++++++++++++++++++++++++++
 hw/vmxnet3.h            |  760 +++++++++++++++
 5 files changed, 3224 insertions(+)
 create mode 100644 hw/vmxnet3.c
 create mode 100644 hw/vmxnet3.h
default-configs/pci.mak +1 -0
index ee2d18d..ce56d58 100644
@@ -13,6 +13,7 @@ CONFIG_LSI_SCSI_PCI=y
 CONFIG_MEGASAS_SCSI_PCI=y
 CONFIG_RTL8139_PCI=y
 CONFIG_E1000_PCI=y
+CONFIG_VMXNET3_PCI=y
 CONFIG_IDE_CORE=y
 CONFIG_IDE_QDEV=y
 CONFIG_IDE_PCI=y
hw/Makefile.objs +1 -0
index 14922cb..026aff6 100644
@@ -120,6 +120,7 @@ common-obj-$(CONFIG_PCNET_COMMON) += pcnet.o
 common-obj-$(CONFIG_E1000_PCI) += e1000.o
 common-obj-$(CONFIG_RTL8139_PCI) += rtl8139.o
 common-obj-$(CONFIG_VMXNET3_PCI) += vmxnet_tx_pkt.o vmxnet_rx_pkt.o
+common-obj-$(CONFIG_VMXNET3_PCI) += vmxnet3.o
 
 common-obj-$(CONFIG_SMC91C111) += smc91c111.o
 common-obj-$(CONFIG_LAN9118) += lan9118.o
hw/pci/pci.h +1 -0
index f340fe5..3beb70b 100644
@@ -60,6 +60,7 @@
 #define PCI_DEVICE_ID_VMWARE_NET         0x0720
 #define PCI_DEVICE_ID_VMWARE_SCSI        0x0730
 #define PCI_DEVICE_ID_VMWARE_IDE         0x1729
+#define PCI_DEVICE_ID_VMWARE_VMXNET3     0x07B0
 
 /* Intel (0x8086) */
 #define PCI_DEVICE_ID_INTEL_82551IT      0x1209
hw/vmxnet3.c new file +2461 -0

2,461 lines not on this page · raw message

hw/vmxnet3.h new file +760 -0

518 lines not on this page · raw message

The rest of this diff is not shown. View the raw message.

Signature · 1 line
1.8.1.2

Part of the diff is not shown here. View the raw message.

#6 lore raw

Re: [Qemu-devel] [PATCH V14 5/5] VMXNET3 device implementation

Quoted 2,552 lines · Dmitry Fleytman

---

default-configs/pci.mak |    1 +
hw/Makefile.objs        |    1 +
hw/pci/pci.h            |    1 +
hw/vmxnet3.c            | 2461 +++++++++++++++++++++++++++++++++++++++++++++++
hw/vmxnet3.h            |  760 +++++++++++++++
5 files changed, 3224 insertions(+)

create mode 100644 hw/vmxnet3.c
create mode 100644 hw/vmxnet3.h

default-configs/pci.mak +0 -0
index ee2d18d..ce56d58 100644
@@ -13,6 +13,7 @@ CONFIG_LSI_SCSI_PCI=y

CONFIG_MEGASAS_SCSI_PCI=y
CONFIG_RTL8139_PCI=y
CONFIG_E1000_PCI=y
+CONFIG_VMXNET3_PCI=y
CONFIG_IDE_CORE=y
CONFIG_IDE_QDEV=y
CONFIG_IDE_PCI=y

hw/Makefile.objs +0 -0
index 14922cb..026aff6 100644
@@ -120,6 +120,7 @@ common-obj-$(CONFIG_PCNET_COMMON) += pcnet.o

common-obj-$(CONFIG_E1000_PCI) += e1000.o common-obj-$(CONFIG_RTL8139_PCI) += rtl8139.o common-obj-$(CONFIG_VMXNET3_PCI) += vmxnet_tx_pkt.o vmxnet_rx_pkt.o +common-obj-$(CONFIG_VMXNET3_PCI) += vmxnet3.o

common-obj-$(CONFIG_SMC91C111) += smc91c111.o common-obj-$(CONFIG_LAN9118) += lan9118.o

hw/pci/pci.h +1 -0
index f340fe5..3beb70b 100644
@@ -60,6 +60,7 @@
 #define PCI_DEVICE_ID_VMWARE_NET         0x0720
 #define PCI_DEVICE_ID_VMWARE_SCSI        0x0730
 #define PCI_DEVICE_ID_VMWARE_IDE         0x1729
+#define PCI_DEVICE_ID_VMWARE_VMXNET3     0x07B0
 
 /* Intel (0x8086) */
 #define PCI_DEVICE_ID_INTEL_82551IT      0x1209
hw/vmxnet3.c new file +2461 -0
new file mode 100644
index 0000000..925be80
@@ -0,0 +1,2461 @@
+/*
+ * QEMU VMWARE VMXNET3 paravirtual NIC
+ *
+ * Copyright (c) 2012 Ravello Systems LTD (http://ravellosystems.com)
+ *
+ * Developed by Daynix Computing LTD (http://www.daynix.com)
+ *
+ * Authors:
+ * Dmitry Fleytman <[email protected]>
+ * Tamir Shomer <[email protected]>
+ * Yan Vugenfirer <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#include "hw.h"
+#include "pci/pci.h"
+#include "net/net.h"
+#include "virtio-net.h"
+#include "net/tap.h"
+#include "net/checksum.h"
+#include "sysemu/sysemu.h"
+#include "qemu-common.h"
+#include "qemu/bswap.h"
+#include "pci/msix.h"
+#include "pci/msi.h"
+
+#include "vmxnet3.h"
+#include "vmxnet_debug.h"
+#include "vmware_utils.h"
+#include "vmxnet_tx_pkt.h"
+#include "vmxnet_rx_pkt.h"
+
+#define PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION 0x1
+#define VMXNET3_MSIX_BAR_SIZE 0x2000
+
+#define VMXNET3_BAR0_IDX      (0)
+#define VMXNET3_BAR1_IDX      (1)
+#define VMXNET3_MSIX_BAR_IDX  (2)
+
+#define VMXNET3_OFF_MSIX_TABLE (0x000)
+#define VMXNET3_OFF_MSIX_PBA   (0x800)
+
+/* Link speed in Mbps should be shifted by 16 */
+#define VMXNET3_LINK_SPEED      (1000 << 16)
+
+/* Link status: 1 - up, 0 - down. */
+#define VMXNET3_LINK_STATUS_UP  0x1
+
+/* Least significant bit should be set for revision and version */
+#define VMXNET3_DEVICE_VERSION    0x1
+#define VMXNET3_DEVICE_REVISION   0x1
+
+/* Macros for rings descriptors access */
+#define VMXNET3_READ_TX_QUEUE_DESCR8(dpa, field) \
+    (vmw_shmem_ld8(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
+
+#define VMXNET3_WRITE_TX_QUEUE_DESCR8(dpa, field, value) \
+    (vmw_shmem_st8(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field, value)))
+
+#define VMXNET3_READ_TX_QUEUE_DESCR32(dpa, field) \
+    (vmw_shmem_ld32(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
+
+#define VMXNET3_WRITE_TX_QUEUE_DESCR32(dpa, field, value) \
+    (vmw_shmem_st32(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
+
+#define VMXNET3_READ_TX_QUEUE_DESCR64(dpa, field) \
+    (vmw_shmem_ld64(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field)))
+
+#define VMXNET3_WRITE_TX_QUEUE_DESCR64(dpa, field, value) \
+    (vmw_shmem_st64(dpa + offsetof(struct Vmxnet3_TxQueueDesc, field), value))
+
+#define VMXNET3_READ_RX_QUEUE_DESCR64(dpa, field) \
+    (vmw_shmem_ld64(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
+
+#define VMXNET3_READ_RX_QUEUE_DESCR32(dpa, field) \
+    (vmw_shmem_ld32(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field)))
+
+#define VMXNET3_WRITE_RX_QUEUE_DESCR64(dpa, field, value) \
+    (vmw_shmem_st64(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
+
+#define VMXNET3_WRITE_RX_QUEUE_DESCR8(dpa, field, value) \
+    (vmw_shmem_st8(dpa + offsetof(struct Vmxnet3_RxQueueDesc, field), value))
+
+/* Macros for guest driver shared area access */
+#define VMXNET3_READ_DRV_SHARED64(shpa, field) \
+    (vmw_shmem_ld64(shpa + offsetof(struct Vmxnet3_DriverShared, field)))
+
+#define VMXNET3_READ_DRV_SHARED32(shpa, field) \
+    (vmw_shmem_ld32(shpa + offsetof(struct Vmxnet3_DriverShared, field)))
+
+#define VMXNET3_WRITE_DRV_SHARED32(shpa, field, val) \
+    (vmw_shmem_st32(shpa + offsetof(struct Vmxnet3_DriverShared, field), val))
+
+#define VMXNET3_READ_DRV_SHARED16(shpa, field) \
+    (vmw_shmem_ld16(shpa + offsetof(struct Vmxnet3_DriverShared, field)))
+
+#define VMXNET3_READ_DRV_SHARED8(shpa, field) \
+    (vmw_shmem_ld8(shpa + offsetof(struct Vmxnet3_DriverShared, field)))
+
+#define VMXNET3_READ_DRV_SHARED(shpa, field, b, l) \
+    (vmw_shmem_read(shpa + offsetof(struct Vmxnet3_DriverShared, field), b, l))
+
+#define VMXNET_FLAG_IS_SET(field, flag) (((field) & (flag)) == (flag))
+
+#define TYPE_VMXNET3 "vmxnet3"
+#define VMXNET3(obj) OBJECT_CHECK(VMXNET3State, (obj), TYPE_VMXNET3)
+
+/* Cyclic ring abstraction */
+typedef struct {
+    hwaddr pa;
+    size_t size;
+    size_t cell_size;
+    size_t next;
+    uint8_t gen;
+} Vmxnet3Ring;
+
+static inline void vmxnet3_ring_init(Vmxnet3Ring *ring,
+                                     hwaddr pa,
+                                     size_t size,
+                                     size_t cell_size,
+                                     bool zero_region)
+{
+    ring->pa = pa;
+    ring->size = size;
+    ring->cell_size = cell_size;
+    ring->gen = VMXNET3_INIT_GEN;
+    ring->next = 0;
+
+    if (zero_region) {
+        vmw_shmem_set(pa, 0, size * cell_size);
+    }
+}
+
+#define VMXNET3_RING_DUMP(macro, ring_name, ridx, r)                         \
+    macro("%s#%d: base %" PRIx64 " size %lu cell_size %lu gen %d next %lu",  \
+          (ring_name), (ridx),                                               \
+          (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next)
+
+static inline void vmxnet3_ring_inc(Vmxnet3Ring *ring)
+{
+    if (++ring->next >= ring->size) {
+        ring->next = 0;
+        ring->gen ^= 1;
+    }
+}
+
+static inline void vmxnet3_ring_dec(Vmxnet3Ring *ring)
+{
+    if (ring->next-- == 0) {
+        ring->next = ring->size - 1;
+        ring->gen ^= 1;
+    }
+}
+
+static inline hwaddr vmxnet3_ring_curr_cell_pa(Vmxnet3Ring *ring)
+{
+    return ring->pa + ring->next * ring->cell_size;
+}
+
+static inline void vmxnet3_ring_read_curr_cell(Vmxnet3Ring *ring, void *buff)
+{
+    vmw_shmem_read(vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
+}
+
+static inline void vmxnet3_ring_write_curr_cell(Vmxnet3Ring *ring, void *buff)
+{
+    vmw_shmem_write(vmxnet3_ring_curr_cell_pa(ring), buff, ring->cell_size);
+}
+
+static inline size_t vmxnet3_ring_curr_cell_idx(Vmxnet3Ring *ring)
+{
+    return ring->next;
+}
+
+static inline uint8_t vmxnet3_ring_curr_gen(Vmxnet3Ring *ring)
+{
+    return ring->gen;
+}
+
+/* Debug trace-related functions */
+static inline void
+vmxnet3_dump_tx_descr(struct Vmxnet3_TxDesc *descr)
+{
+    VMW_PKPRN("TX DESCR: "
+              "addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
+              "dtype: %d, ext1: %d, msscof: %d, hlen: %d, om: %d, "
+              "eop: %d, cq: %d, ext2: %d, ti: %d, tci: %d",
+              le64_to_cpu(descr->addr), descr->len, descr->gen, descr->rsvd,
+              descr->dtype, descr->ext1, descr->msscof, descr->hlen, descr->om,
+              descr->eop, descr->cq, descr->ext2, descr->ti, descr->tci);
+}
+
+static inline void
+vmxnet3_dump_virt_hdr(struct virtio_net_hdr *vhdr)
+{
+    VMW_PKPRN("VHDR: flags 0x%x, gso_type: 0x%x, hdr_len: %d, gso_size: %d, "
+              "csum_start: %d, csum_offset: %d",
+              vhdr->flags, vhdr->gso_type, vhdr->hdr_len, vhdr->gso_size,
+              vhdr->csum_start, vhdr->csum_offset);
+}
+
+static inline void
+vmxnet3_dump_rx_descr(struct Vmxnet3_RxDesc *descr)
+{
+    VMW_PKPRN("RX DESCR: addr %" PRIx64 ", len: %d, gen: %d, rsvd: %d, "
+              "dtype: %d, ext1: %d, btype: %d",
+              le64_to_cpu(descr->addr), descr->len, descr->gen,
+              descr->rsvd, descr->dtype, descr->ext1, descr->btype);
+}
+
+/* Device state and helper functions */
+#define VMXNET3_RX_RINGS_PER_QUEUE (2)
+
+typedef struct {
+    Vmxnet3Ring tx_ring;
+    Vmxnet3Ring comp_ring;
+
+    uint8_t intr_idx;
+    hwaddr tx_stats_pa;
+    struct UPT1_TxStats txq_stats;
+} Vmxnet3TxqDescr;
+
+typedef struct {
+    Vmxnet3Ring rx_ring[VMXNET3_RX_RINGS_PER_QUEUE];
+    Vmxnet3Ring comp_ring;
+    uint8_t intr_idx;
+    hwaddr rx_stats_pa;
+    struct UPT1_RxStats rxq_stats;
+} Vmxnet3RxqDescr;
+
+typedef struct {
+    bool is_masked;
+    bool is_pending;
+    bool is_asserted;
+} Vmxnet3IntState;
+
+typedef struct {
+        PCIDevice parent_obj;
+        NICState *nic;
+        NICConf conf;
+        MemoryRegion bar0;
+        MemoryRegion bar1;
+        MemoryRegion msix_bar;
+
+        Vmxnet3RxqDescr rxq_descr[VMXNET3_DEVICE_MAX_RX_QUEUES];
+        Vmxnet3TxqDescr txq_descr[VMXNET3_DEVICE_MAX_TX_QUEUES];
+
+        /* Whether MSI-X support was installed successfully */
+        bool msix_used;
+        /* Whether MSI support was installed successfully */
+        bool msi_used;
+        hwaddr drv_shmem;
+        hwaddr temp_shared_guest_driver_memory;
+
+        uint8_t txq_num;
+
+        /* This boolean tells whether RX packet being indicated has to */
+        /* be split into head and body chunks from different RX rings  */
+        bool rx_packets_compound;
+
+        bool rx_vlan_stripping;
+        bool lro_supported;
+
+        uint8_t rxq_num;
+
+        /* Network MTU */
+        uint32_t mtu;
+
+        /* Maximum number of fragments for indicated TX packets */
+        uint32_t max_tx_frags;
+
+        /* Maximum number of fragments for indicated RX packets */
+        uint16_t max_rx_frags;
+
+        /* Index for events interrupt */
+        uint8_t event_int_idx;
+
+        /* Whether automatic interrupts masking enabled */
+        bool auto_int_masking;
+
+        bool peer_has_vhdr;
+
+        /* TX packets to QEMU interface */
+        struct VmxnetTxPkt *tx_pkt;
+        uint32_t offload_mode;
+        uint32_t cso_or_gso_size;
+        uint16_t tci;
+        bool needs_vlan;
+
+        struct VmxnetRxPkt *rx_pkt;
+
+        bool tx_sop;
+        bool skip_current_tx_pkt;
+
+        uint32_t device_active;
+        uint32_t last_command;
+
+        uint32_t link_status_and_speed;
+
+        Vmxnet3IntState interrupt_states[VMXNET3_MAX_INTRS];
+
+        uint32_t temp_mac;   /* To store the low part first */
+
+        MACAddr perm_mac;
+        uint32_t vlan_table[VMXNET3_VFT_SIZE];
+        uint32_t rx_mode;
+        MACAddr *mcast_list;
+        uint32_t mcast_list_len;
+        uint32_t mcast_list_buff_size; /* needed for live migration. */
+} VMXNET3State;
+
+/* Interrupt management */
+
+/*
+ *This function returns sign whether interrupt line is in asserted state
+ * This depends on the type of interrupt used. For INTX interrupt line will
+ * be asserted until explicit deassertion, for MSI(X) interrupt line will
+ * be deasserted automatically due to notification semantics of the MSI(X)
+ * interrupts
+ */
+static bool _vmxnet3_assert_interrupt_line(VMXNET3State *s, uint32_t int_idx)
+{
+    PCIDevice *d = PCI_DEVICE(s);
+
+    if (s->msix_used && msix_enabled(d)) {
+        VMW_IRPRN("Sending MSI-X notification for vector %u", int_idx);
+        msix_notify(d, int_idx);
+        return false;
+    }
+    if (s->msi_used && msi_enabled(d)) {
+        VMW_IRPRN("Sending MSI notification for vector %u", int_idx);
+        msi_notify(d, int_idx);
+        return false;
+    }
+
+    VMW_IRPRN("Asserting line for interrupt %u", int_idx);
+    qemu_set_irq(d->irq[int_idx], 1);
+    return true;
+}
+
+static void _vmxnet3_deassert_interrupt_line(VMXNET3State *s, int lidx)
+{
+    PCIDevice *d = PCI_DEVICE(s);
+
+    /*
+     * This function should never be called for MSI(X) interrupts
+     * because deassertion never required for message interrupts
+     */
+    assert(!s->msix_used || !msix_enabled(d));
+    /*
+     * This function should never be called for MSI(X) interrupts
+     * because deassertion never required for message interrupts
+     */
+    assert(!s->msi_used || !msi_enabled(d));
+
+    VMW_IRPRN("Deasserting line for interrupt %u", lidx);
+    qemu_set_irq(d->irq[lidx], 0);
+}
+
+static void vmxnet3_update_interrupt_line_state(VMXNET3State *s, int lidx)
+{
+    if (!s->interrupt_states[lidx].is_pending &&
+       s->interrupt_states[lidx].is_asserted) {
+        VMW_IRPRN("New interrupt line state for index %d is DOWN", lidx);
+        _vmxnet3_deassert_interrupt_line(s, lidx);
+        s->interrupt_states[lidx].is_asserted = false;
+        return;
+    }
+
+    if (s->interrupt_states[lidx].is_pending &&
+       !s->interrupt_states[lidx].is_masked &&
+       !s->interrupt_states[lidx].is_asserted) {
+        VMW_IRPRN("New interrupt line state for index %d is UP", lidx);
+        s->interrupt_states[lidx].is_asserted =
+            _vmxnet3_assert_interrupt_line(s, lidx);
+        s->interrupt_states[lidx].is_pending = false;
+        return;
+    }
+}
+
+static void vmxnet3_trigger_interrupt(VMXNET3State *s, int lidx)
+{
+    PCIDevice *d = PCI_DEVICE(s);
+    s->interrupt_states[lidx].is_pending = true;
+    vmxnet3_update_interrupt_line_state(s, lidx);
+
+    if (s->msix_used && msix_enabled(d) && s->auto_int_masking) {
+        goto do_automask;
+    }
+
+    if (s->msi_used && msi_enabled(d) && s->auto_int_masking) {
+        goto do_automask;
+    }
+
+    return;
+
+do_automask:
+    s->interrupt_states[lidx].is_masked = true;
+    vmxnet3_update_interrupt_line_state(s, lidx);
+}
+
+static bool vmxnet3_interrupt_asserted(VMXNET3State *s, int lidx)
+{
+    return s->interrupt_states[lidx].is_asserted;
+}
+
+static void vmxnet3_clear_interrupt(VMXNET3State *s, int int_idx)
+{
+    s->interrupt_states[int_idx].is_pending = false;
+    if (s->auto_int_masking) {
+        s->interrupt_states[int_idx].is_masked = true;
+    }
+    vmxnet3_update_interrupt_line_state(s, int_idx);
+}
+
+static void
+vmxnet3_on_interrupt_mask_changed(VMXNET3State *s, int lidx, bool is_masked)
+{
+    s->interrupt_states[lidx].is_masked = is_masked;
+    vmxnet3_update_interrupt_line_state(s, lidx);
+}
+
+static bool vmxnet3_verify_driver_magic(hwaddr dshmem)
+{
+    return (VMXNET3_READ_DRV_SHARED32(dshmem, magic) == VMXNET3_REV1_MAGIC);
+}
+
+#define VMXNET3_GET_BYTE(x, byte_num) (((x) >> (byte_num)*8) & 0xFF)
+#define VMXNET3_MAKE_BYTE(byte_num, val) \
+    (((uint32_t)((val) & 0xFF)) << (byte_num)*8)
+
+static void vmxnet3_set_variable_mac(VMXNET3State *s, uint32_t h, uint32_t l)
+{
+    s->conf.macaddr.a[0] = VMXNET3_GET_BYTE(l,  0);
+    s->conf.macaddr.a[1] = VMXNET3_GET_BYTE(l,  1);
+    s->conf.macaddr.a[2] = VMXNET3_GET_BYTE(l,  2);
+    s->conf.macaddr.a[3] = VMXNET3_GET_BYTE(l,  3);
+    s->conf.macaddr.a[4] = VMXNET3_GET_BYTE(h, 0);
+    s->conf.macaddr.a[5] = VMXNET3_GET_BYTE(h, 1);
+
+    VMW_CFPRN("Variable MAC: " VMXNET_MF, VMXNET_MA(s->conf.macaddr.a));
+
+    qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
+}
+
+static uint64_t vmxnet3_get_mac_low(MACAddr *addr)
+{
+    return VMXNET3_MAKE_BYTE(0, addr->a[0]) |
+           VMXNET3_MAKE_BYTE(1, addr->a[1]) |
+           VMXNET3_MAKE_BYTE(2, addr->a[2]) |
+           VMXNET3_MAKE_BYTE(3, addr->a[3]);
+}
+
+static uint64_t vmxnet3_get_mac_high(MACAddr *addr)
+{
+    return VMXNET3_MAKE_BYTE(0, addr->a[4]) |
+           VMXNET3_MAKE_BYTE(1, addr->a[5]);
+}
+
+static void
+vmxnet3_inc_tx_consumption_counter(VMXNET3State *s, int qidx)
+{
+    vmxnet3_ring_inc(&s->txq_descr[qidx].tx_ring);
+}
+
+static inline void
+vmxnet3_inc_rx_consumption_counter(VMXNET3State *s, int qidx, int ridx)
+{
+    vmxnet3_ring_inc(&s->rxq_descr[qidx].rx_ring[ridx]);
+}
+
+static inline void
+vmxnet3_inc_tx_completion_counter(VMXNET3State *s, int qidx)
+{
+    vmxnet3_ring_inc(&s->txq_descr[qidx].comp_ring);
+}
+
+static void
+vmxnet3_inc_rx_completion_counter(VMXNET3State *s, int qidx)
+{
+    vmxnet3_ring_inc(&s->rxq_descr[qidx].comp_ring);
+}
+
+static void
+vmxnet3_dec_rx_completion_counter(VMXNET3State *s, int qidx)
+{
+    vmxnet3_ring_dec(&s->rxq_descr[qidx].comp_ring);
+}
+
+static void vmxnet3_complete_packet(VMXNET3State *s, int qidx, uint32 tx_ridx)
+{
+    struct Vmxnet3_TxCompDesc txcq_descr;
+
+    VMXNET3_RING_DUMP(VMW_RIPRN, "TXC", qidx, &s->txq_descr[qidx].comp_ring);
+
+    txcq_descr.txdIdx = tx_ridx;
+    txcq_descr.gen = vmxnet3_ring_curr_gen(&s->txq_descr[qidx].comp_ring);
+
+    vmxnet3_ring_write_curr_cell(&s->txq_descr[qidx].comp_ring, &txcq_descr);
+
+    /* Flush changes in TX descriptor before changing the counter value */
+    smp_wmb();
+
+    vmxnet3_inc_tx_completion_counter(s, qidx);
+    vmxnet3_trigger_interrupt(s, s->txq_descr[qidx].intr_idx);
+}
+
+static bool
+vmxnet3_setup_tx_offloads(VMXNET3State *s)
+{
+    switch (s->offload_mode) {
+    case VMXNET3_OM_NONE:
+        vmxnet_tx_pkt_build_vheader(s->tx_pkt, false, false, 0);
+        break;
+
+    case VMXNET3_OM_CSUM:
+        vmxnet_tx_pkt_build_vheader(s->tx_pkt, false, true, 0);
+        VMW_PKPRN("L4 CSO requested\n");
+        break;
+
+    case VMXNET3_OM_TSO:
+        vmxnet_tx_pkt_build_vheader(s->tx_pkt, true, true,
+            s->cso_or_gso_size);
+        vmxnet_tx_pkt_update_ip_checksums(s->tx_pkt);
+        VMW_PKPRN("GSO offload requested.");
+        break;
+
+    default:
+        assert(false);
+        return false;
+    }
+
+    return true;
+}
+
+static void
+vmxnet3_tx_retrieve_metadata(VMXNET3State *s,
+                             const struct Vmxnet3_TxDesc *txd)
+{
+    s->offload_mode = txd->om;
+    s->cso_or_gso_size = txd->msscof;
+    s->tci = txd->tci;
+    s->needs_vlan = txd->ti;
+}
+
+typedef enum {
+    VMXNET3_PKT_STATUS_OK,
+    VMXNET3_PKT_STATUS_ERROR,
+    VMXNET3_PKT_STATUS_DISCARD,/* only for tx */
+    VMXNET3_PKT_STATUS_OUT_OF_BUF /* only for rx */
+} Vmxnet3PktStatus;
+
+static void
+vmxnet3_on_tx_done_update_stats(VMXNET3State *s, int qidx,
+    Vmxnet3PktStatus status)
+{
+    size_t tot_len = vmxnet_tx_pkt_get_total_len(s->tx_pkt);
+    struct UPT1_TxStats *stats = &s->txq_descr[qidx].txq_stats;
+
+    switch (status) {
+    case VMXNET3_PKT_STATUS_OK:
+        switch (vmxnet_tx_pkt_get_packet_type(s->tx_pkt)) {
+        case ETH_PKT_BCAST:
+            stats->bcastPktsTxOK++;
+            stats->bcastBytesTxOK += tot_len;
+            break;
+        case ETH_PKT_MCAST:
+            stats->mcastPktsTxOK++;
+            stats->mcastBytesTxOK += tot_len;
+            break;
+        case ETH_PKT_UCAST:
+            stats->ucastPktsTxOK++;
+            stats->ucastBytesTxOK += tot_len;
+            break;
+        default:
+            assert(false);
+        }
+
+        if (s->offload_mode == VMXNET3_OM_TSO) {
+            /*
+             * According to VMWARE headers this statistic is a number
+             * of packets after segmentation but since we don't have
+             * this information in QEMU model, the best we can do is to
+             * provide number of non-segmented packets
+             */
+            stats->TSOPktsTxOK++;
+            stats->TSOBytesTxOK += tot_len;
+        }
+        break;
+
+    case VMXNET3_PKT_STATUS_DISCARD:
+        stats->pktsTxDiscard++;
+        break;
+
+    case VMXNET3_PKT_STATUS_ERROR:
+        stats->pktsTxError++;
+        break;
+
+    default:
+        assert(false);
+    }
+}
+
+static void
+vmxnet3_on_rx_done_update_stats(VMXNET3State *s,
+                                int qidx,
+                                Vmxnet3PktStatus status)
+{
+    struct UPT1_RxStats *stats = &s->rxq_descr[qidx].rxq_stats;
+    size_t tot_len = vmxnet_rx_pkt_get_total_len(s->rx_pkt);
+
+    switch (status) {
+    case VMXNET3_PKT_STATUS_OUT_OF_BUF:
+        stats->pktsRxOutOfBuf++;
+        break;
+
+    case VMXNET3_PKT_STATUS_ERROR:
+        stats->pktsRxError++;
+        break;
+    case VMXNET3_PKT_STATUS_OK:
+        switch (vmxnet_rx_pkt_get_packet_type(s->rx_pkt)) {
+        case ETH_PKT_BCAST:
+            stats->bcastPktsRxOK++;
+            stats->bcastBytesRxOK += tot_len;
+            break;
+        case ETH_PKT_MCAST:
+            stats->mcastPktsRxOK++;
+            stats->mcastBytesRxOK += tot_len;
+            break;
+        case ETH_PKT_UCAST:
+            stats->ucastPktsRxOK++;
+            stats->ucastBytesRxOK += tot_len;
+            break;
+        default:
+            assert(false);
+        }
+
+        if (tot_len > s->mtu) {
+            stats->LROPktsRxOK++;
+            stats->LROBytesRxOK += tot_len;
+        }
+        break;
+    default:
+        assert(false);
+    }
+}
+
+static inline bool
+vmxnet3_pop_next_tx_descr(VMXNET3State *s,
+                          int qidx,
+                          struct Vmxnet3_TxDesc *txd,
+                          uint32_t *descr_idx)
+{
+    Vmxnet3Ring *ring = &s->txq_descr[qidx].tx_ring;
+
+    vmxnet3_ring_read_curr_cell(ring, txd);
+    if (txd->gen == vmxnet3_ring_curr_gen(ring)) {
+        /* Only read after generation field verification */
+        smp_rmb();
+        /* Re-read to be sure we got the latest version */
+        vmxnet3_ring_read_curr_cell(ring, txd);
+        VMXNET3_RING_DUMP(VMW_RIPRN, "TX", qidx, ring);
+        *descr_idx = vmxnet3_ring_curr_cell_idx(ring);
+        vmxnet3_inc_tx_consumption_counter(s, qidx);
+        return true;
+    }
+
+    return false;
+}
+
+static bool
+vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
+{
+    Vmxnet3PktStatus status = VMXNET3_PKT_STATUS_OK;
+
+    if (!vmxnet3_setup_tx_offloads(s)) {
+        status = VMXNET3_PKT_STATUS_ERROR;
+        goto func_exit;
+    }
+
+    /* debug prints */
+    vmxnet3_dump_virt_hdr(vmxnet_tx_pkt_get_vhdr(s->tx_pkt));
+    vmxnet_tx_pkt_dump(s->tx_pkt);
+
+    if (!vmxnet_tx_pkt_send(s->tx_pkt, qemu_get_queue(s->nic))) {
+        status = VMXNET3_PKT_STATUS_DISCARD;
+        goto func_exit;
+    }
+
+func_exit:
+    vmxnet3_on_tx_done_update_stats(s, qidx, status);
+    return (status == VMXNET3_PKT_STATUS_OK);
+}
+
+static void vmxnet3_process_tx_queue(VMXNET3State *s, int qidx)
+{
+    struct Vmxnet3_TxDesc txd;
+    uint32_t txd_idx;
+    uint32_t data_len;
+    hwaddr data_pa;
+
+    for (;;) {
+        if (!vmxnet3_pop_next_tx_descr(s, qidx, &txd, &txd_idx)) {
+            break;
+        }
+
+        vmxnet3_dump_tx_descr(&txd);
+
+        if (!s->skip_current_tx_pkt) {
+            data_len = (txd.len > 0) ? txd.len : VMXNET3_MAX_TX_BUF_SIZE;
+            data_pa = le64_to_cpu(txd.addr);
+
+            if (!vmxnet_tx_pkt_add_raw_fragment(s->tx_pkt,
+                                                data_pa,
+                                                data_len)) {
+                s->skip_current_tx_pkt = true;
+            }
+        }
+
+        if (s->tx_sop) {
+            vmxnet3_tx_retrieve_metadata(s, &txd);
+            s->tx_sop = false;
+        }
+
+        if (txd.eop) {
+            if (!s->skip_current_tx_pkt) {
+                vmxnet_tx_pkt_parse(s->tx_pkt);
+
+                if (s->needs_vlan) {
+                    vmxnet_tx_pkt_setup_vlan_header(s->tx_pkt, s->tci);
+                }
+
+                vmxnet3_send_packet(s, qidx);
+            } else {
+                vmxnet3_on_tx_done_update_stats(s, qidx,
+                                                VMXNET3_PKT_STATUS_ERROR);
+            }
+
+            vmxnet3_complete_packet(s, qidx, txd_idx);
+            s->tx_sop = true;
+            s->skip_current_tx_pkt = false;
+            vmxnet_tx_pkt_reset(s->tx_pkt);
+        }
+    }
+}
+
+static inline void
+vmxnet3_read_next_rx_descr(VMXNET3State *s, int qidx, int ridx,
+                           struct Vmxnet3_RxDesc *dbuf, uint32_t *didx)
+{
+    Vmxnet3Ring *ring = &s->rxq_descr[qidx].rx_ring[ridx];
+    *didx = vmxnet3_ring_curr_cell_idx(ring);
+    vmxnet3_ring_read_curr_cell(ring, dbuf);
+}
+
+static inline uint8_t
+vmxnet3_get_rx_ring_gen(VMXNET3State *s, int qidx, int ridx)
+{
+    return s->rxq_descr[qidx].rx_ring[ridx].gen;
+}
+
+static inline hwaddr
+vmxnet3_pop_rxc_descr(VMXNET3State *s, int qidx, uint32_t *descr_gen)
+{
+    uint8_t ring_gen;
+    struct Vmxnet3_RxCompDesc rxcd;
+
+    hwaddr daddr =
+        vmxnet3_ring_curr_cell_pa(&s->rxq_descr[qidx].comp_ring);
+
+    cpu_physical_memory_read(daddr, &rxcd, sizeof(struct Vmxnet3_RxCompDesc));
+    ring_gen = vmxnet3_ring_curr_gen(&s->rxq_descr[qidx].comp_ring);
+
+    if (rxcd.gen != ring_gen) {
+        *descr_gen = ring_gen;
+        vmxnet3_inc_rx_completion_counter(s, qidx);
+        return daddr;
+    }
+
+    return 0;
+}
+
+static inline void
+vmxnet3_revert_rxc_descr(VMXNET3State *s, int qidx)
+{
+    vmxnet3_dec_rx_completion_counter(s, qidx);
+}
+
+#define RXQ_IDX      (0)
+#define RX_HEAD_BODY_RING (0)
+#define RX_BODY_ONLY_RING (1)
+
+static bool
+vmxnet3_get_next_head_rx_descr(VMXNET3State *s,
+                               struct Vmxnet3_RxDesc *descr_buf,
+                               uint32_t *descr_idx,
+                               uint32_t *ridx)
+{
+    for (;;) {
+        uint32_t ring_gen;
+        vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
+                                   descr_buf, descr_idx);
+
+        /* If no more free descriptors - return */
+        ring_gen = vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING);
+        if (descr_buf->gen != ring_gen) {
+            return false;
+        }
+
+        /* Only read after generation field verification */
+        smp_rmb();
+        /* Re-read to be sure we got the latest version */
+        vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING,
+                                   descr_buf, descr_idx);
+
+        /* Mark current descriptor as used/skipped */
+        vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
+
+        /* If this is what we are looking for - return */
+        if (descr_buf->btype == VMXNET3_RXD_BTYPE_HEAD) {
+            *ridx = RX_HEAD_BODY_RING;
+            return true;
+        }
+    }
+}
+
+static bool
+vmxnet3_get_next_body_rx_descr(VMXNET3State *s,
+                               struct Vmxnet3_RxDesc *d,
+                               uint32_t *didx,
+                               uint32_t *ridx)
+{
+    vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
+
+    /* Try to find corresponding descriptor in head/body ring */
+    if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_HEAD_BODY_RING)) {
+        /* Only read after generation field verification */
+        smp_rmb();
+        /* Re-read to be sure we got the latest version */
+        vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_HEAD_BODY_RING, d, didx);
+        if (d->btype == VMXNET3_RXD_BTYPE_BODY) {
+            vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_HEAD_BODY_RING);
+            *ridx = RX_HEAD_BODY_RING;
+            return true;
+        }
+    }
+
+    /*
+     * If there is no free descriptors on head/body ring or next free
+     * descriptor is a head descriptor switch to body only ring
+     */
+    vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
+
+    /* If no more free descriptors - return */
+    if (d->gen == vmxnet3_get_rx_ring_gen(s, RXQ_IDX, RX_BODY_ONLY_RING)) {
+        /* Only read after generation field verification */
+        smp_rmb();
+        /* Re-read to be sure we got the latest version */
+        vmxnet3_read_next_rx_descr(s, RXQ_IDX, RX_BODY_ONLY_RING, d, didx);
+        assert(d->btype == VMXNET3_RXD_BTYPE_BODY);
+        *ridx = RX_BODY_ONLY_RING;
+        vmxnet3_inc_rx_consumption_counter(s, RXQ_IDX, RX_BODY_ONLY_RING);
+        return true;
+    }
+
+    return false;
+}
+
+static inline bool
+vmxnet3_get_next_rx_descr(VMXNET3State *s, bool is_head,
+                          struct Vmxnet3_RxDesc *descr_buf,
+                          uint32_t *descr_idx,
+                          uint32_t *ridx)
+{
+    if (is_head || !s->rx_packets_compound) {
+        return vmxnet3_get_next_head_rx_descr(s, descr_buf, descr_idx, ridx);
+    } else {
+        return vmxnet3_get_next_body_rx_descr(s, descr_buf, descr_idx, ridx);
+    }
+}
+
+static void vmxnet3_rx_update_descr(struct VmxnetRxPkt *pkt,
+    struct Vmxnet3_RxCompDesc *rxcd)
+{
+    int csum_ok, is_gso;
+    bool isip4, isip6, istcp, isudp;
+    struct virtio_net_hdr *vhdr;
+    uint8_t offload_type;
+
+    if (vmxnet_rx_pkt_is_vlan_stripped(pkt)) {
+        rxcd->ts = 1;
+        rxcd->tci = vmxnet_rx_pkt_get_vlan_tag(pkt);
+    }
+
+    if (!vmxnet_rx_pkt_has_virt_hdr(pkt)) {
+        goto nocsum;
+    }
+
+    vhdr = vmxnet_rx_pkt_get_vhdr(pkt);
+    /*
+     * Checksum is valid when lower level tell so or when lower level
+     * requires checksum offload telling that packet produced/bridged
+     * locally and did travel over network after last checksum calculation
+     * or production
+     */
+    csum_ok = VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_DATA_VALID) ||
+              VMXNET_FLAG_IS_SET(vhdr->flags, VIRTIO_NET_HDR_F_NEEDS_CSUM);
+
+    offload_type = vhdr->gso_type & ~VIRTIO_NET_HDR_GSO_ECN;
+    is_gso = (offload_type != VIRTIO_NET_HDR_GSO_NONE) ? 1 : 0;
+
+    if (!csum_ok && !is_gso) {
+        goto nocsum;
+    }
+
+    vmxnet_rx_pkt_get_protocols(pkt, &isip4, &isip6, &isudp, &istcp);
+    if ((!istcp && !isudp) || (!isip4 && !isip6)) {
+        goto nocsum;
+    }
+
+    rxcd->cnc = 0;
+    rxcd->v4 = isip4 ? 1 : 0;
+    rxcd->v6 = isip6 ? 1 : 0;
+    rxcd->tcp = istcp ? 1 : 0;
+    rxcd->udp = isudp ? 1 : 0;
+    rxcd->fcs = rxcd->tuc = rxcd->ipc = 1;
+    return;
+
+nocsum:
+    rxcd->cnc = 1;
+    return;
+}
+
+static void
+vmxnet3_physical_memory_writev(const struct iovec *iov,
+                               size_t start_iov_off,
+                               hwaddr target_addr,
+                               size_t bytes_to_copy)
+{
+    size_t curr_off = 0;
+    size_t copied = 0;
+
+    while (bytes_to_copy) {
+        if (start_iov_off < (curr_off + iov->iov_len)) {
+            size_t chunk_len =
+                MIN((curr_off + iov->iov_len) - start_iov_off, bytes_to_copy);
+
+            cpu_physical_memory_write(target_addr + copied,
+                                      iov->iov_base + start_iov_off - curr_off,
+                                      chunk_len);
+
+            copied += chunk_len;
+            start_iov_off += chunk_len;
+            curr_off = start_iov_off;
+            bytes_to_copy -= chunk_len;
+        } else {
+            curr_off += iov->iov_len;
+        }
+        iov++;
+    }
+}
+
+static bool
+vmxnet3_indicate_packet(VMXNET3State *s)
+{
+    struct Vmxnet3_RxDesc rxd;
+    bool is_head = true;
+    uint32_t rxd_idx;
+    uint32_t rx_ridx;
+
+    struct Vmxnet3_RxCompDesc rxcd;
+    uint32_t new_rxcd_gen = VMXNET3_INIT_GEN;
+    hwaddr new_rxcd_pa = 0;
+    hwaddr ready_rxcd_pa = 0;
+    struct iovec *data = vmxnet_rx_pkt_get_iovec(s->rx_pkt);
+    size_t bytes_copied = 0;
+    size_t bytes_left = vmxnet_rx_pkt_get_total_len(s->rx_pkt);
+    uint16_t num_frags = 0;
+    size_t chunk_size;
+
+    vmxnet_rx_pkt_dump(s->rx_pkt);
+
+    while (bytes_left > 0) {
+
+        /* cannot add more frags to packet */
+        if (num_frags == s->max_rx_frags) {
+            break;
+        }
+
+        new_rxcd_pa = vmxnet3_pop_rxc_descr(s, RXQ_IDX, &new_rxcd_gen);
+        if (!new_rxcd_pa) {
+            break;
+        }
+
+        if (!vmxnet3_get_next_rx_descr(s, is_head, &rxd, &rxd_idx, &rx_ridx)) {
+            break;
+        }
+
+        chunk_size = MIN(bytes_left, rxd.len);
+        vmxnet3_physical_memory_writev(data, bytes_copied,
+                                       le64_to_cpu(rxd.addr), chunk_size);
+        bytes_copied += chunk_size;
+        bytes_left -= chunk_size;
+
+        vmxnet3_dump_rx_descr(&rxd);
+
+        if (0 != ready_rxcd_pa) {
+            cpu_physical_memory_write(ready_rxcd_pa, &rxcd, sizeof(rxcd));
+        }
+
+        memset(&rxcd, 0, sizeof(struct Vmxnet3_RxCompDesc));
+        rxcd.rxdIdx = rxd_idx;
+        rxcd.len = chunk_size;
+        rxcd.sop = is_head;
+        rxcd.gen = new_rxcd_gen;
+        rxcd.rqID = RXQ_IDX + rx_ridx * s->rxq_num;
+
+        if (0 == bytes_left) {
+            vmxnet3_rx_update_descr(s->rx_pkt, &rxcd);
+        }
+
+        VMW_RIPRN("RX Completion descriptor: rxRing: %lu rxIdx %lu len %lu "
+                  "sop %d csum_correct %lu",
+                  (unsigned long) rx_ridx,
+                  (unsigned long) rxcd.rxdIdx,
+                  (unsigned long) rxcd.len,
+                  (int) rxcd.sop,
+                  (unsigned long) rxcd.tuc);
+
+        is_head = false;
+        ready_rxcd_pa = new_rxcd_pa;
+        new_rxcd_pa = 0;
+    }
+
+    if (0 != ready_rxcd_pa) {
+        rxcd.eop = 1;
+        rxcd.err = (0 != bytes_left);
+        cpu_physical_memory_write(ready_rxcd_pa, &rxcd, sizeof(rxcd));
+
+        /* Flush RX descriptor changes */
+        smp_wmb();
+    }
+
+    if (0 != new_rxcd_pa) {
+        vmxnet3_revert_rxc_descr(s, RXQ_IDX);
+    }
+
+    vmxnet3_trigger_interrupt(s, s->rxq_descr[RXQ_IDX].intr_idx);
+
+    if (bytes_left == 0) {
+        vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_OK);
+        return true;
+    } else if (num_frags == s->max_rx_frags) {
+        vmxnet3_on_rx_done_update_stats(s, RXQ_IDX, VMXNET3_PKT_STATUS_ERROR);
+        return false;
+    } else {
+        vmxnet3_on_rx_done_update_stats(s, RXQ_IDX,
+                                        VMXNET3_PKT_STATUS_OUT_OF_BUF);
+        return false;
+    }
+}
+
+static void
+vmxnet3_io_bar0_write(void *opaque, hwaddr addr,
+                      uint64_t val, unsigned size)
+{
+    VMXNET3State *s = opaque;
+
+    if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_TXPROD,
+                        VMXNET3_DEVICE_MAX_TX_QUEUES, VMXNET3_REG_ALIGN)) {
+        int tx_queue_idx =
+            VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_TXPROD,
+                                     VMXNET3_REG_ALIGN);
+        assert(tx_queue_idx <= s->txq_num);
+        vmxnet3_process_tx_queue(s, tx_queue_idx);
+        return;
+    }
+
+    if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
+                        VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
+        int l = VMW_MULTIREG_IDX_BY_ADDR(addr, VMXNET3_REG_IMR,
+                                         VMXNET3_REG_ALIGN);
+
+        VMW_CBPRN("Interrupt mask for line %d written: 0x%" PRIx64, l, val);
+
+        vmxnet3_on_interrupt_mask_changed(s, l, val);
+        return;
+    }
+
+    if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD,
+                        VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN) ||
+       VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_RXPROD2,
+                        VMXNET3_DEVICE_MAX_RX_QUEUES, VMXNET3_REG_ALIGN)) {
+        return;
+    }
+
+    VMW_WRPRN("BAR0 unknown write [%" PRIx64 "] = %" PRIx64 ", size %d",
+              (uint64_t) addr, val, size);
+}
+
+static uint64_t
+vmxnet3_io_bar0_read(void *opaque, hwaddr addr, unsigned size)
+{
+    if (VMW_IS_MULTIREG_ADDR(addr, VMXNET3_REG_IMR,
+                        VMXNET3_MAX_INTRS, VMXNET3_REG_ALIGN)) {
+        assert(false);
+    }
+
+    VMW_CBPRN("BAR0 unknown read [%" PRIx64 "], size %d", addr, size);
+    return 0;
+}
+
+static void vmxnet3_reset_interrupt_states(VMXNET3State *s)
+{
+    int i;
+    for (i = 0; i < ARRAY_SIZE(s->interrupt_states); i++) {
+        s->interrupt_states[i].is_asserted = false;
+        s->interrupt_states[i].is_pending = false;
+        s->interrupt_states[i].is_masked = true;
+    }
+}
+
+static void vmxnet3_reset_mac(VMXNET3State *s)
+{
+    memcpy(&s->conf.macaddr.a, &s->perm_mac.a, sizeof(s->perm_mac.a));
+    VMW_CFPRN("MAC address set to: " VMXNET_MF, VMXNET_MA(s->conf.macaddr.a));
+}
+
+static void vmxnet3_deactivate_device(VMXNET3State *s)
+{
+    VMW_CBPRN("Deactivating vmxnet3...");
+    s->device_active = false;
+}
+
+static void vmxnet3_reset(VMXNET3State *s)
+{
+    VMW_CBPRN("Resetting vmxnet3...");
+
+    vmxnet3_deactivate_device(s);
+    vmxnet3_reset_interrupt_states(s);
+    vmxnet_tx_pkt_reset(s->tx_pkt);
+    s->drv_shmem = 0;
+    s->tx_sop = true;
+    s->skip_current_tx_pkt = false;
+}
+
+static void vmxnet3_update_rx_mode(VMXNET3State *s)
+{
+    s->rx_mode = VMXNET3_READ_DRV_SHARED32(s->drv_shmem,
+                                           devRead.rxFilterConf.rxMode);
+    VMW_CFPRN("RX mode: 0x%08X", s->rx_mode);
+}
+
+static void vmxnet3_update_vlan_filters(VMXNET3State *s)
+{
+    int i;
+
+    /* Copy configuration from shared memory */
+    VMXNET3_READ_DRV_SHARED(s->drv_shmem,
+                            devRead.rxFilterConf.vfTable,
+                            s->vlan_table,
+                            sizeof(s->vlan_table));
+
+    /* Invert byte order when needed */
+    for (i = 0; i < ARRAY_SIZE(s->vlan_table); i++) {
+        s->vlan_table[i] = le32_to_cpu(s->vlan_table[i]);
+    }
+
+    /* Dump configuration for debugging purposes */
+    VMW_CFPRN("Configured VLANs:");
+    for (i = 0; i < sizeof(s->vlan_table) * 8; i++) {
+        if (VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, i)) {
+            VMW_CFPRN("\tVLAN %d is present", i);
+        }
+    }
+}
+
+static void vmxnet3_update_mcast_filters(VMXNET3State *s)
+{
+    uint16_t list_bytes =
+        VMXNET3_READ_DRV_SHARED16(s->drv_shmem,
+                                  devRead.rxFilterConf.mfTableLen);
+
+    s->mcast_list_len = list_bytes / sizeof(s->mcast_list[0]);
+
+    s->mcast_list = g_realloc(s->mcast_list, list_bytes);
+    if (NULL == s->mcast_list) {
+        if (0 == s->mcast_list_len) {
+            VMW_CFPRN("Current multicast list is empty");
+        } else {
+            VMW_ERPRN("Failed to allocate multicast list of %d elements",
+                      s->mcast_list_len);
+        }
+        s->mcast_list_len = 0;
+    } else {
+        int i;
+        hwaddr mcast_list_pa =
+            VMXNET3_READ_DRV_SHARED64(s->drv_shmem,
+                                      devRead.rxFilterConf.mfTablePA);
+
+        cpu_physical_memory_read(mcast_list_pa, s->mcast_list, list_bytes);
+        VMW_CFPRN("Current multicast list len is %d:", s->mcast_list_len);
+        for (i = 0; i < s->mcast_list_len; i++) {
+            VMW_CFPRN("\t" VMXNET_MF, VMXNET_MA(s->mcast_list[i].a));
+        }
+    }
+}
+
+static void vmxnet3_setup_rx_filtering(VMXNET3State *s)
+{
+    vmxnet3_update_rx_mode(s);
+    vmxnet3_update_vlan_filters(s);
+    vmxnet3_update_mcast_filters(s);
+}
+
+static uint32_t vmxnet3_get_interrupt_config(VMXNET3State *s)
+{
+    uint32_t interrupt_mode = VMXNET3_IT_AUTO | (VMXNET3_IMM_AUTO << 2);
+    VMW_CFPRN("Interrupt config is 0x%X", interrupt_mode);
+    return interrupt_mode;
+}
+
+static void vmxnet3_fill_stats(VMXNET3State *s)
+{
+    int i;
+    for (i = 0; i < s->txq_num; i++) {
+        cpu_physical_memory_write(s->txq_descr[i].tx_stats_pa,
+                                  &s->txq_descr[i].txq_stats,
+                                  sizeof(s->txq_descr[i].txq_stats));
+    }
+
+    for (i = 0; i < s->rxq_num; i++) {
+        cpu_physical_memory_write(s->rxq_descr[i].rx_stats_pa,
+                                  &s->rxq_descr[i].rxq_stats,
+                                  sizeof(s->rxq_descr[i].rxq_stats));
+    }
+}
+
+static void vmxnet3_adjust_by_guest_type(VMXNET3State *s)
+{
+    struct Vmxnet3_GOSInfo gos;
+
+    VMXNET3_READ_DRV_SHARED(s->drv_shmem, devRead.misc.driverInfo.gos,
+                            &gos, sizeof(gos));
+    s->rx_packets_compound =
+        (gos.gosType == VMXNET3_GOS_TYPE_WIN) ? false : true;
+
+    VMW_CFPRN("Guest type specifics: RXCOMPOUND: %d", s->rx_packets_compound);
+}
+
+static void
+vmxnet3_dump_conf_descr(const char *name,
+                        struct Vmxnet3_VariableLenConfDesc *pm_descr)
+{
+    VMW_CFPRN("%s descriptor dump: Version %u, Length %u",
+              name, pm_descr->confVer, pm_descr->confLen);
+
+};
+
+static void vmxnet3_update_pm_state(VMXNET3State *s)
+{
+    struct Vmxnet3_VariableLenConfDesc pm_descr;
+
+    pm_descr.confLen =
+        VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.pmConfDesc.confLen);
+    pm_descr.confVer =
+        VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.pmConfDesc.confVer);
+    pm_descr.confPA =
+        VMXNET3_READ_DRV_SHARED64(s->drv_shmem, devRead.pmConfDesc.confPA);
+
+    vmxnet3_dump_conf_descr("PM State", &pm_descr);
+}
+
+static void vmxnet3_update_features(VMXNET3State *s)
+{
+    uint32_t guest_features;
+    int rxcso_supported;
+
+    guest_features = VMXNET3_READ_DRV_SHARED32(s->drv_shmem,
+                                               devRead.misc.uptFeatures);
+
+    rxcso_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXCSUM);
+    s->rx_vlan_stripping = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_RXVLAN);
+    s->lro_supported = VMXNET_FLAG_IS_SET(guest_features, UPT1_F_LRO);
+
+    VMW_CFPRN("Features configuration: LRO: %d, RXCSUM: %d, VLANSTRIP: %d",
+              s->lro_supported, rxcso_supported,
+              s->rx_vlan_stripping);
+    if (s->peer_has_vhdr) {
+        tap_set_offload(qemu_get_queue(s->nic)->peer,
+                        rxcso_supported,
+                        s->lro_supported,
+                        s->lro_supported,
+                        0,
+                        0);
+    }
+}
+
+static void vmxnet3_activate_device(VMXNET3State *s)
+{
+    int i;
+    static const uint32_t VMXNET3_DEF_TX_THRESHOLD = 1;
+    hwaddr qdescr_table_pa;
+    uint64_t pa;
+    uint32_t size;
+
+    /* Verify configuration consistency */
+    if (!vmxnet3_verify_driver_magic(s->drv_shmem)) {
+        VMW_ERPRN("Device configuration received from driver is invalid");
+        return;
+    }
+
+    vmxnet3_adjust_by_guest_type(s);
+    vmxnet3_update_features(s);
+    vmxnet3_update_pm_state(s);
+    vmxnet3_setup_rx_filtering(s);
+    /* Cache fields from shared memory */
+    s->mtu = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, devRead.misc.mtu);
+    VMW_CFPRN("MTU is %u", s->mtu);
+
+    s->max_rx_frags =
+        VMXNET3_READ_DRV_SHARED16(s->drv_shmem, devRead.misc.maxNumRxSG);
+
+    VMW_CFPRN("Max RX fragments is %u", s->max_rx_frags);
+
+    s->event_int_idx =
+        VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.intrConf.eventIntrIdx);
+    VMW_CFPRN("Events interrupt line is %u", s->event_int_idx);
+
+    s->auto_int_masking =
+        VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.intrConf.autoMask);
+    VMW_CFPRN("Automatic interrupt masking is %d", (int)s->auto_int_masking);
+
+    s->txq_num =
+        VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.misc.numTxQueues);
+    s->rxq_num =
+        VMXNET3_READ_DRV_SHARED8(s->drv_shmem, devRead.misc.numRxQueues);
+
+    VMW_CFPRN("Number of TX/RX queues %u/%u", s->txq_num, s->rxq_num);
+    assert(s->txq_num <= VMXNET3_DEVICE_MAX_TX_QUEUES);
+
+    qdescr_table_pa =
+        VMXNET3_READ_DRV_SHARED64(s->drv_shmem, devRead.misc.queueDescPA);
+    VMW_CFPRN("TX queues descriptors table is at 0x%" PRIx64, qdescr_table_pa);
+
+    /*
+     * Worst-case scenario is a packet that holds all TX rings space so
+     * we calculate total size of all TX rings for max TX fragments number
+     */
+    s->max_tx_frags = 0;
+
+    /* TX queues */
+    for (i = 0; i < s->txq_num; i++) {
+        hwaddr qdescr_pa =
+            qdescr_table_pa + i * sizeof(struct Vmxnet3_TxQueueDesc);
+
+        /* Read interrupt number for this TX queue */
+        s->txq_descr[i].intr_idx =
+            VMXNET3_READ_TX_QUEUE_DESCR8(qdescr_pa, conf.intrIdx);
+
+        VMW_CFPRN("TX Queue %d interrupt: %d", i, s->txq_descr[i].intr_idx);
+
+        /* Read rings memory locations for TX queues */
+        pa = VMXNET3_READ_TX_QUEUE_DESCR64(qdescr_pa, conf.txRingBasePA);
+        size = VMXNET3_READ_TX_QUEUE_DESCR32(qdescr_pa, conf.txRingSize);
+
+        vmxnet3_ring_init(&s->txq_descr[i].tx_ring, pa, size,
+                          sizeof(struct Vmxnet3_TxDesc), false);
+        VMXNET3_RING_DUMP(VMW_CFPRN, "TX", i, &s->txq_descr[i].tx_ring);
+
+        s->max_tx_frags += size;
+
+        /* TXC ring */
+        pa = VMXNET3_READ_TX_QUEUE_DESCR64(qdescr_pa, conf.compRingBasePA);
+        size = VMXNET3_READ_TX_QUEUE_DESCR32(qdescr_pa, conf.compRingSize);
+        vmxnet3_ring_init(&s->txq_descr[i].comp_ring, pa, size,
+                          sizeof(struct Vmxnet3_TxCompDesc), true);
+        VMXNET3_RING_DUMP(VMW_CFPRN, "TXC", i, &s->txq_descr[i].comp_ring);
+
+        s->txq_descr[i].tx_stats_pa =
+            qdescr_pa + offsetof(struct Vmxnet3_TxQueueDesc, stats);
+
+        memset(&s->txq_descr[i].txq_stats, 0,
+               sizeof(s->txq_descr[i].txq_stats));
+
+        /* Fill device-managed parameters for queues */
+        VMXNET3_WRITE_TX_QUEUE_DESCR32(qdescr_pa,
+                                       ctrl.txThreshold,
+                                       VMXNET3_DEF_TX_THRESHOLD);
+    }
+
+    /* Preallocate TX packet wrapper */
+    VMW_CFPRN("Max TX fragments is %u", s->max_tx_frags);
+    vmxnet_tx_pkt_init(&s->tx_pkt, s->max_tx_frags, s->peer_has_vhdr);
+    vmxnet_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
+
+    /* Read rings memory locations for RX queues */
+    for (i = 0; i < s->rxq_num; i++) {
+        int j;
+        hwaddr qd_pa =
+            qdescr_table_pa + s->txq_num * sizeof(struct Vmxnet3_TxQueueDesc) +
+            i * sizeof(struct Vmxnet3_RxQueueDesc);
+
+        /* Read interrupt number for this RX queue */
+        s->rxq_descr[i].intr_idx =
+            VMXNET3_READ_TX_QUEUE_DESCR8(qd_pa, conf.intrIdx);
+
+        VMW_CFPRN("RX Queue %d interrupt: %d", i, s->rxq_descr[i].intr_idx);
+
+        /* Read rings memory locations */
+        for (j = 0; j < VMXNET3_RX_RINGS_PER_QUEUE; j++) {
+            /* RX rings */
+            pa = VMXNET3_READ_RX_QUEUE_DESCR64(qd_pa, conf.rxRingBasePA[j]);
+            size = VMXNET3_READ_RX_QUEUE_DESCR32(qd_pa, conf.rxRingSize[j]);
+            vmxnet3_ring_init(&s->rxq_descr[i].rx_ring[j], pa, size,
+                              sizeof(struct Vmxnet3_RxDesc), false);
+            VMW_CFPRN("RX queue %d:%d: Base: %" PRIx64 ", Size: %d",
+                      i, j, pa, size);
+        }
+
+        /* RXC ring */
+        pa = VMXNET3_READ_RX_QUEUE_DESCR64(qd_pa, conf.compRingBasePA);
+        size = VMXNET3_READ_RX_QUEUE_DESCR32(qd_pa, conf.compRingSize);
+        vmxnet3_ring_init(&s->rxq_descr[i].comp_ring, pa, size,
+                          sizeof(struct Vmxnet3_RxCompDesc), true);
+        VMW_CFPRN("RXC queue %d: Base: %" PRIx64 ", Size: %d", i, pa, size);
+
+        s->rxq_descr[i].rx_stats_pa =
+            qd_pa + offsetof(struct Vmxnet3_RxQueueDesc, stats);
+        memset(&s->rxq_descr[i].rxq_stats, 0,
+               sizeof(s->rxq_descr[i].rxq_stats));
+    }
+
+    /* Make sure everything is in place before device activation */
+    smp_wmb();
+
+    vmxnet3_reset_mac(s);
+
+    s->device_active = true;
+}
+
+static void vmxnet3_handle_command(VMXNET3State *s, uint64_t cmd)
+{
+    s->last_command = cmd;
+
+    switch (cmd) {
+    case VMXNET3_CMD_GET_PERM_MAC_HI:
+        VMW_CBPRN("Set: Get upper part of permanent MAC");
+        break;
+
+    case VMXNET3_CMD_GET_PERM_MAC_LO:
+        VMW_CBPRN("Set: Get lower part of permanent MAC");
+        break;
+
+    case VMXNET3_CMD_GET_STATS:
+        VMW_CBPRN("Set: Get device statistics");
+        vmxnet3_fill_stats(s);
+        break;
+
+    case VMXNET3_CMD_ACTIVATE_DEV:
+        VMW_CBPRN("Set: Activating vmxnet3 device");
+        vmxnet3_activate_device(s);
+        break;
+
+    case VMXNET3_CMD_UPDATE_RX_MODE:
+        VMW_CBPRN("Set: Update rx mode");
+        vmxnet3_update_rx_mode(s);
+        break;
+
+    case VMXNET3_CMD_UPDATE_VLAN_FILTERS:
+        VMW_CBPRN("Set: Update VLAN filters");
+        vmxnet3_update_vlan_filters(s);
+        break;
+
+    case VMXNET3_CMD_UPDATE_MAC_FILTERS:
+        VMW_CBPRN("Set: Update MAC filters");
+        vmxnet3_update_mcast_filters(s);
+        break;
+
+    case VMXNET3_CMD_UPDATE_FEATURE:
+        VMW_CBPRN("Set: Update features");
+        vmxnet3_update_features(s);
+        break;
+
+    case VMXNET3_CMD_UPDATE_PMCFG:
+        VMW_CBPRN("Set: Update power management config");
+        vmxnet3_update_pm_state(s);
+        break;
+
+    case VMXNET3_CMD_GET_LINK:
+        VMW_CBPRN("Set: Get link");
+        break;
+
+    case VMXNET3_CMD_RESET_DEV:
+        VMW_CBPRN("Set: Reset device");
+        vmxnet3_reset(s);
+        break;
+
+    case VMXNET3_CMD_QUIESCE_DEV:
+        VMW_CBPRN("Set: VMXNET3_CMD_QUIESCE_DEV - pause the device");
+        vmxnet3_deactivate_device(s);
+        break;
+
+    case VMXNET3_CMD_GET_CONF_INTR:
+        VMW_CBPRN("Set: VMXNET3_CMD_GET_CONF_INTR - interrupt configuration");
+        break;
+
+    default:
+        VMW_CBPRN("Received unknown command: %" PRIx64, cmd);
+        break;
+    }
+}
+
+static uint64_t vmxnet3_get_command_status(VMXNET3State *s)
+{
+    uint64_t ret;
+
+    switch (s->last_command) {
+    case VMXNET3_CMD_ACTIVATE_DEV:
+        ret = (s->device_active) ? 0 : -1;
+        VMW_CFPRN("Device active: %" PRIx64, ret);
+        break;
+
+    case VMXNET3_CMD_GET_LINK:
+        ret = s->link_status_and_speed;
+        VMW_CFPRN("Link and speed: %" PRIx64, ret);
+        break;
+
+    case VMXNET3_CMD_GET_PERM_MAC_LO:
+        ret = vmxnet3_get_mac_low(&s->perm_mac);
+        break;
+
+    case VMXNET3_CMD_GET_PERM_MAC_HI:
+        ret = vmxnet3_get_mac_high(&s->perm_mac);
+        break;
+
+    case VMXNET3_CMD_GET_CONF_INTR:
+        ret = vmxnet3_get_interrupt_config(s);
+        break;
+
+    default:
+        VMW_WRPRN("Received request for unknown command: %x", s->last_command);
+        ret = -1;
+        break;
+    }
+
+    return ret;
+}
+
+static void vmxnet3_set_events(VMXNET3State *s, uint32_t val)
+{
+    uint32_t events;
+
+    VMW_CBPRN("Setting events: 0x%x", val);
+    events = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, ecr) | val;
+    VMXNET3_WRITE_DRV_SHARED32(s->drv_shmem, ecr, events);
+}
+
+static void vmxnet3_ack_events(VMXNET3State *s, uint32_t val)
+{
+    uint32_t events;
+
+    VMW_CBPRN("Clearing events: 0x%x", val);
+    events = VMXNET3_READ_DRV_SHARED32(s->drv_shmem, ecr) & ~val;
+    VMXNET3_WRITE_DRV_SHARED32(s->drv_shmem, ecr, events);
+}
+
+static void
+vmxnet3_io_bar1_write(void *opaque,
+                      hwaddr addr,
+                      uint64_t val,
+                      unsigned size)
+{
+    VMXNET3State *s = opaque;
+
+    switch (addr) {
+    /* Vmxnet3 Revision Report Selection */
+    case VMXNET3_REG_VRRS:
+        VMW_CBPRN("Write BAR1 [VMXNET3_REG_VRRS] = %" PRIx64 ", size %d",
+                  val, size);
+        break;
+
+    /* UPT Version Report Selection */
+    case VMXNET3_REG_UVRS:
+        VMW_CBPRN("Write BAR1 [VMXNET3_REG_UVRS] = %" PRIx64 ", size %d",
+                  val, size);
+        break;
+
+    /* Driver Shared Address Low */
+    case VMXNET3_REG_DSAL:
+        VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAL] = %" PRIx64 ", size %d",
+                  val, size);
+        /*
+         * Guest driver will first write the low part of the shared
+         * memory address. We save it to temp variable and set the
+         * shared address only after we get the high part
+         */
+        if (0 == val) {
+            s->device_active = false;
+        }
+        s->temp_shared_guest_driver_memory = val;
+        s->drv_shmem = 0;
+        break;
+
+    /* Driver Shared Address High */
+    case VMXNET3_REG_DSAH:
+        VMW_CBPRN("Write BAR1 [VMXNET3_REG_DSAH] = %" PRIx64 ", size %d",
+                  val, size);
+        /*
+         * Set the shared memory between guest driver and device.
+         * We already should have low address part.
+         */
+        s->drv_shmem = s->temp_shared_guest_driver_memory | (val << 32);
+        break;
+
+    /* Command */
+    case VMXNET3_REG_CMD:
+        VMW_CBPRN("Write BAR1 [VMXNET3_REG_CMD] = %" PRIx64 ", size %d",
+                  val, size);
+        vmxnet3_handle_command(s, val);
+        break;
+
+    /* MAC Address Low */
+    case VMXNET3_REG_MACL:
+        VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACL] = %" PRIx64 ", size %d",
+                  val, size);
+        s->temp_mac = val;
+        break;
+
+    /* MAC Address High */
+    case VMXNET3_REG_MACH:
+        VMW_CBPRN("Write BAR1 [VMXNET3_REG_MACH] = %" PRIx64 ", size %d",
+                  val, size);
+        vmxnet3_set_variable_mac(s, val, s->temp_mac);
+        break;
+
+    /* Interrupt Cause Register */
+    case VMXNET3_REG_ICR:
+        VMW_CBPRN("Write BAR1 [VMXNET3_REG_ICR] = %" PRIx64 ", size %d",
+                  val, size);
+        assert(false);
+        break;
+
+    /* Event Cause Register */
+    case VMXNET3_REG_ECR:
+        VMW_CBPRN("Write BAR1 [VMXNET3_REG_ECR] = %" PRIx64 ", size %d",
+                  val, size);
+        vmxnet3_ack_events(s, val);
+        break;
+
+    default:
+        VMW_CBPRN("Unknown Write to BAR1 [%" PRIx64 "] = %" PRIx64 ", size %d",
+                  addr, val, size);
+        break;
+    }
+}
+
+static uint64_t
+vmxnet3_io_bar1_read(void *opaque, hwaddr addr, unsigned size)
+{
+        VMXNET3State *s = opaque;
+        uint64_t ret = 0;
+
+        switch (addr) {
+        /* Vmxnet3 Revision Report Selection */
+        case VMXNET3_REG_VRRS:
+            VMW_CBPRN("Read BAR1 [VMXNET3_REG_VRRS], size %d", size);
+            ret = VMXNET3_DEVICE_REVISION;
+            break;
+
+        /* UPT Version Report Selection */
+        case VMXNET3_REG_UVRS:
+            VMW_CBPRN("Read BAR1 [VMXNET3_REG_UVRS], size %d", size);
+            ret = VMXNET3_DEVICE_VERSION;
+            break;
+
+        /* Command */
+        case VMXNET3_REG_CMD:
+            VMW_CBPRN("Read BAR1 [VMXNET3_REG_CMD], size %d", size);
+            ret = vmxnet3_get_command_status(s);
+            break;
+
+        /* MAC Address Low */
+        case VMXNET3_REG_MACL:
+            VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACL], size %d", size);
+            ret = vmxnet3_get_mac_low(&s->conf.macaddr);
+            break;
+
+        /* MAC Address High */
+        case VMXNET3_REG_MACH:
+            VMW_CBPRN("Read BAR1 [VMXNET3_REG_MACH], size %d", size);
+            ret = vmxnet3_get_mac_high(&s->conf.macaddr);
+            break;
+
+        /*
+         * Interrupt Cause Register
+         * Used for legacy interrupts only so interrupt index always 0
+         */
+        case VMXNET3_REG_ICR:
+            VMW_CBPRN("Read BAR1 [VMXNET3_REG_ICR], size %d", size);
+            if (vmxnet3_interrupt_asserted(s, 0)) {
+                vmxnet3_clear_interrupt(s, 0);
+                ret = true;
+            } else {
+                ret = false;
+            }
+            break;
+
+        default:
+            VMW_CBPRN("Unknow read BAR1[%" PRIx64 "], %d bytes", addr, size);
+            break;
+        }
+
+        return ret;
+}
+
+static int
+vmxnet3_can_receive(NetClientState *nc)
+{
+    VMXNET3State *s = qemu_get_nic_opaque(nc);
+    return s->device_active &&
+           VMXNET_FLAG_IS_SET(s->link_status_and_speed, VMXNET3_LINK_STATUS_UP);
+}
+
+static inline bool
+vmxnet3_is_registered_vlan(VMXNET3State *s, const void *data)
+{
+    uint16_t vlan_tag = eth_get_pkt_tci(data) & VLAN_VID_MASK;
+    if (IS_SPECIAL_VLAN_ID(vlan_tag)) {
+        return true;
+    }
+
+    return VMXNET3_VFTABLE_ENTRY_IS_SET(s->vlan_table, vlan_tag);
+}
+
+static bool
+vmxnet3_is_allowed_mcast_group(VMXNET3State *s, const uint8_t *group_mac)
+{
+    int i;
+    for (i = 0; i < s->mcast_list_len; i++) {
+        if (!memcmp(group_mac, s->mcast_list[i].a, sizeof(s->mcast_list[i]))) {
+            return true;
+        }
+    }
+    return false;
+}
+
+static bool
+vmxnet3_rx_filter_may_indicate(VMXNET3State *s, const void *data,
+    size_t size)
+{
+    struct eth_header *ehdr = PKT_GET_ETH_HDR(data);
+
+    if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_PROMISC)) {
+        return true;
+    }
+
+    if (!vmxnet3_is_registered_vlan(s, data)) {
+        return false;
+    }
+
+    switch (vmxnet_rx_pkt_get_packet_type(s->rx_pkt)) {
+    case ETH_PKT_UCAST:
+        if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_UCAST)) {
+            return false;
+        }
+        if (memcmp(s->conf.macaddr.a, ehdr->h_dest, ETH_ALEN)) {
+            return false;
+        }
+        break;
+
+    case ETH_PKT_BCAST:
+        if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_BCAST)) {
+            return false;
+        }
+        break;
+
+    case ETH_PKT_MCAST:
+        if (VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_ALL_MULTI)) {
+            return true;
+        }
+        if (!VMXNET_FLAG_IS_SET(s->rx_mode, VMXNET3_RXM_MCAST)) {
+            return false;
+        }
+        if (!vmxnet3_is_allowed_mcast_group(s, ehdr->h_dest)) {
+            return false;
+        }
+        break;
+
+    default:
+        assert(false);
+    }
+
+    return true;
+}
+
+static ssize_t
+vmxnet3_receive(NetClientState *nc, const uint8_t *buf, size_t size)
+{
+    VMXNET3State *s = qemu_get_nic_opaque(nc);
+    size_t bytes_indicated;
+
+    if (!vmxnet3_can_receive(nc)) {
+        VMW_PKPRN("Cannot receive now");
+        return -1;
+    }
+
+    if (s->peer_has_vhdr) {
+        vmxnet_rx_pkt_set_vhdr(s->rx_pkt, (struct virtio_net_hdr *)buf);
+        buf += sizeof(struct virtio_net_hdr);
+        size -= sizeof(struct virtio_net_hdr);
+    }
+
+    vmxnet_rx_pkt_set_packet_type(s->rx_pkt,
+        get_eth_packet_type(PKT_GET_ETH_HDR(buf)));
+
+    if (vmxnet3_rx_filter_may_indicate(s, buf, size)) {
+        vmxnet_rx_pkt_attach_data(s->rx_pkt, buf, size, s->rx_vlan_stripping);
+        bytes_indicated = vmxnet3_indicate_packet(s) ? size : -1;
+        if (bytes_indicated < size) {
+            VMW_PKPRN("RX: %lu of %lu bytes indicated", bytes_indicated, size);
+        }
+    } else {
+        VMW_PKPRN("Packet dropped by RX filter");
+        bytes_indicated = size;
+    }
+
+    assert(size > 0);
+    assert(bytes_indicated != 0);
+    return bytes_indicated;
+}
+
+static void vmxnet3_cleanup(NetClientState *nc)
+{
+    VMXNET3State *s = qemu_get_nic_opaque(nc);
+    s->nic = NULL;
+}
+
+static void vmxnet3_set_link_status(NetClientState *nc)
+{
+    VMXNET3State *s = qemu_get_nic_opaque(nc);
+
+    if (nc->link_down) {
+        s->link_status_and_speed &= ~VMXNET3_LINK_STATUS_UP;
+    } else {
+        s->link_status_and_speed |= VMXNET3_LINK_STATUS_UP;
+    }
+
+    vmxnet3_set_events(s, VMXNET3_ECR_LINK);
+    vmxnet3_trigger_interrupt(s, s->event_int_idx);
+}
+
+static NetClientInfo net_vmxnet3_info = {
+        .type = NET_CLIENT_OPTIONS_KIND_NIC,
+        .size = sizeof(NICState),
+        .can_receive = vmxnet3_can_receive,
+        .receive = vmxnet3_receive,
+        .cleanup = vmxnet3_cleanup,
+        .link_status_changed = vmxnet3_set_link_status,
+};
+
+static bool vmxnet3_peer_has_vnet_hdr(VMXNET3State *s)
+{
+    NetClientState *peer = qemu_get_queue(s->nic)->peer;
+
+    if ((NULL != peer)                              &&
+        (peer->info->type == NET_CLIENT_OPTIONS_KIND_TAP)   &&
+        tap_has_vnet_hdr(peer)) {
+        return true;
+    }
+
+    VMW_WRPRN("Peer has no virtio extension. Task offloads will be emulated.");
+    return false;
+}
+
+static void vmxnet3_net_uninit(VMXNET3State *s)
+{
+    g_free(s->mcast_list);
+    vmxnet_tx_pkt_reset(s->tx_pkt);
+    vmxnet_tx_pkt_uninit(s->tx_pkt);
+    vmxnet_rx_pkt_uninit(s->rx_pkt);
+    qemu_del_net_client(qemu_get_queue(s->nic));
+}
+
+static void vmxnet3_net_init(VMXNET3State *s)
+{
+    DeviceState *d = DEVICE(s);
+
+    VMW_CBPRN("vmxnet3_net_init called...");
+
+    qemu_macaddr_default_if_unset(&s->conf.macaddr);
+
+    /* Windows guest will query the address that was set on init */
+    memcpy(&s->perm_mac.a, &s->conf.macaddr.a, sizeof(s->perm_mac.a));
+
+    s->mcast_list = NULL;
+    s->mcast_list_len = 0;
+
+    s->link_status_and_speed = VMXNET3_LINK_SPEED | VMXNET3_LINK_STATUS_UP;
+
+    VMW_CFPRN("Permanent MAC: " MAC_FMT, MAC_ARG(s->perm_mac.a));
+
+    s->nic = qemu_new_nic(&net_vmxnet3_info, &s->conf,
+                          object_get_typename(OBJECT(s)),
+                          d->id, s);
+
+    s->peer_has_vhdr = vmxnet3_peer_has_vnet_hdr(s);
+    s->tx_sop = true;
+    s->skip_current_tx_pkt = false;
+    s->tx_pkt = NULL;
+    s->rx_pkt = NULL;
+    s->rx_vlan_stripping = false;
+    s->lro_supported = false;
+
+    if (s->peer_has_vhdr) {
+        tap_set_vnet_hdr_len(qemu_get_queue(s->nic)->peer,
+            sizeof(struct virtio_net_hdr));
+
+        tap_using_vnet_hdr(qemu_get_queue(s->nic)->peer, 1);
+    }
+
+    qemu_format_nic_info_str(qemu_get_queue(s->nic), s->conf.macaddr.a);
+}
+
+static void
+vmxnet3_unuse_msix_vectors(VMXNET3State *s, int num_vectors)
+{
+    PCIDevice *d = PCI_DEVICE(s);
+    int i;
+    for (i = 0; i < num_vectors; i++) {
+        msix_vector_unuse(d, i);
+    }
+}
+
+static bool
+vmxnet3_use_msix_vectors(VMXNET3State *s, int num_vectors)
+{
+    PCIDevice *d = PCI_DEVICE(s);
+    int i;
+    for (i = 0; i < num_vectors; i++) {
+        int res = msix_vector_use(d, i);
+        if (0 > res) {
+            VMW_WRPRN("Failed to use MSI-X vector %d, error %d", i, res);
+            vmxnet3_unuse_msix_vectors(s, i);
+            return false;
+        }
+    }
+    return true;
+}
+
+static bool
+vmxnet3_init_msix(VMXNET3State *s)
+{
+    PCIDevice *d = PCI_DEVICE(s);
+    int res = msix_init(d, VMXNET3_MAX_INTRS,
+                        &s->msix_bar,
+                        VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_TABLE,
+                        &s->msix_bar,
+                        VMXNET3_MSIX_BAR_IDX, VMXNET3_OFF_MSIX_PBA,
+                        0);
+
+    if (0 > res) {
+        VMW_WRPRN("Failed to initialize MSI-X, error %d", res);
+        s->msix_used = false;
+    } else {
+        if (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
+            VMW_WRPRN("Failed to use MSI-X vectors, error %d", res);
+            msix_uninit(d, &s->msix_bar, &s->msix_bar);
+            s->msix_used = false;
+        } else {
+            s->msix_used = true;
+        }
+    }
+    return s->msix_used;
+}
+
+static void
+vmxnet3_cleanup_msix(VMXNET3State *s)
+{
+    PCIDevice *d = PCI_DEVICE(s);
+
+    if (s->msix_used) {
+        msix_vector_unuse(d, VMXNET3_MAX_INTRS);
+        msix_uninit(d, &s->msix_bar, &s->msix_bar);
+    }
+}
+
+#define VMXNET3_MSI_NUM_VECTORS   (1)
+#define VMXNET3_MSI_OFFSET        (0x50)
+#define VMXNET3_USE_64BIT         (true)
+#define VMXNET3_PER_VECTOR_MASK   (false)
+
+static bool
+vmxnet3_init_msi(VMXNET3State *s)
+{
+    PCIDevice *d = PCI_DEVICE(s);
+    int res;
+
+    res = msi_init(d, VMXNET3_MSI_OFFSET, VMXNET3_MSI_NUM_VECTORS,
+                   VMXNET3_USE_64BIT, VMXNET3_PER_VECTOR_MASK);
+    if (0 > res) {
+        VMW_WRPRN("Failed to initialize MSI, error %d", res);
+        s->msi_used = false;
+    } else {
+        s->msi_used = true;
+    }
+
+    return s->msi_used;
+}
+
+static void
+vmxnet3_cleanup_msi(VMXNET3State *s)
+{
+    PCIDevice *d = PCI_DEVICE(s);
+
+    if (s->msi_used) {
+        msi_uninit(d);
+    }
+}
+
+static void
+vmxnet3_msix_save(QEMUFile *f, void *opaque)
+{
+    PCIDevice *d = PCI_DEVICE(opaque);
+    msix_save(d, f);
+}
+
+static int
+vmxnet3_msix_load(QEMUFile *f, void *opaque, int version_id)
+{
+    PCIDevice *d = PCI_DEVICE(opaque);
+    msix_load(d, f);
+    return 0;
+}
+
+static const MemoryRegionOps b0_ops = {
+    .read = vmxnet3_io_bar0_read,
+    .write = vmxnet3_io_bar0_write,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .impl = {
+            .min_access_size = 4,
+            .max_access_size = 4,
+    },
+};
+
+static const MemoryRegionOps b1_ops = {
+    .read = vmxnet3_io_bar1_read,
+    .write = vmxnet3_io_bar1_write,
+    .endianness = DEVICE_LITTLE_ENDIAN,
+    .impl = {
+            .min_access_size = 4,
+            .max_access_size = 4,
+    },
+};
+
+static int vmxnet3_pci_init(PCIDevice *pci_dev)
+{
+    DeviceState *dev = DEVICE(pci_dev);
+    VMXNET3State *s = VMXNET3(pci_dev);
+
+    VMW_CBPRN("Starting init...");
+
+    memory_region_init_io(&s->bar0, &b0_ops, s,
+                          "vmxnet3-b0", VMXNET3_PT_REG_SIZE);
+    pci_register_bar(pci_dev, VMXNET3_BAR0_IDX,
+                     PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar0);
+
+    memory_region_init_io(&s->bar1, &b1_ops, s,
+                          "vmxnet3-b1", VMXNET3_VD_REG_SIZE);
+    pci_register_bar(pci_dev, VMXNET3_BAR1_IDX,
+                     PCI_BASE_ADDRESS_SPACE_MEMORY, &s->bar1);
+
+    memory_region_init(&s->msix_bar, "vmxnet3-msix-bar",
+                       VMXNET3_MSIX_BAR_SIZE);
+    pci_register_bar(pci_dev, VMXNET3_MSIX_BAR_IDX,
+                     PCI_BASE_ADDRESS_SPACE_MEMORY, &s->msix_bar);
+
+    vmxnet3_reset_interrupt_states(s);
+
+    /* Interrupt pin A */
+    pci_dev->config[PCI_INTERRUPT_PIN] = 0x01;
+
+    if (!vmxnet3_init_msix(s)) {
+        VMW_WRPRN("Failed to initialize MSI-X, configuration is inconsistent.");
+    }
+
+    if (!vmxnet3_init_msi(s)) {
+        VMW_WRPRN("Failed to initialize MSI, configuration is inconsistent.");
+    }
+
+    vmxnet3_net_init(s);
+
+    register_savevm(dev, "vmxnet3-msix", -1, 1,
+                    vmxnet3_msix_save, vmxnet3_msix_load, s);
+
+    add_boot_device_path(s->conf.bootindex, dev, "/ethernet-phy@0");
+
+    return 0;
+}
+
+
+static void vmxnet3_pci_uninit(PCIDevice *pci_dev)
+{
+    DeviceState *dev = DEVICE(pci_dev);
+    VMXNET3State *s = VMXNET3(pci_dev);
+
+    VMW_CBPRN("Starting uninit...");
+
+    unregister_savevm(dev, "vmxnet3-msix", s);
+
+    vmxnet3_net_uninit(s);
+
+    vmxnet3_cleanup_msix(s);
+
+    vmxnet3_cleanup_msi(s);
+
+    memory_region_destroy(&s->bar0);
+    memory_region_destroy(&s->bar1);
+    memory_region_destroy(&s->msix_bar);
+}
+
+static void vmxnet3_qdev_reset(DeviceState *dev)
+{
+    PCIDevice *d = PCI_DEVICE(dev);
+    VMXNET3State *s = VMXNET3(d);
+
+    VMW_CBPRN("Starting QDEV reset...");
+    vmxnet3_reset(s);
+}
+
+static bool vmxnet3_mc_list_needed(void *opaque)
+{
+    return true;
+}
+
+static int vmxnet3_mcast_list_pre_load(void *opaque)
+{
+    VMXNET3State *s = opaque;
+
+    s->mcast_list = g_malloc(s->mcast_list_buff_size);
+
+    return 0;
+}
+
+
+static void vmxnet3_pre_save(void *opaque)
+{
+    VMXNET3State *s = opaque;
+
+    s->mcast_list_buff_size = s->mcast_list_len * sizeof(MACAddr);
+}
+
+static const VMStateDescription vmxstate_vmxnet3_mcast_list = {
+    .name = "vmxnet3/mcast_list",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .minimum_version_id_old = 1,
+    .pre_load = vmxnet3_mcast_list_pre_load,
+    .fields = (VMStateField[]) {
+        VMSTATE_VBUFFER_UINT32(mcast_list, VMXNET3State, 0, NULL, 0,
+            mcast_list_buff_size),
+        VMSTATE_END_OF_LIST()
+    }
+};
+
+static void vmxnet3_get_ring_from_file(QEMUFile *f, Vmxnet3Ring *r)
+{
+    r->pa = qemu_get_be64(f);
+    r->size = qemu_get_be32(f);
+    r->cell_size = qemu_get_be32(f);
+    r->next = qemu_get_be32(f);
+    r->gen = qemu_get_byte(f);
+}
+
+static void vmxnet3_put_ring_to_file(QEMUFile *f, Vmxnet3Ring *r)
+{
+    qemu_put_be64(f, r->pa);
+    qemu_put_be32(f, r->size);
+    qemu_put_be32(f, r->cell_size);
+    qemu_put_be32(f, r->next);
+    qemu_put_byte(f, r->gen);
+}
+
+static void vmxnet3_get_tx_stats_from_file(QEMUFile *f,
+    struct UPT1_TxStats *tx_stat)
+{
+    tx_stat->TSOPktsTxOK = qemu_get_be64(f);
+    tx_stat->TSOBytesTxOK = qemu_get_be64(f);
+    tx_stat->ucastPktsTxOK = qemu_get_be64(f);
+    tx_stat->ucastBytesTxOK = qemu_get_be64(f);
+    tx_stat->mcastPktsTxOK = qemu_get_be64(f);
+    tx_stat->mcastBytesTxOK = qemu_get_be64(f);
+    tx_stat->bcastPktsTxOK = qemu_get_be64(f);
+    tx_stat->bcastBytesTxOK = qemu_get_be64(f);
+    tx_stat->pktsTxError = qemu_get_be64(f);
+    tx_stat->pktsTxDiscard = qemu_get_be64(f);
+}
+
+static void vmxnet3_put_tx_stats_to_file(QEMUFile *f,
+    struct UPT1_TxStats *tx_stat)
+{
+    qemu_put_be64(f, tx_stat->TSOPktsTxOK);
+    qemu_put_be64(f, tx_stat->TSOBytesTxOK);
+    qemu_put_be64(f, tx_stat->ucastPktsTxOK);
+    qemu_put_be64(f, tx_stat->ucastBytesTxOK);
+    qemu_put_be64(f, tx_stat->mcastPktsTxOK);
+    qemu_put_be64(f, tx_stat->mcastBytesTxOK);
+    qemu_put_be64(f, tx_stat->bcastPktsTxOK);
+    qemu_put_be64(f, tx_stat->bcastBytesTxOK);
+    qemu_put_be64(f, tx_stat->pktsTxError);
+    qemu_put_be64(f, tx_stat->pktsTxDiscard);
+}
+
+static int vmxnet3_get_txq_descr(QEMUFile *f, void *pv, size_t size)
+{
+    Vmxnet3TxqDescr *r = pv;
+
+    vmxnet3_get_ring_from_file(f, &r->tx_ring);
+    vmxnet3_get_ring_from_file(f, &r->comp_ring);
+    r->intr_idx = qemu_get_byte(f);
+    r->tx_stats_pa = qemu_get_be64(f);
+
+    vmxnet3_get_tx_stats_from_file(f, &r->txq_stats);
+
+    return 0;
+}
+
+static void vmxnet3_put_txq_descr(QEMUFile *f, void *pv, size_t size)
+{
+    Vmxnet3TxqDescr *r = pv;
+
+    vmxnet3_put_ring_to_file(f, &r->tx_ring);
+    vmxnet3_put_ring_to_file(f, &r->comp_ring);
+    qemu_put_byte(f, r->intr_idx);
+    qemu_put_be64(f, r->tx_stats_pa);
+    vmxnet3_put_tx_stats_to_file(f, &r->txq_stats);
+}
+
+const VMStateInfo txq_descr_info = {
+    .name = "txq_descr",
+    .get = vmxnet3_get_txq_descr,
+    .put = vmxnet3_put_txq_descr
+};
+
+static void vmxnet3_get_rx_stats_from_file(QEMUFile *f,
+    struct UPT1_RxStats *rx_stat)
+{
+    rx_stat->LROPktsRxOK = qemu_get_be64(f);
+    rx_stat->LROBytesRxOK = qemu_get_be64(f);
+    rx_stat->ucastPktsRxOK = qemu_get_be64(f);
+    rx_stat->ucastBytesRxOK = qemu_get_be64(f);
+    rx_stat->mcastPktsRxOK = qemu_get_be64(f);
+    rx_stat->mcastBytesRxOK = qemu_get_be64(f);
+    rx_stat->bcastPktsRxOK = qemu_get_be64(f);
+    rx_stat->bcastBytesRxOK = qemu_get_be64(f);
+    rx_stat->pktsRxOutOfBuf = qemu_get_be64(f);
+    rx_stat->pktsRxError = qemu_get_be64(f);
+}
+
+static void vmxnet3_put_rx_stats_to_file(QEMUFile *f,
+    struct UPT1_RxStats *rx_stat)
+{
+    qemu_put_be64(f, rx_stat->LROPktsRxOK);
+    qemu_put_be64(f, rx_stat->LROBytesRxOK);
+    qemu_put_be64(f, rx_stat->ucastPktsRxOK);
+    qemu_put_be64(f, rx_stat->ucastBytesRxOK);
+    qemu_put_be64(f, rx_stat->mcastPktsRxOK);
+    qemu_put_be64(f, rx_stat->mcastBytesRxOK);
+    qemu_put_be64(f, rx_stat->bcastPktsRxOK);
+    qemu_put_be64(f, rx_stat->bcastBytesRxOK);
+    qemu_put_be64(f, rx_stat->pktsRxOutOfBuf);
+    qemu_put_be64(f, rx_stat->pktsRxError);
+}
+
+static int vmxnet3_get_rxq_descr(QEMUFile *f, void *pv, size_t size)
+{
+    Vmxnet3RxqDescr *r = pv;
+    int i;
+
+    for (i = 0; i < VMXNET3_RX_RINGS_PER_QUEUE; i++) {
+        vmxnet3_get_ring_from_file(f, &r->rx_ring[i]);
+    }
+
+    vmxnet3_get_ring_from_file(f, &r->comp_ring);
+    r->intr_idx = qemu_get_byte(f);
+    r->rx_stats_pa = qemu_get_be64(f);
+
+    vmxnet3_get_rx_stats_from_file(f, &r->rxq_stats);
+
+    return 0;
+}
+
+static void vmxnet3_put_rxq_descr(QEMUFile *f, void *pv, size_t size)
+{
+    Vmxnet3RxqDescr *r = pv;
+    int i;
+
+    for (i = 0; i < VMXNET3_RX_RINGS_PER_QUEUE; i++) {
+        vmxnet3_put_ring_to_file(f, &r->rx_ring[i]);
+    }
+
+    vmxnet3_put_ring_to_file(f, &r->comp_ring);
+    qemu_put_byte(f, r->intr_idx);
+    qemu_put_be64(f, r->rx_stats_pa);
+    vmxnet3_put_rx_stats_to_file(f, &r->rxq_stats);
+}
+
+static int vmxnet3_post_load(void *opaque, int version_id)
+{
+    VMXNET3State *s = opaque;
+    PCIDevice *d = PCI_DEVICE(s);
+
+    vmxnet_tx_pkt_init(&s->tx_pkt, s->max_tx_frags, s->peer_has_vhdr);
+    vmxnet_rx_pkt_init(&s->rx_pkt, s->peer_has_vhdr);
+
+    if (s->msix_used) {
+        if  (!vmxnet3_use_msix_vectors(s, VMXNET3_MAX_INTRS)) {
+            VMW_WRPRN("Failed to re-use MSI-X vectors");
+            msix_uninit(d, &s->msix_bar, &s->msix_bar);
+            s->msix_used = false;
+            return -1;
+        }
+    }
+
+    return 0;
+}
+
+const VMStateInfo rxq_descr_info = {
+    .name = "rxq_descr",
+    .get = vmxnet3_get_rxq_descr,
+    .put = vmxnet3_put_rxq_descr
+};
+
+static int vmxnet3_get_int_state(QEMUFile *f, void *pv, size_t size)
+{
+    Vmxnet3IntState *r = pv;
+
+    r->is_masked = qemu_get_byte(f);
+    r->is_pending = qemu_get_byte(f);
+    r->is_asserted = qemu_get_byte(f);
+
+    return 0;
+}
+
+static void vmxnet3_put_int_state(QEMUFile *f, void *pv, size_t size)
+{
+    Vmxnet3IntState *r = pv;
+
+    qemu_put_byte(f, r->is_masked);
+    qemu_put_byte(f, r->is_pending);
+    qemu_put_byte(f, r->is_asserted);
+}
+
+const VMStateInfo int_state_info = {
+    .name = "int_state",
+    .get = vmxnet3_get_int_state,
+    .put = vmxnet3_put_int_state
+};
+
+static const VMStateDescription vmstate_vmxnet3 = {
+    .name = "vmxnet3",
+    .version_id = 1,
+    .minimum_version_id = 1,
+    .minimum_version_id_old = 1,
+    .pre_save = vmxnet3_pre_save,
+    .post_load = vmxnet3_post_load,
+    .fields      = (VMStateField[]) {
+            VMSTATE_PCI_DEVICE(parent_obj, VMXNET3State),
+            VMSTATE_BOOL(rx_packets_compound, VMXNET3State),
+            VMSTATE_BOOL(rx_vlan_stripping, VMXNET3State),
+            VMSTATE_BOOL(lro_supported, VMXNET3State),
+            VMSTATE_UINT32(rx_mode, VMXNET3State),
+            VMSTATE_UINT32(mcast_list_len, VMXNET3State),
+            VMSTATE_UINT32(mcast_list_buff_size, VMXNET3State),
+            VMSTATE_UINT32_ARRAY(vlan_table, VMXNET3State, VMXNET3_VFT_SIZE),
+            VMSTATE_UINT32(mtu, VMXNET3State),
+            VMSTATE_UINT16(max_rx_frags, VMXNET3State),
+            VMSTATE_UINT32(max_tx_frags, VMXNET3State),
+            VMSTATE_UINT8(event_int_idx, VMXNET3State),
+            VMSTATE_BOOL(auto_int_masking, VMXNET3State),
+            VMSTATE_UINT8(txq_num, VMXNET3State),
+            VMSTATE_UINT8(rxq_num, VMXNET3State),
+            VMSTATE_UINT32(device_active, VMXNET3State),
+            VMSTATE_UINT32(last_command, VMXNET3State),
+            VMSTATE_UINT32(link_status_and_speed, VMXNET3State),
+            VMSTATE_UINT32(temp_mac, VMXNET3State),
+            VMSTATE_UINT64(drv_shmem, VMXNET3State),
+            VMSTATE_UINT64(temp_shared_guest_driver_memory, VMXNET3State),
+
+            VMSTATE_ARRAY(txq_descr, VMXNET3State,
+                VMXNET3_DEVICE_MAX_TX_QUEUES, 0, txq_descr_info,
+                Vmxnet3TxqDescr),
+            VMSTATE_ARRAY(rxq_descr, VMXNET3State,
+                VMXNET3_DEVICE_MAX_RX_QUEUES, 0, rxq_descr_info,
+                Vmxnet3RxqDescr),
+            VMSTATE_ARRAY(interrupt_states, VMXNET3State, VMXNET3_MAX_INTRS,
+                0, int_state_info, Vmxnet3IntState),
+
+            VMSTATE_END_OF_LIST()
+    },
+    .subsections = (VMStateSubsection[]) {
+        {
+            .vmsd = &vmxstate_vmxnet3_mcast_list,
+            .needed = vmxnet3_mc_list_needed
+        },
+        {
+            /* empty element. */
+        }
+    }
+};
+
+static void
+vmxnet3_write_config(PCIDevice *pci_dev, uint32_t addr, uint32_t val, int len)
+{
+    pci_default_write_config(pci_dev, addr, val, len);
+    msix_write_config(pci_dev, addr, val, len);
+    msi_write_config(pci_dev, addr, val, len);
+}
+
+static Property vmxnet3_properties[] = {
+    DEFINE_NIC_PROPERTIES(VMXNET3State, conf),
+    DEFINE_PROP_END_OF_LIST(),
+};
+
+static void vmxnet3_class_init(ObjectClass *class, void *data)
+{
+    DeviceClass *dc = DEVICE_CLASS(class);
+    PCIDeviceClass *c = PCI_DEVICE_CLASS(class);
+
+    c->init = vmxnet3_pci_init;
+    c->exit = vmxnet3_pci_uninit;
+    c->vendor_id = PCI_VENDOR_ID_VMWARE;
+    c->device_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
+    c->revision = PCI_DEVICE_ID_VMWARE_VMXNET3_REVISION;
+    c->class_id = PCI_CLASS_NETWORK_ETHERNET;
+    c->subsystem_vendor_id = PCI_VENDOR_ID_VMWARE;
+    c->subsystem_id = PCI_DEVICE_ID_VMWARE_VMXNET3;
+    c->config_write = vmxnet3_write_config,
+    dc->desc = "VMWare Paravirtualized Ethernet v3";
+    dc->reset = vmxnet3_qdev_reset;
+    dc->vmsd = &vmstate_vmxnet3;
+    dc->props = vmxnet3_properties;
+}
+
+static const TypeInfo vmxnet3_info = {
+    .name          = TYPE_VMXNET3,
+    .parent        = TYPE_PCI_DEVICE,
+    .instance_size = sizeof(VMXNET3State),
+    .class_init    = vmxnet3_class_init,
+};
+
+static void vmxnet3_register_types(void)
+{
+    VMW_CBPRN("vmxnet3_register_types called...");
+    type_register_static(&vmxnet3_info);
+}
+
+type_init(vmxnet3_register_types)
hw/vmxnet3.h new file +31 -0
new file mode 100644
index 0000000..7db0c8f
@@ -0,0 +1,760 @@
+/*
+ * QEMU VMWARE VMXNET3 paravirtual NIC interface definitions
+ *
+ * Copyright (c) 2012 Ravello Systems LTD (http://ravellosystems.com)
+ *
+ * Developed by Daynix Computing LTD (http://www.daynix.com)
+ *
+ * Authors:
+ * Dmitry Fleytman <[email protected]>
+ * Tamir Shomer <[email protected]>
+ * Yan Vugenfirer <[email protected]>
+ *
+ * This work is licensed under the terms of the GNU GPL, version 2.
+ * See the COPYING file in the top-level directory.
+ *
+ */
+
+#ifndef _QEMU_VMXNET3_H
+#define _QEMU_VMXNET3_H
+
+#define VMXNET3_DEVICE_MAX_TX_QUEUES 8
+#define VMXNET3_DEVICE_MAX_RX_QUEUES 8   /* Keep this value as a power of 2 */
+
+/*
+ * VMWARE headers we got from Linux kernel do not fully comply QEMU coding
+ * standards in sense of types and defines used.
+ * Since we didn't want to change VMWARE code, following set of typedefs
+ * and defines needed to compile these headers with QEMU introduced.
+ */
+#define u64     uint64_t
+#define u32     uint32_t

hw/vmxnet3.h line 41

Quoted 10 lines
hw/vmxnet3.h new file +10 -0
+#define u16     uint16_t
+#define u8      uint8_t
+#define __le16  uint16_t
+#define __le32  uint32_t
+#define __le64  uint64_t
+#define __packed QEMU_PACKED
+
+#if defined(HOST_WORDS_BIGENDIAN)
+#define const_cpu_to_le64(x) bswap_64(x)
+#define __BIG_ENDIAN_BITFIELD

Ah, sorry, I replied to the wrong version.

ARE YOU KIDDING ME? BIG ENDIAN BITFIELD? BITFIELDS ARE _IMPLEMENTATION SPECIFIC_!

Can we please revert this whole patch set and send the authors back to school?

Alex

#7 lore raw

Re: [Qemu-devel] [PATCH V14 5/5] VMXNET3 device implementation

Il 27/03/2013 15:49, Alexander Graf ha scritto:

Quoted 8 lines
Quoted 3 lines

hw/vmxnet3.h line 41

Quoted 3 lines
hw/vmxnet3.h new file +3 -0
+#if defined(HOST_WORDS_BIGENDIAN)
+#define const_cpu_to_le64(x) bswap_64(x)
+#define __BIG_ENDIAN_BITFIELD

Ah, sorry, I replied to the wrong version.

ARE YOU KIDDING ME? BIG ENDIAN BITFIELD? BITFIELDS ARE _IMPLEMENTATION SPECIFIC_!

Can we please revert this whole patch set and send the authors back to school?

Can we please maintain a decent tone?

First, this file comes from Linux. __BIG_ENDIAN_BITFIELD is a Linux #define. No doubt it is wrong to define it based on HOST_WORDS_BIGENDIAN, it is better to use a configure check. But it's not the reason why PPC compilation fails.

Second, you haven't said _how_ it breaks PPC compilation. Just cut-and-paste from the compiler is enough. Ok, I can guess it but not always.

Third, there is no need to revert the patch set. The const_cpu_to_le64 should simply be removed, since little-endian conversion is already done in vmw_shmem_ld32.

Paolo

#8 lore raw

Re: [Qemu-devel] [PATCH V14 5/5] VMXNET3 device implementation

Quoted 8 lines · Paolo Bonzini

Il 27/03/2013 15:49, Alexander Graf ha scritto:

Quoted 7 lines
Quoted 3 lines

hw/vmxnet3.h line 41

Quoted 3 lines
hw/vmxnet3.h new file +3 -0
+#if defined(HOST_WORDS_BIGENDIAN)
+#define const_cpu_to_le64(x) bswap_64(x)
+#define __BIG_ENDIAN_BITFIELD

Ah, sorry, I replied to the wrong version.

ARE YOU KIDDING ME? BIG ENDIAN BITFIELD? BITFIELDS ARE _IMPLEMENTATION SPECIFIC_!

Quoted 8 lines
Quoted 1 line

Can we please revert this whole patch set and send the authors back to school?

Can we please maintain a decent tone?

First, this file comes from Linux. __BIG_ENDIAN_BITFIELD is a Linux #define. No doubt it is wrong to define it based on HOST_WORDS_BIGENDIAN, it is better to use a configure check. But it's not the reason why PPC compilation fails.

No, but it indicates that the code isn't written with portability in mind.

It's simply wrong to define it in the first place. You shouldn't do any assumptions how bitfields are laid out in memory / registers. Linux gets away with it mostly because it's heavily tied to gcc, but we shouldn't take the same assumptions in QEMU code.

Quoted 3 lines

Second, you haven't said _how_ it breaks PPC compilation. Just cut-and-paste from the compiler is enough. Ok, I can guess it but not always.

In file included from hw/vmxnet3.c:30:
hw/vmxnet3.h:140: error: braced-group within expression allowed only inside a function hw/vmxnet3.h:141: error: braced-group within expression allowed only inside a function hw/vmxnet3.h:142: error: braced-group within expression allowed only inside a function hw/vmxnet3.h:143: error: braced-group within expression allowed only inside a function

But it doesn't really matter what the reason for the breakage is, the code won't work on big endian hosts even with that bswap removed. In fact, the bswap was even a 64bit bswap on a 32bit guest field, so that code wouldn't even have worked if gcc hadn't complained (which probably means the Linux code here is broken).

Quoted 3 lines

Third, there is no need to revert the patch set. The const_cpu_to_le64 should simply be removed, since little-endian conversion is already done in vmw_shmem_ld32.

Yes, and bitfields should be converted to bitmasks. Then you don't have to worry at all about anything big or little endian except for the few interfaces to guest memory.

Whenever you see an explicit endianness swap, be sure to get very suspicious.

Alex

#9 lore raw

Re: [Qemu-devel] [PATCH V14 5/5] VMXNET3 device implementation

Quoted 22 lines · Alexander Graf
Quoted 7 lines · Paolo Bonzini

Il 27/03/2013 15:49, Alexander Graf ha scritto:

Quoted 6 lines
Quoted 3 lines

hw/vmxnet3.h line 41

Quoted 3 lines
hw/vmxnet3.h new file +3 -0
+#if defined(HOST_WORDS_BIGENDIAN)
+#define const_cpu_to_le64(x) bswap_64(x)
+#define __BIG_ENDIAN_BITFIELD

Ah, sorry, I replied to the wrong version.

ARE YOU KIDDING ME? BIG ENDIAN BITFIELD? BITFIELDS ARE _IMPLEMENTATION

SPECIFIC_!

Quoted 2 lines
Quoted 2 lines

Can we please revert this whole patch set and send the authors back to

school?

Quoted 7 lines

Can we please maintain a decent tone?

First, this file comes from Linux. __BIG_ENDIAN_BITFIELD is a Linux #define. No doubt it is wrong to define it based on HOST_WORDS_BIGENDIAN, it is better to use a configure check. But it's not the reason why PPC compilation fails.

Quoted 7 lines

No, but it indicates that the code isn't written with portability in mind.

It's simply wrong to define it in the first place. You shouldn't do any assumptions how bitfields are laid out in memory / registers. Linux gets away with it mostly because it's heavily tied to gcc, but we shouldn't take the same assumptions in QEMU code.

First of all I'd like to emphasize that vmxnet3 and pvscsi patches are our first experience in writing QEMU devices so most probably we miss some common practices of the project.
The codebase is huge and it takes time to learn things like this. At the same time we are open for constructive criticism and ready to improve the code we submit.

Regarding this specific case, we used an approach from current QEMU code, namely bitfield definitions technique from slirp/ip.h and slirp/tcp.h files. We do realize this is not the best way, if there are examples of better techniques in QEMU codebase please point it out.

Quoted 15 lines
Quoted 4 lines

Second, you haven't said _how_ it breaks PPC compilation. Just cut-and-paste from the compiler is enough. Ok, I can guess it but not always.

In file included from hw/vmxnet3.c:30:
hw/vmxnet3.h:140: error: braced-group within expression allowed only inside a function
hw/vmxnet3.h:141: error: braced-group within expression allowed only inside a function
hw/vmxnet3.h:142: error: braced-group within expression allowed only inside a function
hw/vmxnet3.h:143: error: braced-group within expression allowed only inside a function

Quoted 11 lines

But it doesn't really matter what the reason for the breakage is, the code won't work on big endian hosts even with that bswap removed. In fact, the bswap was even a 64bit bswap on a 32bit guest field, so that code wouldn't even have worked if gcc hadn't complained (which probably means the Linux code here is broken).

Quoted 3 lines

Third, there is no need to revert the patch set. The const_cpu_to_le64 should simply be removed, since little-endian conversion is already done in vmw_shmem_ld32.

Right, we got these definitions as is from VMXNET3 Linux driver header as is including cpu_to_le64() wrapping.
Indeed it defined improperly and should be removed/redefined empty as shared memory accessors doing byte swap automatically.
This is a bug, we'll fix this.

Quoted 8 lines

Yes, and bitfields should be converted to bitmasks. Then you don't have
to worry at all about anything big or little endian except for the few interfaces to guest memory.

Whenever you see an explicit endianness swap, be sure to get very suspicious.

The issue is we got device interface definition from Linux and we'd like to leave it unchanged where possible for easier merges of future driver changes.
Indeed bitmasks are much more clear.

Dmitry.

Quoted 4 lines

Alex

#10 lore raw

Re: [Qemu-devel] [PATCH V14 5/5] VMXNET3 device implementation

Quoted 39 lines · Dmitry Fleytman
Quoted 22 lines · Alexander Graf
Quoted 7 lines · Paolo Bonzini

Il 27/03/2013 15:49, Alexander Graf ha scritto:

Quoted 6 lines
Quoted 3 lines

hw/vmxnet3.h line 41

Quoted 3 lines
hw/vmxnet3.h new file +3 -0
+#if defined(HOST_WORDS_BIGENDIAN)
+#define const_cpu_to_le64(x) bswap_64(x)
+#define __BIG_ENDIAN_BITFIELD

Ah, sorry, I replied to the wrong version.

ARE YOU KIDDING ME? BIG ENDIAN BITFIELD? BITFIELDS ARE _IMPLEMENTATION

SPECIFIC_!

Quoted 2 lines
Quoted 2 lines

Can we please revert this whole patch set and send the authors back to

school?

Quoted 7 lines

Can we please maintain a decent tone?

First, this file comes from Linux. __BIG_ENDIAN_BITFIELD is a Linux #define. No doubt it is wrong to define it based on HOST_WORDS_BIGENDIAN, it is better to use a configure check. But it's not the reason why PPC compilation fails.

Quoted 7 lines

No, but it indicates that the code isn't written with portability in mind.

It's simply wrong to define it in the first place. You shouldn't do any assumptions how bitfields are laid out in memory / registers. Linux gets away with it mostly because it's heavily tied to gcc, but we shouldn't take the same assumptions in QEMU code.

First of all I'd like to emphasize that vmxnet3 and pvscsi patches are our first experience in writing QEMU devices so most probably we miss some common practices of the project.
The codebase is huge and it takes time to learn things like this. At the same time we are open for constructive criticism and ready to improve the code we submit.

Quoted 4 lines

Regarding this specific case, we used an approach from current QEMU code, namely bitfield definitions technique from slirp/ip.h and slirp/tcp.h files. We do realize this is not the best way, if there are examples of better techniques in QEMU codebase please point it out.

It may not be obvious since it's part of the source tree but slirp is not a good example of QEMU code. It actually comes from
http://slirp.sourceforge.net/.

Stefan

#11 lore raw
In reply to Dmitry Fleytman
Quoted 491 lines · Dmitry Fleytman

This set of patches implements VMWare VMXNET3 paravirtual NIC device. The device supports of all the device features including offload capabilties, VLANs and etc.
The device is tested on different OSes:

    Fedora 15
    Ubuntu 10.4
    Centos 6.2
    Windows 2008R2
    Windows 2008 64bit
    Windows 2008 32bit
    Windows 2003 64bit
    Windows 2003 32bit

Changes in V14:
QOM-related fixes

Changes in V13:
Licensing of vmxnet3.h changed to match original VMWare's header file license

Changes in V12:

QOM-related fixes

Changes in V11:
Rebased to master HEAD to fix multiqueue compilation issues

Changes in V10:

Issues reported by Stefan Hajnoczi fixed.

Changes in V9:

Issues reported by Stefan Hajnoczi fixed.

Changes in V8:

Issues reported by Stefan Hajnoczi reviewed and mostly fixed:

Review

Quoted 5 lines
(quoted diff) +5 -0
+        }
+        curr_src_off += src[i].iov_len;
+    }
+    return j;
+}

The existing iov_copy() function provides equivalent functionality. I don't think iov_rebuild() is needed.

[DF] Done. Thanks, missed it.

Review

Quoted 6 lines
(quoted diff) +6 -0
+            size -= len;
+        }
+        iovec_off += iov[i].iov_len;
+    }
+    return res;
+}

Rename this net_checksum_add_iov() and place it in net/checksum.c, then the new dependency on net from block can be dropped.

[DF] Done.

Quoted 5 lines

+vmw_shmem_read(hwaddr addr, void *buf, int len)

 {
     VMW_SHPRN("SHMEM r: %" PRIx64 ", len: %d to %p", addr, len, buf);
     cpu_physical_memory_read(addr, buf, len);

}

All changes to this file should be squashed with the previous patch.

[DF] Done

Review

Quoted 13 lines
(quoted diff) +9 -0
+#ifdef VMXNET_DEBUG_SHMEM_ACCESS
+#define VMW_SHPRN(fmt, ...)                                                  
 \
+    do {                                                                     
 \
+        printf("[%s][SH][%s]: " fmt "\n", VMXNET_DEVICE_NAME, __func__,      
 \
+            ## __VA_ARGS__);                                                 
 \
+    } while (0)
+#else
+#define VMW_SHPRN(fmt, ...) do {} while (0)
+#endif

Please use QEMU tracing. It eliminates all this boilerplate and conditional compilation. Tracing can be enabled/disabled at runtime and works with SystemTap/DTrace. See docs/tracing.txt.

[DF] We'd like to stick with compile time logic in this case becase of 2 reasons:
[DF] 1. These printouts are intended for reverse engineering/development only and there is
[DF] no need to enable them at run time
[DF] 2. There is a big number of printouts, all driver-device communication is traced,
[DF] they hit performance even on strongest x86 in case of run-time logic

Review

Quoted 5 lines
(quoted diff) +5 -0
+struct eth_header {
+    uint8_t  h_dest[ETH_ALEN];   /* destination eth addr */
+    uint8_t  h_source[ETH_ALEN]; /* source ether addr    */
+    uint16_t h_proto;            /* packet type ID field */
+};

Looks like it's copy-pasted stuff from /usr/include/linux/if_*.h, /usr/include/netinet/*.h, and friends. If the system-wide headers are included names will collide for some of the macros at least.

Did you check if the slirp/ definitions can be reused?

[DF] Yes, you are right. This is copy-pasted from different places. [DF] Slips definishing do not fully cover our needs.

I'd rather we import network header definitions once in a generic place into the source tree. That way vmxnet and other components don't need to redefine these structs.

[DF] Exaclty! Our intention is to create generic header with network definitions and make everyone use it. [DF] We can move our header to some shared place if you want, however I'd do it in parallel with cleanup [DF] of similar definitions in existing code and this is a big change that os out of scope of these patches.

Quoted 2 lines

+ *===========================================================================*/

Is this huge comment box a sign that the code should be split into a foo_tx.c and an foo_rx.c file?

[DF] As for me this file is not that big to be splitted (<800 lines), however I'll do this if you insist :)

Quoted 1 line

+size_t vmxnet_tx_pkt_send(VmxnetTxPktH pkt, NetClientState *vc)

'vc' is an old name that was used for VLANClientState. The struct has since been renamed to NetClientState and the rest of QEMU uses 'nc' instead of 'vc'.

[DF] Fixed. Thanks.

Review

Quoted 2 lines
(quoted diff) +2 -0
+/* tx module context handle */
+typedef void *VmxnetTxPktH;

Forward-declaring the struct is nicer:

typedef struct VmxnetTxPkt VmxnetTxPkt;

The definition of VmxnetTxPkt is still hidden from the caller but you avoid the void* and casting. In vmxnet_pkt.c define using:

struct VmxnetTxPkt {
    ...
};

[DF] Agree, fixed. Thanks.

hw/vmxnet_utils.h

Quoted 4 lines
hw/vmxnet_utils.h +0 -0
index 7fd9a01..fac3b7b 100644

Please squash these fixes into the previous patch.

[DF] Oops, my bad. Fixed.

hw/vmxnet_utils.h

Quoted 3 lines
hw/vmxnet_utils.h +3 -0
+        uint8_t msix_used;
+        /* Whether MSI support was installed successfully */
+        uint8_t msi_used;

These two fields should be bool.

hw/vmxnet_utils.h

Quoted 2 lines
hw/vmxnet_utils.h +2 -0
+        /* Whether automatic interrupts masking enabled */
+        uint8_t auto_int_masking;

bool

[DF] Done.

hw/vmxnet_utils.h

Quoted 9 lines
hw/vmxnet_utils.h +9 -0
+static inline void vmxnet3_flush_shmem_changes(void)
+{
+    /*
+     * Flush shared memory changes
+     * Needed before sending interrupt to guest to ensure
+     * it gets consistent memory state
+     */
+    smp_wmb();
+}

It's useful to document why a memory barrier is being used in each instance. Therefore hiding smp_wmb() inside a wrapper function isn't great.

[DF] Fixed.

Also, it's suspicious that smb_wmb() is used but no other barriers are used. What about a read memory barrier when accessing shared memory written by the guest?

[DF] VMWARE interface build a in safe way - you always read out data buffer (packet descriptor) with memcopy, [DF] and then check its last byte to see whether it was fully filled by driver.
[DF] In this case device doesn't need read barriers. Drivers need write barriers of course to secure [DF] proper order of writes.

Changes in V7:

Issues reported by Michael S. Tsirkin reviewed and mostly fixed:

File vmware_utils.h:

...

hw/vmxnet_utils.h

Quoted 7 lines
hw/vmxnet_utils.h +7 -0
+static inline void
+vmw_shmem_st64(target_phys_addr_t addr, uint64_t value)
+{
+    VMW_SHPRN("SHMEM store64: %" PRIx64 " (value %" PRIx64 ")", addr, value);
+    stq_le_phys(addr, value);
+}
+

Pls remove these wrappers. These are just memory stores. Our codebase is too large as it is without every driver wrapping all standard calls.

[DF] Idea behind this macros is to have printout in each of them [DF] Printouts are really needed when reverse-engineering windows guest drivers [DF] Since there is no windows drivers code this is the only way to understand [DF] sequences they use

Quoted 1 line

+/* MACROS for simplification of operations on array-style registers */

UPPERCASE ABUSE

[DF] Fixed

hw/vmxnet_utils.h

Quoted 2 lines
hw/vmxnet_utils.h +2 -0
+#define IS_MULTIREG_ADDR(addr, base, cnt, regsize)                 \
+    (((addr + 1) > (base)) && ((addr) < (base) + (cnt) * (regsize)))

Same as range_covers_byte(base, cnt * regsize, addr)?

[DF] Fixed, thanks

hw/vmxnet_utils.h

Quoted 4 lines
hw/vmxnet_utils.h +4 -0
+
+#define MULTIREG_IDX_BY_ADDR(addr, base, regsize)                  \
+    (((addr) - (base)) / (regsize))
+

Above two macros is all that's left. No objection but it does not say what they do - want to add minimal documentation?
And please prefix with VMWARE_ or something.

[DF] Prefix added, macros documented

File vmxnet_utils.h (and related with the same problems):

...

hw/vmxnet_utils.h

Quoted 6 lines

+ * See the COPYING file in the top-level directory.
+ *
+ */

hw/vmxnet_utils.h +3 -0
+
+#ifndef _VMXNET_UTILS_H_
+#define _VMXNET_UTILS_H_

Please do not start identifiers with _ followed by an uppercase lattters.

[DF] Fixed

...

hw/vmxnet_utils.h

Quoted 7 lines
hw/vmxnet_utils.h +7 -0
+
+struct eth_header {
+    uint8_t  h_dest[ETH_ALEN];   /* destination eth addr */
+    uint8_t  h_source[ETH_ALEN]; /* source ether addr    */
+    uint16_t h_proto;            /* packet type ID field */
+};
+

And fix struct definitions to
1. follow qemu coding style
[DF] It's not clear what's wrong with the coding style. Please, elaborate.
2. start with vmxnet
[DF] Since this is generic network definitions with no device specifics
[DF] I'm not sure it makes sense to add vmxnet prefix.
[DF] I'd say it makes sense to put them into some generic header files under "net" directory instead [DF] and clean up other devices that have their local definitions of the same structures. [DF] Please, advise.

I also don't really understand why are these functions split out - vmxnet is the only user, no?

[DF] Currently vmxnet3 is the only user, hovewer we have vmxnet2 implementation as well (not submitted yet) [DF] and it uses the same header

File vmxnet_pkt.h:

...

hw/vmxnet_utils.h

Quoted 3 lines
hw/vmxnet_utils.h +3 -0
+
+/* tx module context handle */
+typedef void *VmxnetTxPktH;

This gets you zero type safety and makes debugging impossible. Just use pointers like everyone does.

[DF] Handle type is added to hide VmxnetTxPkt structure into .c header [DF] for better encapsulation.
[DF] Should we drop it anyway?

Files vmxnet3.*:

...

hw/vmxnet_utils.h

Quoted 3 lines
hw/vmxnet_utils.h +3 -0
+
+    if (zero_region) {
+        vmw_shmem_set(pa, 0, size*cell_size);

spaces around *

[DF] Fixed

...

Quoted 7 lines

+#define vmxnet3_ring_dump(macro, ring_name, ridx, r) \
+ macro("%s#%d: base %" PRIx64 " size %lu cell_size %lu gen %d next %lu",
\
+ (ring_name), (ridx),
\
+ (r)->pa, (r)->size, (r)->cell_size, (r)->gen, (r)->next)

make macros upper case

[DF] Fixed

...

hw/vmxnet_utils.h

Quoted 5 lines
hw/vmxnet_utils.h +5 -0
+static inline void vmxnet3_flush_shmem_changes(void)
+{
+    /*
+     * Flush shared memory changes
+     * Needed before transferring control to guest

what does 'transferring control to guest' mean?
[DF] Changed to
[DF] /*

[DF]     * Flush shared memory changes
[DF]     * Needed before sending interrupt to guest to ensure
[DF]     * it gets consistent memory state
[DF]     */

...

hw/vmxnet_utils.h

Quoted 3 lines
hw/vmxnet_utils.h +3 -0
+     */
+    smp_wmb();
+}

Don't use wrappers like this. They just hide bugs. For example it's not helpful before an interrupt in the function below.

[DF] I guess you are talking about vmxnet3_complete_packet() [DF] Strictly speaking barrier is a must because we change shared memory in [DF] vmxnet3_complete_packet()

[DF] And the wrapper is a good thing because its name explains its effect [DF] in a formal way as opposed to comments

...

hw/vmxnet_utils.h

Quoted 2 lines
hw/vmxnet_utils.h +2 -0
+    switch (status) {
+    case VMXNET3_PKT_STATUS_OK: {

don't put {} around cases: they align incorrectly
if it's too big move to a function.

[DF] Fixed

...

hw/vmxnet_utils.h

Quoted 5 lines
hw/vmxnet_utils.h +5 -0
+static bool
+vmxnet3_send_packet(VMXNET3State *s, uint32_t qidx)
+{
+    size_t bytes_sent = 0;
+    bool res = true;

why = true? don't initialize just because.

[DF] Fixed

...

Quoted 6 lines

+/*
+ * VMWARE headers we got from Linux kernel do not fully comply QEMU coding
+ * standards in sense of types and defines used.
+ * Since we didn't want to change VMWARE code, following set of typedefs
+ * and defines needed to compile these headers with QEMU introduced.
+ */

No need for this now.
You can export headers and put them under linux-headers.

[DF] Not sure it is possible because the header as-is is not stand-alone and won't compile [DF] without changes. We extracted definitions we use from their header and dropped unused [DF] and kernel-specific stuff.
[DF} Please, advise.

...

Quoted 1 line

+ if (VMXNET3_OM_TSO == s->offload_mode) {

Don't do Yoda style like this

[DF] "Yoda" style removed everywhere

Changes in V6:
Fixed most of problems pointed out by Michael S. Tsirkin
The only issue still open is creation of shared place
with generic network structures and functions. Currently
all generic network code introduced by VMXNET3 resides in vmxnet_utils.c/h files. It could be moved to some shared location however we believe it is a matter of separate refactoring as there are a lot of copy-pasted definitions in almost every device and code cleanup efforts requred in order to create truly shared codebase.

Implemented suggestions by Anthony Liguori

Fixed incorrect checksum caclulation for some packets in SW offloads mode

Changes in V5:
MSI-X save/load implemented in the device instead of pci bus as suggested by Michael S. Tsirkin

Patches regrouped as suggested by Paolo Bonzini

Changes in V4:
Fixed a few problems uncovered by NETIO test suit Assertion on failure to initialize MSI/MSI-X replaced with warning message and fallback to Legacy/MSI respectively

Various coding style adjustments and patch split-up as suggested by Anthony
Liguori

Live migration support added

Changes in V3:
Fixed crash when net device that is used as network fronted has no virtio HDR support.
Task offloads emulation for cases when net device that is used as network fronted has no virtio HDR support.

Changes in V2:
License text changed accoring to community suggestions Standard license header from GPLv2+ - licensed QEMU files used

Dmitry Fleytman (5):
Checksum-related utility functions
net: iovec checksum calculator
Common definitions for VMWARE devices
Packet abstraction for VMWARE network devices
VMXNET3 device implementation

 default-configs/pci.mak |    1 +
 hw/Makefile.objs        |    2 +
 hw/pci/pci.h            |    1 +
 hw/vmware_utils.h       |  143 +++
 hw/vmxnet3.c            | 2461 +++++++++++++++++++++++++++++++++++++++++++++++
 hw/vmxnet3.h            |  760 +++++++++++++++
 hw/vmxnet_debug.h       |  115 +++
 hw/vmxnet_rx_pkt.c      |  187 ++++
 hw/vmxnet_rx_pkt.h      |  174 ++++
 hw/vmxnet_tx_pkt.c      |  567 +++++++++++
 hw/vmxnet_tx_pkt.h      |  148 +++
 include/net/checksum.h  |   26 +-
 include/net/eth.h       |  347 +++++++
 net/Makefile.objs       |    1 +
 net/checksum.c          |   42 +-
 net/eth.c               |  217 +++++
 16 files changed, 5185 insertions(+), 7 deletions(-)
 create mode 100644 hw/vmware_utils.h
 create mode 100644 hw/vmxnet3.c
Quoted 12 lines
 create mode 100644 hw/vmxnet3.h
 create mode 100644 hw/vmxnet_debug.h
 create mode 100644 hw/vmxnet_rx_pkt.c
 create mode 100644 hw/vmxnet_rx_pkt.h
 create mode 100644 hw/vmxnet_tx_pkt.c
 create mode 100644 hw/vmxnet_tx_pkt.h
 create mode 100644 include/net/eth.h
 create mode 100644 net/eth.c
Signature · 1 line
1.8.1.2

Thanks, applied to my net tree:
https://github.com/stefanha/qemu/commits/net

Stefan