freebsd-dev/sys/dev/sfxge/common/hunt_tx.c
Andrew Rybchenko 9dd0e15fab sfxge: switch to TxQ creation specific flags
It is better do not mix TxQ creation and receive event flags since only
checksum flags are applicable to TxQ.
Also it will allow to add a new TxQ creation specific flags.

Reviewed by:    gnn
Sponsored by:   Solarflare Communications, Inc.
MFC after: 2    days
Differential Revision: https://reviews.freebsd.org/D4389
2015-12-07 06:04:24 +00:00

684 lines
15 KiB
C
Executable File

/*-
* Copyright (c) 2012-2015 Solarflare Communications Inc.
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions are met:
*
* 1. Redistributions of source code must retain the above copyright notice,
* this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright notice,
* this list of conditions and the following disclaimer in the documentation
* and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
* AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
* THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
* CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
* OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
* WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
* OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
* EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
*
* The views and conclusions contained in the software and documentation are
* those of the authors and should not be interpreted as representing official
* policies, either expressed or implied, of the FreeBSD Project.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "efsys.h"
#include "efx.h"
#include "efx_impl.h"
#if EFSYS_OPT_HUNTINGTON
#if EFSYS_OPT_QSTATS
#define EFX_TX_QSTAT_INCR(_etp, _stat) \
do { \
(_etp)->et_stat[_stat]++; \
_NOTE(CONSTANTCONDITION) \
} while (B_FALSE)
#else
#define EFX_TX_QSTAT_INCR(_etp, _stat)
#endif
static __checkReturn efx_rc_t
efx_mcdi_init_txq(
__in efx_nic_t *enp,
__in uint32_t size,
__in uint32_t target_evq,
__in uint32_t label,
__in uint32_t instance,
__in uint16_t flags,
__in efsys_mem_t *esmp)
{
efx_mcdi_req_t req;
uint8_t payload[MAX(MC_CMD_INIT_TXQ_IN_LEN(EFX_TXQ_MAX_BUFS),
MC_CMD_INIT_TXQ_OUT_LEN)];
efx_qword_t *dma_addr;
uint64_t addr;
int npages;
int i;
efx_rc_t rc;
EFSYS_ASSERT(EFX_TXQ_MAX_BUFS >=
EFX_TXQ_NBUFS(EFX_TXQ_MAXNDESCS(&enp->en_nic_cfg)));
npages = EFX_TXQ_NBUFS(size);
if (npages > MC_CMD_INIT_TXQ_IN_DMA_ADDR_MAXNUM) {
rc = EINVAL;
goto fail1;
}
(void) memset(payload, 0, sizeof (payload));
req.emr_cmd = MC_CMD_INIT_TXQ;
req.emr_in_buf = payload;
req.emr_in_length = MC_CMD_INIT_TXQ_IN_LEN(npages);
req.emr_out_buf = payload;
req.emr_out_length = MC_CMD_INIT_TXQ_OUT_LEN;
MCDI_IN_SET_DWORD(req, INIT_TXQ_IN_SIZE, size);
MCDI_IN_SET_DWORD(req, INIT_TXQ_IN_TARGET_EVQ, target_evq);
MCDI_IN_SET_DWORD(req, INIT_TXQ_IN_LABEL, label);
MCDI_IN_SET_DWORD(req, INIT_TXQ_IN_INSTANCE, instance);
MCDI_IN_POPULATE_DWORD_6(req, INIT_TXQ_IN_FLAGS,
INIT_TXQ_IN_FLAG_BUFF_MODE, 0,
INIT_TXQ_IN_FLAG_IP_CSUM_DIS,
(flags & EFX_TXQ_CKSUM_IPV4) ? 0 : 1,
INIT_TXQ_IN_FLAG_TCP_CSUM_DIS,
(flags & EFX_TXQ_CKSUM_TCPUDP) ? 0 : 1,
INIT_TXQ_IN_FLAG_TCP_UDP_ONLY, 0,
INIT_TXQ_IN_CRC_MODE, 0,
INIT_TXQ_IN_FLAG_TIMESTAMP, 0);
MCDI_IN_SET_DWORD(req, INIT_TXQ_IN_OWNER_ID, 0);
MCDI_IN_SET_DWORD(req, INIT_TXQ_IN_PORT_ID, EVB_PORT_ID_ASSIGNED);
dma_addr = MCDI_IN2(req, efx_qword_t, INIT_TXQ_IN_DMA_ADDR);
addr = EFSYS_MEM_ADDR(esmp);
for (i = 0; i < npages; i++) {
EFX_POPULATE_QWORD_2(*dma_addr,
EFX_DWORD_1, (uint32_t)(addr >> 32),
EFX_DWORD_0, (uint32_t)(addr & 0xffffffff));
dma_addr++;
addr += EFX_BUF_SIZE;
}
efx_mcdi_execute(enp, &req);
if (req.emr_rc != 0) {
rc = req.emr_rc;
goto fail2;
}
return (0);
fail2:
EFSYS_PROBE(fail2);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
static __checkReturn efx_rc_t
efx_mcdi_fini_txq(
__in efx_nic_t *enp,
__in uint32_t instance)
{
efx_mcdi_req_t req;
uint8_t payload[MAX(MC_CMD_FINI_TXQ_IN_LEN,
MC_CMD_FINI_TXQ_OUT_LEN)];
efx_rc_t rc;
(void) memset(payload, 0, sizeof (payload));
req.emr_cmd = MC_CMD_FINI_TXQ;
req.emr_in_buf = payload;
req.emr_in_length = MC_CMD_FINI_TXQ_IN_LEN;
req.emr_out_buf = payload;
req.emr_out_length = MC_CMD_FINI_TXQ_OUT_LEN;
MCDI_IN_SET_DWORD(req, FINI_TXQ_IN_INSTANCE, instance);
efx_mcdi_execute(enp, &req);
if ((req.emr_rc != 0) && (req.emr_rc != MC_CMD_ERR_EALREADY)) {
rc = req.emr_rc;
goto fail1;
}
return (0);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
__checkReturn efx_rc_t
hunt_tx_init(
__in efx_nic_t *enp)
{
_NOTE(ARGUNUSED(enp))
return (0);
}
void
hunt_tx_fini(
__in efx_nic_t *enp)
{
_NOTE(ARGUNUSED(enp))
}
__checkReturn efx_rc_t
hunt_tx_qcreate(
__in efx_nic_t *enp,
__in unsigned int index,
__in unsigned int label,
__in efsys_mem_t *esmp,
__in size_t n,
__in uint32_t id,
__in uint16_t flags,
__in efx_evq_t *eep,
__in efx_txq_t *etp,
__out unsigned int *addedp)
{
efx_qword_t desc;
efx_rc_t rc;
if ((rc = efx_mcdi_init_txq(enp, n, eep->ee_index, label, index, flags,
esmp)) != 0)
goto fail1;
/*
* A previous user of this TX queue may have written a descriptor to the
* TX push collector, but not pushed the doorbell (e.g. after a crash).
* The next doorbell write would then push the stale descriptor.
*
* Ensure the (per network port) TX push collector is cleared by writing
* a no-op TX option descriptor. See bug29981 for details.
*/
*addedp = 1;
EFX_POPULATE_QWORD_4(desc,
ESF_DZ_TX_DESC_IS_OPT, 1,
ESF_DZ_TX_OPTION_TYPE, ESE_DZ_TX_OPTION_DESC_CRC_CSUM,
ESF_DZ_TX_OPTION_UDP_TCP_CSUM,
(flags & EFX_TXQ_CKSUM_TCPUDP) ? 1 : 0,
ESF_DZ_TX_OPTION_IP_CSUM,
(flags & EFX_TXQ_CKSUM_IPV4) ? 1 : 0);
EFSYS_MEM_WRITEQ(etp->et_esmp, 0, &desc);
hunt_tx_qpush(etp, *addedp, 0);
return (0);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
void
hunt_tx_qdestroy(
__in efx_txq_t *etp)
{
/* FIXME */
_NOTE(ARGUNUSED(etp))
/* FIXME */
}
__checkReturn efx_rc_t
hunt_tx_qpio_enable(
__in efx_txq_t *etp)
{
efx_nic_t *enp = etp->et_enp;
efx_piobuf_handle_t handle;
efx_rc_t rc;
if (etp->et_pio_size != 0) {
rc = EALREADY;
goto fail1;
}
/* Sub-allocate a PIO block from a piobuf */
if ((rc = hunt_nic_pio_alloc(enp,
&etp->et_pio_bufnum,
&handle,
&etp->et_pio_blknum,
&etp->et_pio_offset,
&etp->et_pio_size)) != 0) {
goto fail2;
}
EFSYS_ASSERT3U(etp->et_pio_size, !=, 0);
/* Link the piobuf to this TXQ */
if ((rc = hunt_nic_pio_link(enp, etp->et_index, handle)) != 0) {
goto fail3;
}
/*
* et_pio_offset is the offset of the sub-allocated block within the
* hardware PIO buffer. It is used as the buffer address in the PIO
* option descriptor.
*
* et_pio_write_offset is the offset of the sub-allocated block from the
* start of the write-combined memory mapping, and is used for writing
* data into the PIO buffer.
*/
etp->et_pio_write_offset =
(etp->et_pio_bufnum * ER_DZ_TX_PIOBUF_STEP) +
ER_DZ_TX_PIOBUF_OFST + etp->et_pio_offset;
return (0);
fail3:
EFSYS_PROBE(fail3);
hunt_nic_pio_free(enp, etp->et_pio_bufnum, etp->et_pio_blknum);
etp->et_pio_size = 0;
fail2:
EFSYS_PROBE(fail2);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
void
hunt_tx_qpio_disable(
__in efx_txq_t *etp)
{
efx_nic_t *enp = etp->et_enp;
if (etp->et_pio_size != 0) {
/* Unlink the piobuf from this TXQ */
hunt_nic_pio_unlink(enp, etp->et_index);
/* Free the sub-allocated PIO block */
hunt_nic_pio_free(enp, etp->et_pio_bufnum, etp->et_pio_blknum);
etp->et_pio_size = 0;
etp->et_pio_write_offset = 0;
}
}
__checkReturn efx_rc_t
hunt_tx_qpio_write(
__in efx_txq_t *etp,
__in_ecount(length) uint8_t *buffer,
__in size_t length,
__in size_t offset)
{
efx_nic_t *enp = etp->et_enp;
efsys_bar_t *esbp = enp->en_esbp;
uint32_t write_offset;
uint32_t write_offset_limit;
efx_qword_t *eqp;
efx_rc_t rc;
EFSYS_ASSERT(length % sizeof (efx_qword_t) == 0);
if (etp->et_pio_size == 0) {
rc = ENOENT;
goto fail1;
}
if (offset + length > etp->et_pio_size) {
rc = ENOSPC;
goto fail2;
}
/*
* Writes to PIO buffers must be 64 bit aligned, and multiples of
* 64 bits.
*/
write_offset = etp->et_pio_write_offset + offset;
write_offset_limit = write_offset + length;
eqp = (efx_qword_t *)buffer;
while (write_offset < write_offset_limit) {
EFSYS_BAR_WC_WRITEQ(esbp, write_offset, eqp);
eqp++;
write_offset += sizeof (efx_qword_t);
}
return (0);
fail2:
EFSYS_PROBE(fail2);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
__checkReturn efx_rc_t
hunt_tx_qpio_post(
__in efx_txq_t *etp,
__in size_t pkt_length,
__in unsigned int completed,
__inout unsigned int *addedp)
{
efx_qword_t pio_desc;
unsigned int id;
size_t offset;
unsigned int added = *addedp;
efx_rc_t rc;
if (added - completed + 1 > EFX_TXQ_LIMIT(etp->et_mask + 1)) {
rc = ENOSPC;
goto fail1;
}
if (etp->et_pio_size == 0) {
rc = ENOENT;
goto fail2;
}
id = added++ & etp->et_mask;
offset = id * sizeof (efx_qword_t);
EFSYS_PROBE4(tx_pio_post, unsigned int, etp->et_index,
unsigned int, id, uint32_t, etp->et_pio_offset,
size_t, pkt_length);
EFX_POPULATE_QWORD_5(pio_desc,
ESF_DZ_TX_DESC_IS_OPT, 1,
ESF_DZ_TX_OPTION_TYPE, 1,
ESF_DZ_TX_PIO_CONT, 0,
ESF_DZ_TX_PIO_BYTE_CNT, pkt_length,
ESF_DZ_TX_PIO_BUF_ADDR, etp->et_pio_offset);
EFSYS_MEM_WRITEQ(etp->et_esmp, offset, &pio_desc);
EFX_TX_QSTAT_INCR(etp, TX_POST_PIO);
*addedp = added;
return (0);
fail2:
EFSYS_PROBE(fail2);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
__checkReturn efx_rc_t
hunt_tx_qpost(
__in efx_txq_t *etp,
__in_ecount(n) efx_buffer_t *eb,
__in unsigned int n,
__in unsigned int completed,
__inout unsigned int *addedp)
{
unsigned int added = *addedp;
unsigned int i;
efx_rc_t rc;
if (added - completed + n > EFX_TXQ_LIMIT(etp->et_mask + 1)) {
rc = ENOSPC;
goto fail1;
}
for (i = 0; i < n; i++) {
efx_buffer_t *ebp = &eb[i];
efsys_dma_addr_t addr = ebp->eb_addr;
size_t size = ebp->eb_size;
boolean_t eop = ebp->eb_eop;
unsigned int id;
size_t offset;
efx_qword_t qword;
/* Fragments must not span 4k boundaries. */
EFSYS_ASSERT(P2ROUNDUP(addr + 1, 4096) >= (addr + size));
id = added++ & etp->et_mask;
offset = id * sizeof (efx_qword_t);
EFSYS_PROBE5(tx_post, unsigned int, etp->et_index,
unsigned int, id, efsys_dma_addr_t, addr,
size_t, size, boolean_t, eop);
EFX_POPULATE_QWORD_5(qword,
ESF_DZ_TX_KER_TYPE, 0,
ESF_DZ_TX_KER_CONT, (eop) ? 0 : 1,
ESF_DZ_TX_KER_BYTE_CNT, (uint32_t)(size),
ESF_DZ_TX_KER_BUF_ADDR_DW0, (uint32_t)(addr & 0xffffffff),
ESF_DZ_TX_KER_BUF_ADDR_DW1, (uint32_t)(addr >> 32));
EFSYS_MEM_WRITEQ(etp->et_esmp, offset, &qword);
}
EFX_TX_QSTAT_INCR(etp, TX_POST);
*addedp = added;
return (0);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
/*
* This improves performance by pushing a TX descriptor at the same time as the
* doorbell. The descriptor must be added to the TXQ, so that can be used if the
* hardware decides not to use the pushed descriptor.
*/
void
hunt_tx_qpush(
__in efx_txq_t *etp,
__in unsigned int added,
__in unsigned int pushed)
{
efx_nic_t *enp = etp->et_enp;
unsigned int wptr;
unsigned int id;
size_t offset;
efx_qword_t desc;
efx_oword_t oword;
wptr = added & etp->et_mask;
id = pushed & etp->et_mask;
offset = id * sizeof (efx_qword_t);
EFSYS_MEM_READQ(etp->et_esmp, offset, &desc);
EFX_POPULATE_OWORD_3(oword,
ERF_DZ_TX_DESC_WPTR, wptr,
ERF_DZ_TX_DESC_HWORD, EFX_QWORD_FIELD(desc, EFX_DWORD_1),
ERF_DZ_TX_DESC_LWORD, EFX_QWORD_FIELD(desc, EFX_DWORD_0));
/* Guarantee ordering of memory (descriptors) and PIO (doorbell) */
EFX_DMA_SYNC_QUEUE_FOR_DEVICE(etp->et_esmp, etp->et_mask + 1, wptr, id);
EFSYS_PIO_WRITE_BARRIER();
EFX_BAR_TBL_DOORBELL_WRITEO(enp, ER_DZ_TX_DESC_UPD_REG, etp->et_index,
&oword);
}
__checkReturn efx_rc_t
hunt_tx_qdesc_post(
__in efx_txq_t *etp,
__in_ecount(n) efx_desc_t *ed,
__in unsigned int n,
__in unsigned int completed,
__inout unsigned int *addedp)
{
unsigned int added = *addedp;
unsigned int i;
efx_rc_t rc;
if (added - completed + n > EFX_TXQ_LIMIT(etp->et_mask + 1)) {
rc = ENOSPC;
goto fail1;
}
for (i = 0; i < n; i++) {
efx_desc_t *edp = &ed[i];
unsigned int id;
size_t offset;
id = added++ & etp->et_mask;
offset = id * sizeof (efx_desc_t);
EFSYS_MEM_WRITEQ(etp->et_esmp, offset, &edp->ed_eq);
}
EFSYS_PROBE3(tx_desc_post, unsigned int, etp->et_index,
unsigned int, added, unsigned int, n);
EFX_TX_QSTAT_INCR(etp, TX_POST);
*addedp = added;
return (0);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
void
hunt_tx_qdesc_dma_create(
__in efx_txq_t *etp,
__in efsys_dma_addr_t addr,
__in size_t size,
__in boolean_t eop,
__out efx_desc_t *edp)
{
/* Fragments must not span 4k boundaries. */
EFSYS_ASSERT(P2ROUNDUP(addr + 1, 4096) >= addr + size);
EFSYS_PROBE4(tx_desc_dma_create, unsigned int, etp->et_index,
efsys_dma_addr_t, addr,
size_t, size, boolean_t, eop);
EFX_POPULATE_QWORD_5(edp->ed_eq,
ESF_DZ_TX_KER_TYPE, 0,
ESF_DZ_TX_KER_CONT, (eop) ? 0 : 1,
ESF_DZ_TX_KER_BYTE_CNT, (uint32_t)(size),
ESF_DZ_TX_KER_BUF_ADDR_DW0, (uint32_t)(addr & 0xffffffff),
ESF_DZ_TX_KER_BUF_ADDR_DW1, (uint32_t)(addr >> 32));
}
void
hunt_tx_qdesc_tso_create(
__in efx_txq_t *etp,
__in uint16_t ipv4_id,
__in uint32_t tcp_seq,
__in uint8_t tcp_flags,
__out efx_desc_t *edp)
{
EFSYS_PROBE4(tx_desc_tso_create, unsigned int, etp->et_index,
uint16_t, ipv4_id, uint32_t, tcp_seq,
uint8_t, tcp_flags);
EFX_POPULATE_QWORD_5(edp->ed_eq,
ESF_DZ_TX_DESC_IS_OPT, 1,
ESF_DZ_TX_OPTION_TYPE,
ESE_DZ_TX_OPTION_DESC_TSO,
ESF_DZ_TX_TSO_TCP_FLAGS, tcp_flags,
ESF_DZ_TX_TSO_IP_ID, ipv4_id,
ESF_DZ_TX_TSO_TCP_SEQNO, tcp_seq);
}
void
hunt_tx_qdesc_vlantci_create(
__in efx_txq_t *etp,
__in uint16_t tci,
__out efx_desc_t *edp)
{
EFSYS_PROBE2(tx_desc_vlantci_create, unsigned int, etp->et_index,
uint16_t, tci);
EFX_POPULATE_QWORD_4(edp->ed_eq,
ESF_DZ_TX_DESC_IS_OPT, 1,
ESF_DZ_TX_OPTION_TYPE,
ESE_DZ_TX_OPTION_DESC_VLAN,
ESF_DZ_TX_VLAN_OP, tci ? 1 : 0,
ESF_DZ_TX_VLAN_TAG1, tci);
}
__checkReturn efx_rc_t
hunt_tx_qpace(
__in efx_txq_t *etp,
__in unsigned int ns)
{
efx_rc_t rc;
/* FIXME */
_NOTE(ARGUNUSED(etp, ns))
if (B_FALSE) {
rc = ENOTSUP;
goto fail1;
}
/* FIXME */
return (0);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
__checkReturn efx_rc_t
hunt_tx_qflush(
__in efx_txq_t *etp)
{
efx_nic_t *enp = etp->et_enp;
efx_rc_t rc;
if ((rc = efx_mcdi_fini_txq(enp, etp->et_index)) != 0)
goto fail1;
return (0);
fail1:
EFSYS_PROBE1(fail1, efx_rc_t, rc);
return (rc);
}
void
hunt_tx_qenable(
__in efx_txq_t *etp)
{
/* FIXME */
_NOTE(ARGUNUSED(etp))
/* FIXME */
}
#if EFSYS_OPT_QSTATS
void
hunt_tx_qstats_update(
__in efx_txq_t *etp,
__inout_ecount(TX_NQSTATS) efsys_stat_t *stat)
{
/*
* TBD: Consider a common Siena/Huntington function. The code is
* essentially identical.
*/
unsigned int id;
for (id = 0; id < TX_NQSTATS; id++) {
efsys_stat_t *essp = &stat[id];
EFSYS_STAT_INCR(essp, etp->et_stat[id]);
etp->et_stat[id] = 0;
}
}
#endif /* EFSYS_OPT_QSTATS */
#endif /* EFSYS_OPT_HUNTINGTON */