* [PATCH v7 net-next 09/19] ionic: Add notifyq support
From: Shannon Nelson @ 2019-09-03 22:28 UTC (permalink / raw)
To: snelson, netdev, davem
In-Reply-To: <20190903222821.46161-1-snelson@pensando.io>
The AdminQ is fine for sending messages and requests to the NIC,
but we also need to have events published from the NIC to the
driver. The NotifyQ handles this for us, using the same interrupt
as AdminQ.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
---
.../ethernet/pensando/ionic/ionic_debugfs.c | 15 +-
.../net/ethernet/pensando/ionic/ionic_lif.c | 169 +++++++++++++++++-
.../net/ethernet/pensando/ionic/ionic_lif.h | 4 +
3 files changed, 186 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c
index d02f81a2b466..7afc4a365b75 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c
@@ -112,7 +112,7 @@ static const struct debugfs_reg32 intr_ctrl_regs[] = {
void ionic_debugfs_add_qcq(struct ionic_lif *lif, struct ionic_qcq *qcq)
{
- struct dentry *q_dentry, *cq_dentry, *intr_dentry;
+ struct dentry *q_dentry, *cq_dentry, *intr_dentry, *stats_dentry;
struct ionic_dev *idev = &lif->ionic->idev;
struct debugfs_regset32 *intr_ctrl_regset;
struct ionic_intr_info *intr = &qcq->intr;
@@ -201,6 +201,19 @@ void ionic_debugfs_add_qcq(struct ionic_lif *lif, struct ionic_qcq *qcq)
debugfs_create_regset32("intr_ctrl", 0400, intr_dentry,
intr_ctrl_regset);
}
+
+ if (qcq->flags & IONIC_QCQ_F_NOTIFYQ) {
+ stats_dentry = debugfs_create_dir("notifyblock", qcq->dentry);
+
+ debugfs_create_u64("eid", 0400, stats_dentry,
+ (u64 *)&lif->info->status.eid);
+ debugfs_create_u16("link_status", 0400, stats_dentry,
+ (u16 *)&lif->info->status.link_status);
+ debugfs_create_u32("link_speed", 0400, stats_dentry,
+ (u32 *)&lif->info->status.link_speed);
+ debugfs_create_u16("link_down_count", 0400, stats_dentry,
+ (u16 *)&lif->info->status.link_down_count);
+ }
}
static int netdev_show(struct seq_file *seq, void *v)
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.c b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
index 6fc2b4bb4c58..6096ed813e20 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_lif.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
@@ -110,12 +110,29 @@ static void ionic_qcq_free(struct ionic_lif *lif, struct ionic_qcq *qcq)
static void ionic_qcqs_free(struct ionic_lif *lif)
{
+ if (lif->notifyqcq) {
+ ionic_qcq_free(lif, lif->notifyqcq);
+ lif->notifyqcq = NULL;
+ }
+
if (lif->adminqcq) {
ionic_qcq_free(lif, lif->adminqcq);
lif->adminqcq = NULL;
}
}
+static void ionic_link_qcq_interrupts(struct ionic_qcq *src_qcq,
+ struct ionic_qcq *n_qcq)
+{
+ if (WARN_ON(n_qcq->flags & IONIC_QCQ_F_INTR)) {
+ ionic_intr_free(n_qcq->cq.lif, n_qcq->intr.index);
+ n_qcq->flags &= ~IONIC_QCQ_F_INTR;
+ }
+
+ n_qcq->intr.vector = src_qcq->intr.vector;
+ n_qcq->intr.index = src_qcq->intr.index;
+}
+
static int ionic_qcq_alloc(struct ionic_lif *lif, unsigned int type,
unsigned int index,
const char *name, unsigned int flags,
@@ -269,7 +286,91 @@ static int ionic_qcqs_alloc(struct ionic_lif *lif)
if (err)
return err;
+ if (lif->ionic->nnqs_per_lif) {
+ flags = IONIC_QCQ_F_NOTIFYQ;
+ err = ionic_qcq_alloc(lif, IONIC_QTYPE_NOTIFYQ, 0, "notifyq",
+ flags, IONIC_NOTIFYQ_LENGTH,
+ sizeof(struct ionic_notifyq_cmd),
+ sizeof(union ionic_notifyq_comp),
+ 0, lif->kern_pid, &lif->notifyqcq);
+ if (err)
+ goto err_out_free_adminqcq;
+
+ /* Let the notifyq ride on the adminq interrupt */
+ ionic_link_qcq_interrupts(lif->adminqcq, lif->notifyqcq);
+ }
+
return 0;
+
+err_out_free_adminqcq:
+ ionic_qcq_free(lif, lif->adminqcq);
+ lif->adminqcq = NULL;
+
+ return err;
+}
+
+static bool ionic_notifyq_service(struct ionic_cq *cq,
+ struct ionic_cq_info *cq_info)
+{
+ union ionic_notifyq_comp *comp = cq_info->cq_desc;
+ struct net_device *netdev;
+ struct ionic_queue *q;
+ struct ionic_lif *lif;
+ u64 eid;
+
+ q = cq->bound_q;
+ lif = q->info[0].cb_arg;
+ netdev = lif->netdev;
+ eid = le64_to_cpu(comp->event.eid);
+
+ /* Have we run out of new completions to process? */
+ if (eid <= lif->last_eid)
+ return false;
+
+ lif->last_eid = eid;
+
+ dev_dbg(lif->ionic->dev, "notifyq event:\n");
+ dynamic_hex_dump("event ", DUMP_PREFIX_OFFSET, 16, 1,
+ comp, sizeof(*comp), true);
+
+ switch (le16_to_cpu(comp->event.ecode)) {
+ case IONIC_EVENT_LINK_CHANGE:
+ netdev_info(netdev, "Notifyq IONIC_EVENT_LINK_CHANGE eid=%lld\n",
+ eid);
+ netdev_info(netdev,
+ " link_status=%d link_speed=%d\n",
+ le16_to_cpu(comp->link_change.link_status),
+ le32_to_cpu(comp->link_change.link_speed));
+ break;
+ case IONIC_EVENT_RESET:
+ netdev_info(netdev, "Notifyq IONIC_EVENT_RESET eid=%lld\n",
+ eid);
+ netdev_info(netdev, " reset_code=%d state=%d\n",
+ comp->reset.reset_code,
+ comp->reset.state);
+ break;
+ default:
+ netdev_warn(netdev, "Notifyq unknown event ecode=%d eid=%lld\n",
+ comp->event.ecode, eid);
+ break;
+ }
+
+ return true;
+}
+
+static int ionic_notifyq_clean(struct ionic_lif *lif, int budget)
+{
+ struct ionic_dev *idev = &lif->ionic->idev;
+ struct ionic_cq *cq = &lif->notifyqcq->cq;
+ u32 work_done;
+
+ work_done = ionic_cq_service(cq, budget, ionic_notifyq_service,
+ NULL, NULL);
+ if (work_done)
+ ionic_intr_credits(idev->intr_ctrl, cq->bound_intr->index,
+ work_done, IONIC_INTR_CRED_RESET_COALESCE);
+
+ return work_done;
}
static bool ionic_adminq_service(struct ionic_cq *cq,
@@ -287,7 +388,15 @@ static bool ionic_adminq_service(struct ionic_cq *cq,
static int ionic_adminq_napi(struct napi_struct *napi, int budget)
{
- return ionic_napi(napi, budget, ionic_adminq_service, NULL, NULL);
+ struct ionic_lif *lif = napi_to_cq(napi)->lif;
+ int n_work = 0;
+ int a_work = 0;
+
+ if (likely(lif->notifyqcq && lif->notifyqcq->flags & IONIC_QCQ_F_INITED))
+ n_work = ionic_notifyq_clean(lif, budget);
+ a_work = ionic_napi(napi, budget, ionic_adminq_service, NULL, NULL);
+
+ return max(n_work, a_work);
}
static struct ionic_lif *ionic_lif_alloc(struct ionic *ionic, unsigned int index)
@@ -417,6 +526,7 @@ static void ionic_lif_deinit(struct ionic_lif *lif)
clear_bit(IONIC_LIF_INITED, lif->state);
napi_disable(&lif->adminqcq->napi);
+ ionic_lif_qcq_deinit(lif, lif->notifyqcq);
ionic_lif_qcq_deinit(lif, lif->adminqcq);
ionic_lif_reset(lif);
@@ -486,6 +596,55 @@ static int ionic_lif_adminq_init(struct ionic_lif *lif)
return 0;
}
+static int ionic_lif_notifyq_init(struct ionic_lif *lif)
+{
+ struct ionic_qcq *qcq = lif->notifyqcq;
+ struct device *dev = lif->ionic->dev;
+ struct ionic_queue *q = &qcq->q;
+ int err;
+
+ struct ionic_admin_ctx ctx = {
+ .work = COMPLETION_INITIALIZER_ONSTACK(ctx.work),
+ .cmd.q_init = {
+ .opcode = IONIC_CMD_Q_INIT,
+ .lif_index = cpu_to_le16(lif->index),
+ .type = q->type,
+ .index = cpu_to_le32(q->index),
+ .flags = cpu_to_le16(IONIC_QINIT_F_IRQ |
+ IONIC_QINIT_F_ENA),
+ .intr_index = cpu_to_le16(lif->adminqcq->intr.index),
+ .pid = cpu_to_le16(q->pid),
+ .ring_size = ilog2(q->num_descs),
+ .ring_base = cpu_to_le64(q->base_pa),
+ }
+ };
+
+ dev_dbg(dev, "notifyq_init.pid %d\n", ctx.cmd.q_init.pid);
+ dev_dbg(dev, "notifyq_init.index %d\n", ctx.cmd.q_init.index);
+ dev_dbg(dev, "notifyq_init.ring_base 0x%llx\n", ctx.cmd.q_init.ring_base);
+ dev_dbg(dev, "notifyq_init.ring_size %d\n", ctx.cmd.q_init.ring_size);
+
+ err = ionic_adminq_post_wait(lif, &ctx);
+ if (err)
+ return err;
+
+ q->hw_type = ctx.comp.q_init.hw_type;
+ q->hw_index = le32_to_cpu(ctx.comp.q_init.hw_index);
+ q->dbval = IONIC_DBELL_QID(q->hw_index);
+
+ dev_dbg(dev, "notifyq->hw_type %d\n", q->hw_type);
+ dev_dbg(dev, "notifyq->hw_index %d\n", q->hw_index);
+
+ /* preset the callback info */
+ q->info[0].cb_arg = lif;
+
+ qcq->flags |= IONIC_QCQ_F_INITED;
+
+ ionic_debugfs_add_qcq(lif, qcq);
+
+ return 0;
+}
+
static int ionic_lif_init(struct ionic_lif *lif)
{
struct ionic_dev *idev = &lif->ionic->idev;
@@ -535,10 +694,18 @@ static int ionic_lif_init(struct ionic_lif *lif)
if (err)
goto err_out_adminq_deinit;
+ if (lif->ionic->nnqs_per_lif) {
+ err = ionic_lif_notifyq_init(lif);
+ if (err)
+ goto err_out_notifyq_deinit;
+ }
+
set_bit(IONIC_LIF_INITED, lif->state);
return 0;
+err_out_notifyq_deinit:
+ ionic_lif_qcq_deinit(lif, lif->notifyqcq);
err_out_adminq_deinit:
ionic_lif_qcq_deinit(lif, lif->adminqcq);
ionic_lif_reset(lif);
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.h b/drivers/net/ethernet/pensando/ionic/ionic_lif.h
index d897048ae77f..7bbe818893b7 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_lif.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.h
@@ -7,6 +7,7 @@
#include <linux/pci.h>
#define IONIC_ADMINQ_LENGTH 16 /* must be a power of two */
+#define IONIC_NOTIFYQ_LENGTH 64 /* must be a power of two */
#define IONIC_MAX_NUM_NAPI_CNTR (NAPI_POLL_WEIGHT + 1)
#define IONIC_MAX_NUM_SG_CNTR (IONIC_TX_MAX_SG_ELEMS + 1)
@@ -24,6 +25,7 @@ struct ionic_rx_stats {
#define IONIC_QCQ_F_INITED BIT(0)
#define IONIC_QCQ_F_SG BIT(1)
#define IONIC_QCQ_F_INTR BIT(2)
+#define IONIC_QCQ_F_NOTIFYQ BIT(5)
struct ionic_napi_stats {
u64 poll_count;
@@ -76,6 +78,8 @@ struct ionic_lif {
u64 __iomem *kern_dbpage;
spinlock_t adminq_lock; /* lock for AdminQ operations */
struct ionic_qcq *adminqcq;
+ struct ionic_qcq *notifyqcq;
+ u64 last_eid;
unsigned int neqs;
unsigned int nxqs;
--
2.17.1
^ permalink raw reply related
* [PATCH v7 net-next 05/19] ionic: Add basic lif support
From: Shannon Nelson @ 2019-09-03 22:28 UTC (permalink / raw)
To: snelson, netdev, davem
In-Reply-To: <20190903222821.46161-1-snelson@pensando.io>
The LIF is the Logical Interface, which represents the external
connections. The NIC can multiplex many LIFs to a single port,
but in most setups, LIF0 is the primary control for the port.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
---
drivers/net/ethernet/pensando/ionic/Makefile | 2 +-
drivers/net/ethernet/pensando/ionic/ionic.h | 7 +
.../net/ethernet/pensando/ionic/ionic_bus.h | 2 +
.../ethernet/pensando/ionic/ionic_bus_pci.c | 47 +++
.../ethernet/pensando/ionic/ionic_debugfs.c | 36 +++
.../ethernet/pensando/ionic/ionic_debugfs.h | 6 +
.../net/ethernet/pensando/ionic/ionic_dev.c | 34 ++
.../net/ethernet/pensando/ionic/ionic_dev.h | 7 +
.../net/ethernet/pensando/ionic/ionic_lif.c | 291 ++++++++++++++++++
.../net/ethernet/pensando/ionic/ionic_lif.h | 45 +++
.../net/ethernet/pensando/ionic/ionic_main.c | 1 +
11 files changed, 477 insertions(+), 1 deletion(-)
create mode 100644 drivers/net/ethernet/pensando/ionic/ionic_lif.c
create mode 100644 drivers/net/ethernet/pensando/ionic/ionic_lif.h
diff --git a/drivers/net/ethernet/pensando/ionic/Makefile b/drivers/net/ethernet/pensando/ionic/Makefile
index a23d58519c63..215ed1ea44df 100644
--- a/drivers/net/ethernet/pensando/ionic/Makefile
+++ b/drivers/net/ethernet/pensando/ionic/Makefile
@@ -4,4 +4,4 @@
obj-$(CONFIG_IONIC) := ionic.o
ionic-y := ionic_main.o ionic_bus_pci.o ionic_devlink.o ionic_dev.o \
- ionic_debugfs.o
+ ionic_debugfs.o ionic_lif.o
diff --git a/drivers/net/ethernet/pensando/ionic/ionic.h b/drivers/net/ethernet/pensando/ionic/ionic.h
index 4960effd2bcc..723b2ba6874e 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic.h
@@ -32,6 +32,13 @@ struct ionic {
struct ionic_dev_bar bars[IONIC_BARS_MAX];
unsigned int num_bars;
struct ionic_identity ident;
+ struct list_head lifs;
+ unsigned int nnqs_per_lif;
+ unsigned int neqs_per_lif;
+ unsigned int ntxqs_per_lif;
+ unsigned int nrxqs_per_lif;
+ DECLARE_BITMAP(lifbits, IONIC_LIFS_MAX);
+ unsigned int nintrs;
};
int ionic_dev_cmd_wait(struct ionic *ionic, unsigned long max_wait);
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_bus.h b/drivers/net/ethernet/pensando/ionic/ionic_bus.h
index 24b4c01ec03f..3b1e2d0ebf8f 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_bus.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_bus.h
@@ -5,6 +5,8 @@
#define _IONIC_BUS_H_
const char *ionic_bus_info(struct ionic *ionic);
+int ionic_bus_alloc_irq_vectors(struct ionic *ionic, unsigned int nintrs);
+void ionic_bus_free_irq_vectors(struct ionic *ionic);
int ionic_bus_register_driver(void);
void ionic_bus_unregister_driver(void);
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c b/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c
index 804dd43e92a6..f0e0daee45bc 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c
@@ -8,6 +8,7 @@
#include "ionic.h"
#include "ionic_bus.h"
+#include "ionic_lif.h"
#include "ionic_debugfs.h"
/* Supported devices */
@@ -23,6 +24,17 @@ const char *ionic_bus_info(struct ionic *ionic)
return pci_name(ionic->pdev);
}
+int ionic_bus_alloc_irq_vectors(struct ionic *ionic, unsigned int nintrs)
+{
+ return pci_alloc_irq_vectors(ionic->pdev, nintrs, nintrs,
+ PCI_IRQ_MSIX);
+}
+
+void ionic_bus_free_irq_vectors(struct ionic *ionic)
+{
+ pci_free_irq_vectors(ionic->pdev);
+}
+
static int ionic_map_bars(struct ionic *ionic)
{
struct pci_dev *pdev = ionic->pdev;
@@ -151,12 +163,44 @@ static int ionic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
goto err_out_reset;
}
+ /* Configure LIFs */
+ err = ionic_lif_identify(ionic, IONIC_LIF_TYPE_CLASSIC,
+ &ionic->ident.lif);
+ if (err) {
+ dev_err(dev, "Cannot identify LIFs: %d, aborting\n", err);
+ goto err_out_port_reset;
+ }
+
+ err = ionic_lifs_size(ionic);
+ if (err) {
+ dev_err(dev, "Cannot size LIFs: %d, aborting\n", err);
+ goto err_out_port_reset;
+ }
+
+ err = ionic_lifs_alloc(ionic);
+ if (err) {
+ dev_err(dev, "Cannot allocate LIFs: %d, aborting\n", err);
+ goto err_out_free_irqs;
+ }
+
+ err = ionic_lifs_init(ionic);
+ if (err) {
+ dev_err(dev, "Cannot init LIFs: %d, aborting\n", err);
+ goto err_out_free_lifs;
+ }
+
err = ionic_devlink_register(ionic);
if (err)
dev_err(dev, "Cannot register devlink: %d\n", err);
return 0;
+err_out_free_lifs:
+ ionic_lifs_free(ionic);
+err_out_free_irqs:
+ ionic_bus_free_irq_vectors(ionic);
+err_out_port_reset:
+ ionic_port_reset(ionic);
err_out_reset:
ionic_reset(ionic);
err_out_teardown:
@@ -185,6 +229,9 @@ static void ionic_remove(struct pci_dev *pdev)
return;
ionic_devlink_unregister(ionic);
+ ionic_lifs_deinit(ionic);
+ ionic_lifs_free(ionic);
+ ionic_bus_free_irq_vectors(ionic);
ionic_port_reset(ionic);
ionic_reset(ionic);
ionic_dev_teardown(ionic);
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c
index c57a7e4f35d1..840b3da5da3e 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c
@@ -1,10 +1,12 @@
// SPDX-License-Identifier: GPL-2.0
/* Copyright(c) 2017 - 2019 Pensando Systems, Inc */
+#include <linux/pci.h>
#include <linux/netdevice.h>
#include "ionic.h"
#include "ionic_bus.h"
+#include "ionic_lif.h"
#include "ionic_debugfs.h"
#ifdef CONFIG_DEBUG_FS
@@ -58,4 +60,38 @@ void ionic_debugfs_add_ident(struct ionic *ionic)
ionic, &identity_fops) ? 0 : -EOPNOTSUPP;
}
+void ionic_debugfs_add_sizes(struct ionic *ionic)
+{
+ debugfs_create_u32("nlifs", 0400, ionic->dentry,
+ (u32 *)&ionic->ident.dev.nlifs);
+ debugfs_create_u32("nintrs", 0400, ionic->dentry, &ionic->nintrs);
+
+ debugfs_create_u32("ntxqs_per_lif", 0400, ionic->dentry,
+ (u32 *)&ionic->ident.lif.eth.config.queue_count[IONIC_QTYPE_TXQ]);
+ debugfs_create_u32("nrxqs_per_lif", 0400, ionic->dentry,
+ (u32 *)&ionic->ident.lif.eth.config.queue_count[IONIC_QTYPE_RXQ]);
+}
+
+static int netdev_show(struct seq_file *seq, void *v)
+{
+ struct net_device *netdev = seq->private;
+
+ seq_printf(seq, "%s\n", netdev->name);
+
+ return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(netdev);
+
+void ionic_debugfs_add_lif(struct ionic_lif *lif)
+{
+ lif->dentry = debugfs_create_dir(lif->name, lif->ionic->dentry);
+ debugfs_create_file("netdev", 0400, lif->dentry,
+ lif->netdev, &netdev_fops);
+}
+
+void ionic_debugfs_del_lif(struct ionic_lif *lif)
+{
+ debugfs_remove_recursive(lif->dentry);
+ lif->dentry = NULL;
+}
#endif
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.h b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.h
index 7073a8b4e2f9..f742acf56adf 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.h
@@ -13,12 +13,18 @@ void ionic_debugfs_destroy(void);
void ionic_debugfs_add_dev(struct ionic *ionic);
void ionic_debugfs_del_dev(struct ionic *ionic);
void ionic_debugfs_add_ident(struct ionic *ionic);
+void ionic_debugfs_add_sizes(struct ionic *ionic);
+void ionic_debugfs_add_lif(struct ionic_lif *lif);
+void ionic_debugfs_del_lif(struct ionic_lif *lif);
#else
static inline void ionic_debugfs_create(void) { }
static inline void ionic_debugfs_destroy(void) { }
static inline void ionic_debugfs_add_dev(struct ionic *ionic) { }
static inline void ionic_debugfs_del_dev(struct ionic *ionic) { }
static inline void ionic_debugfs_add_ident(struct ionic *ionic) { }
+static inline void ionic_debugfs_add_sizes(struct ionic *ionic) { }
+static inline void ionic_debugfs_add_lif(struct ionic_lif *lif) { }
+static inline void ionic_debugfs_del_lif(struct ionic_lif *lif) { }
#endif
#endif /* _IONIC_DEBUGFS_H_ */
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_dev.c b/drivers/net/ethernet/pensando/ionic/ionic_dev.c
index 3137776e9191..01e922fa9366 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_dev.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_dev.c
@@ -226,3 +226,37 @@ void ionic_dev_cmd_port_pause(struct ionic_dev *idev, u8 pause_type)
ionic_dev_cmd_go(idev, &cmd);
}
+
+/* LIF commands */
+void ionic_dev_cmd_lif_identify(struct ionic_dev *idev, u8 type, u8 ver)
+{
+ union ionic_dev_cmd cmd = {
+ .lif_identify.opcode = IONIC_CMD_LIF_IDENTIFY,
+ .lif_identify.type = type,
+ .lif_identify.ver = ver,
+ };
+
+ ionic_dev_cmd_go(idev, &cmd);
+}
+
+void ionic_dev_cmd_lif_init(struct ionic_dev *idev, u16 lif_index,
+ dma_addr_t info_pa)
+{
+ union ionic_dev_cmd cmd = {
+ .lif_init.opcode = IONIC_CMD_LIF_INIT,
+ .lif_init.index = cpu_to_le16(lif_index),
+ .lif_init.info_pa = cpu_to_le64(info_pa),
+ };
+
+ ionic_dev_cmd_go(idev, &cmd);
+}
+
+void ionic_dev_cmd_lif_reset(struct ionic_dev *idev, u16 lif_index)
+{
+ union ionic_dev_cmd cmd = {
+ .lif_init.opcode = IONIC_CMD_LIF_RESET,
+ .lif_init.index = cpu_to_le16(lif_index),
+ };
+
+ ionic_dev_cmd_go(idev, &cmd);
+}
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_dev.h b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
index 81b6910aabc1..e8d4fc888333 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_dev.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
@@ -10,6 +10,8 @@
#include "ionic_if.h"
#include "ionic_regs.h"
+#define IONIC_LIFS_MAX 1024
+
struct ionic_dev_bar {
void __iomem *vaddr;
phys_addr_t bus_addr;
@@ -148,4 +150,9 @@ void ionic_dev_cmd_port_autoneg(struct ionic_dev *idev, u8 an_enable);
void ionic_dev_cmd_port_fec(struct ionic_dev *idev, u8 fec_type);
void ionic_dev_cmd_port_pause(struct ionic_dev *idev, u8 pause_type);
+void ionic_dev_cmd_lif_identify(struct ionic_dev *idev, u8 type, u8 ver);
+void ionic_dev_cmd_lif_init(struct ionic_dev *idev, u16 lif_index,
+ dma_addr_t addr);
+void ionic_dev_cmd_lif_reset(struct ionic_dev *idev, u16 lif_index);
+
#endif /* _IONIC_DEV_H_ */
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.c b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
new file mode 100644
index 000000000000..5528043095d8
--- /dev/null
+++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
@@ -0,0 +1,291 @@
+// SPDX-License-Identifier: GPL-2.0
+/* Copyright(c) 2017 - 2019 Pensando Systems, Inc */
+
+#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
+#include <linux/interrupt.h>
+#include <linux/pci.h>
+#include <linux/cpumask.h>
+
+#include "ionic.h"
+#include "ionic_bus.h"
+#include "ionic_lif.h"
+#include "ionic_debugfs.h"
+
+static struct ionic_lif *ionic_lif_alloc(struct ionic *ionic, unsigned int index)
+{
+ struct device *dev = ionic->dev;
+ struct net_device *netdev;
+ struct ionic_lif *lif;
+ int err;
+
+ netdev = alloc_etherdev_mqs(sizeof(*lif),
+ ionic->ntxqs_per_lif, ionic->ntxqs_per_lif);
+ if (!netdev) {
+ dev_err(dev, "Cannot allocate netdev, aborting\n");
+ return ERR_PTR(-ENOMEM);
+ }
+
+ SET_NETDEV_DEV(netdev, dev);
+
+ lif = netdev_priv(netdev);
+ lif->netdev = netdev;
+
+ lif->neqs = ionic->neqs_per_lif;
+ lif->nxqs = ionic->ntxqs_per_lif;
+
+ lif->ionic = ionic;
+ lif->index = index;
+
+ snprintf(lif->name, sizeof(lif->name), "lif%u", index);
+
+ /* allocate lif info */
+ lif->info_sz = ALIGN(sizeof(*lif->info), PAGE_SIZE);
+ lif->info = dma_alloc_coherent(dev, lif->info_sz,
+ &lif->info_pa, GFP_KERNEL);
+ if (!lif->info) {
+ dev_err(dev, "Failed to allocate lif info, aborting\n");
+ err = -ENOMEM;
+ goto err_out_free_netdev;
+ }
+
+ list_add_tail(&lif->list, &ionic->lifs);
+
+ return lif;
+
+err_out_free_netdev:
+ free_netdev(lif->netdev);
+ lif = NULL;
+
+ return ERR_PTR(err);
+}
+
+int ionic_lifs_alloc(struct ionic *ionic)
+{
+ struct ionic_lif *lif;
+
+ INIT_LIST_HEAD(&ionic->lifs);
+
+ /* only build the first lif, others are for later features */
+ set_bit(0, ionic->lifbits);
+ lif = ionic_lif_alloc(ionic, 0);
+
+ return PTR_ERR_OR_ZERO(lif);
+}
+
+static void ionic_lif_reset(struct ionic_lif *lif)
+{
+ struct ionic_dev *idev = &lif->ionic->idev;
+
+ mutex_lock(&lif->ionic->dev_cmd_lock);
+ ionic_dev_cmd_lif_reset(idev, lif->index);
+ ionic_dev_cmd_wait(lif->ionic, DEVCMD_TIMEOUT);
+ mutex_unlock(&lif->ionic->dev_cmd_lock);
+}
+
+static void ionic_lif_free(struct ionic_lif *lif)
+{
+ struct device *dev = lif->ionic->dev;
+
+ ionic_lif_reset(lif);
+
+ /* free lif info */
+ dma_free_coherent(dev, lif->info_sz, lif->info, lif->info_pa);
+ lif->info = NULL;
+ lif->info_pa = 0;
+
+ /* free netdev & lif */
+ ionic_debugfs_del_lif(lif);
+ list_del(&lif->list);
+ free_netdev(lif->netdev);
+}
+
+void ionic_lifs_free(struct ionic *ionic)
+{
+ struct list_head *cur, *tmp;
+ struct ionic_lif *lif;
+
+ list_for_each_safe(cur, tmp, &ionic->lifs) {
+ lif = list_entry(cur, struct ionic_lif, list);
+
+ ionic_lif_free(lif);
+ }
+}
+
+static void ionic_lif_deinit(struct ionic_lif *lif)
+{
+ if (!test_bit(IONIC_LIF_INITED, lif->state))
+ return;
+
+ clear_bit(IONIC_LIF_INITED, lif->state);
+
+ ionic_lif_reset(lif);
+}
+
+void ionic_lifs_deinit(struct ionic *ionic)
+{
+ struct list_head *cur, *tmp;
+ struct ionic_lif *lif;
+
+ list_for_each_safe(cur, tmp, &ionic->lifs) {
+ lif = list_entry(cur, struct ionic_lif, list);
+ ionic_lif_deinit(lif);
+ }
+}
+
+static int ionic_lif_init(struct ionic_lif *lif)
+{
+ struct ionic_dev *idev = &lif->ionic->idev;
+ struct ionic_lif_init_comp comp;
+ int err;
+
+ ionic_debugfs_add_lif(lif);
+
+ mutex_lock(&lif->ionic->dev_cmd_lock);
+ ionic_dev_cmd_lif_init(idev, lif->index, lif->info_pa);
+ err = ionic_dev_cmd_wait(lif->ionic, DEVCMD_TIMEOUT);
+ ionic_dev_cmd_comp(idev, (union ionic_dev_cmd_comp *)&comp);
+ mutex_unlock(&lif->ionic->dev_cmd_lock);
+ if (err)
+ return err;
+
+ lif->hw_index = le16_to_cpu(comp.hw_index);
+
+ set_bit(IONIC_LIF_INITED, lif->state);
+
+ return 0;
+}
+
+int ionic_lifs_init(struct ionic *ionic)
+{
+ struct list_head *cur, *tmp;
+ struct ionic_lif *lif;
+ int err;
+
+ list_for_each_safe(cur, tmp, &ionic->lifs) {
+ lif = list_entry(cur, struct ionic_lif, list);
+ err = ionic_lif_init(lif);
+ if (err)
+ return err;
+ }
+
+ return 0;
+}
+
+int ionic_lif_identify(struct ionic *ionic, u8 lif_type,
+ union ionic_lif_identity *lid)
+{
+ struct ionic_dev *idev = &ionic->idev;
+ size_t sz;
+ int err;
+
+ sz = min(sizeof(*lid), sizeof(idev->dev_cmd_regs->data));
+
+ mutex_lock(&ionic->dev_cmd_lock);
+ ionic_dev_cmd_lif_identify(idev, lif_type, IONIC_IDENTITY_VERSION_1);
+ err = ionic_dev_cmd_wait(ionic, DEVCMD_TIMEOUT);
+ memcpy_fromio(lid, &idev->dev_cmd_regs->data, sz);
+ mutex_unlock(&ionic->dev_cmd_lock);
+ if (err)
+ return (err);
+
+ dev_dbg(ionic->dev, "capabilities 0x%llx\n",
+ le64_to_cpu(lid->capabilities));
+
+ dev_dbg(ionic->dev, "eth.max_ucast_filters %d\n",
+ le32_to_cpu(lid->eth.max_ucast_filters));
+ dev_dbg(ionic->dev, "eth.max_mcast_filters %d\n",
+ le32_to_cpu(lid->eth.max_mcast_filters));
+ dev_dbg(ionic->dev, "eth.features 0x%llx\n",
+ le64_to_cpu(lid->eth.config.features));
+ dev_dbg(ionic->dev, "eth.queue_count[IONIC_QTYPE_ADMINQ] %d\n",
+ le32_to_cpu(lid->eth.config.queue_count[IONIC_QTYPE_ADMINQ]));
+ dev_dbg(ionic->dev, "eth.queue_count[IONIC_QTYPE_NOTIFYQ] %d\n",
+ le32_to_cpu(lid->eth.config.queue_count[IONIC_QTYPE_NOTIFYQ]));
+ dev_dbg(ionic->dev, "eth.queue_count[IONIC_QTYPE_RXQ] %d\n",
+ le32_to_cpu(lid->eth.config.queue_count[IONIC_QTYPE_RXQ]));
+ dev_dbg(ionic->dev, "eth.queue_count[IONIC_QTYPE_TXQ] %d\n",
+ le32_to_cpu(lid->eth.config.queue_count[IONIC_QTYPE_TXQ]));
+ dev_dbg(ionic->dev, "eth.config.name %s\n", lid->eth.config.name);
+ dev_dbg(ionic->dev, "eth.config.mac %pM\n", lid->eth.config.mac);
+ dev_dbg(ionic->dev, "eth.config.mtu %d\n",
+ le32_to_cpu(lid->eth.config.mtu));
+
+ return 0;
+}
+
+int ionic_lifs_size(struct ionic *ionic)
+{
+ struct ionic_identity *ident = &ionic->ident;
+ unsigned int nintrs, dev_nintrs;
+ union ionic_lif_config *lc;
+ unsigned int ntxqs_per_lif;
+ unsigned int nrxqs_per_lif;
+ unsigned int neqs_per_lif;
+ unsigned int nnqs_per_lif;
+ unsigned int nxqs, neqs;
+ unsigned int min_intrs;
+ int err;
+
+ lc = &ident->lif.eth.config;
+ dev_nintrs = le32_to_cpu(ident->dev.nintrs);
+ neqs_per_lif = le32_to_cpu(ident->lif.rdma.eq_qtype.qid_count);
+ nnqs_per_lif = le32_to_cpu(lc->queue_count[IONIC_QTYPE_NOTIFYQ]);
+ ntxqs_per_lif = le32_to_cpu(lc->queue_count[IONIC_QTYPE_TXQ]);
+ nrxqs_per_lif = le32_to_cpu(lc->queue_count[IONIC_QTYPE_RXQ]);
+
+ nxqs = min(ntxqs_per_lif, nrxqs_per_lif);
+ nxqs = min(nxqs, num_online_cpus());
+ neqs = min(neqs_per_lif, num_online_cpus());
+
+try_again:
+ /* interrupt usage:
+ * 1 for master lif adminq/notifyq
+ * 1 for each CPU for master lif TxRx queue pairs
+ * whatever's left is for RDMA queues
+ */
+ nintrs = 1 + nxqs + neqs;
+ min_intrs = 2; /* adminq + 1 TxRx queue pair */
+
+ if (nintrs > dev_nintrs)
+ goto try_fewer;
+
+ err = ionic_bus_alloc_irq_vectors(ionic, nintrs);
+ if (err < 0 && err != -ENOSPC) {
+ dev_err(ionic->dev, "Can't get intrs from OS: %d\n", err);
+ return err;
+ }
+ if (err == -ENOSPC)
+ goto try_fewer;
+
+ if (err != nintrs) {
+ ionic_bus_free_irq_vectors(ionic);
+ goto try_fewer;
+ }
+
+ ionic->nnqs_per_lif = nnqs_per_lif;
+ ionic->neqs_per_lif = neqs;
+ ionic->ntxqs_per_lif = nxqs;
+ ionic->nrxqs_per_lif = nxqs;
+ ionic->nintrs = nintrs;
+
+ ionic_debugfs_add_sizes(ionic);
+
+ return 0;
+
+try_fewer:
+ if (nnqs_per_lif > 1) {
+ nnqs_per_lif >>= 1;
+ goto try_again;
+ }
+ if (neqs > 1) {
+ neqs >>= 1;
+ goto try_again;
+ }
+ if (nxqs > 1) {
+ nxqs >>= 1;
+ goto try_again;
+ }
+ dev_err(ionic->dev, "Can't get minimum %d intrs from OS\n", min_intrs);
+ return -ENOSPC;
+}
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.h b/drivers/net/ethernet/pensando/ionic/ionic_lif.h
new file mode 100644
index 000000000000..fff4fc287b89
--- /dev/null
+++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.h
@@ -0,0 +1,45 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/* Copyright(c) 2017 - 2019 Pensando Systems, Inc */
+
+#ifndef _IONIC_LIF_H_
+#define _IONIC_LIF_H_
+
+#include <linux/pci.h>
+
+enum ionic_lif_state_flags {
+ IONIC_LIF_INITED,
+
+ /* leave this as last */
+ IONIC_LIF_STATE_SIZE
+};
+
+#define IONIC_LIF_NAME_MAX_SZ 32
+struct ionic_lif {
+ char name[IONIC_LIF_NAME_MAX_SZ];
+ struct list_head list;
+ struct net_device *netdev;
+ DECLARE_BITMAP(state, IONIC_LIF_STATE_SIZE);
+ struct ionic *ionic;
+ bool registered;
+ unsigned int index;
+ unsigned int hw_index;
+ unsigned int neqs;
+ unsigned int nxqs;
+
+ struct ionic_lif_info *info;
+ dma_addr_t info_pa;
+ u32 info_sz;
+
+ struct dentry *dentry;
+ u32 flags;
+};
+
+int ionic_lifs_alloc(struct ionic *ionic);
+void ionic_lifs_free(struct ionic *ionic);
+void ionic_lifs_deinit(struct ionic *ionic);
+int ionic_lifs_init(struct ionic *ionic);
+int ionic_lif_identify(struct ionic *ionic, u8 lif_type,
+ union ionic_lif_identity *lif_ident);
+int ionic_lifs_size(struct ionic *ionic);
+
+#endif /* _IONIC_LIF_H_ */
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_main.c b/drivers/net/ethernet/pensando/ionic/ionic_main.c
index d7e62d31bff0..d6d77abbc4af 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_main.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_main.c
@@ -8,6 +8,7 @@
#include "ionic.h"
#include "ionic_bus.h"
+#include "ionic_lif.h"
#include "ionic_debugfs.h"
MODULE_DESCRIPTION(IONIC_DRV_DESCRIPTION);
--
2.17.1
^ permalink raw reply related
* [PATCH v7 net-next 07/19] ionic: Add basic adminq support
From: Shannon Nelson @ 2019-09-03 22:28 UTC (permalink / raw)
To: snelson, netdev, davem
In-Reply-To: <20190903222821.46161-1-snelson@pensando.io>
Most of the NIC configuration happens through the AdminQ message
queue. NAPI is used for basic interrupt handling and message
queue management. These routines are set up to be shared among
different types of queues when used in slow-path handling.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
---
drivers/net/ethernet/pensando/ionic/ionic.h | 3 +
.../net/ethernet/pensando/ionic/ionic_bus.h | 1 +
.../ethernet/pensando/ionic/ionic_bus_pci.c | 5 +
.../ethernet/pensando/ionic/ionic_debugfs.c | 138 +++++++
.../ethernet/pensando/ionic/ionic_debugfs.h | 4 +
.../net/ethernet/pensando/ionic/ionic_dev.c | 232 ++++++++++++
.../net/ethernet/pensando/ionic/ionic_dev.h | 109 ++++++
.../net/ethernet/pensando/ionic/ionic_lif.c | 356 ++++++++++++++++++
.../net/ethernet/pensando/ionic/ionic_lif.h | 67 ++++
.../net/ethernet/pensando/ionic/ionic_main.c | 26 ++
10 files changed, 941 insertions(+)
diff --git a/drivers/net/ethernet/pensando/ionic/ionic.h b/drivers/net/ethernet/pensando/ionic/ionic.h
index affd7a88b58b..23d39be54064 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic.h
@@ -44,6 +44,9 @@ struct ionic {
DECLARE_BITMAP(intrs, IONIC_INTR_CTRL_REGS_MAX);
};
+int ionic_napi(struct napi_struct *napi, int budget, ionic_cq_cb cb,
+ ionic_cq_done_cb done_cb, void *done_arg);
+
int ionic_dev_cmd_wait(struct ionic *ionic, unsigned long max_wait);
int ionic_set_dma_mask(struct ionic *ionic);
int ionic_setup(struct ionic *ionic);
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_bus.h b/drivers/net/ethernet/pensando/ionic/ionic_bus.h
index 6b29e94f81d6..2f4d08c64910 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_bus.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_bus.h
@@ -4,6 +4,7 @@
#ifndef _IONIC_BUS_H_
#define _IONIC_BUS_H_
+int ionic_bus_get_irq(struct ionic *ionic, unsigned int num);
const char *ionic_bus_info(struct ionic *ionic);
int ionic_bus_alloc_irq_vectors(struct ionic *ionic, unsigned int nintrs);
void ionic_bus_free_irq_vectors(struct ionic *ionic);
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c b/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c
index 4f08d915c3d2..7b6190b96a46 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c
@@ -19,6 +19,11 @@ static const struct pci_device_id ionic_id_table[] = {
};
MODULE_DEVICE_TABLE(pci, ionic_id_table);
+int ionic_bus_get_irq(struct ionic *ionic, unsigned int num)
+{
+ return pci_irq_vector(ionic->pdev, num);
+}
+
const char *ionic_bus_info(struct ionic *ionic)
{
return pci_name(ionic->pdev);
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c
index 840b3da5da3e..d02f81a2b466 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.c
@@ -72,6 +72,137 @@ void ionic_debugfs_add_sizes(struct ionic *ionic)
(u32 *)&ionic->ident.lif.eth.config.queue_count[IONIC_QTYPE_RXQ]);
}
+static int q_tail_show(struct seq_file *seq, void *v)
+{
+ struct ionic_queue *q = seq->private;
+
+ seq_printf(seq, "%d\n", q->tail->index);
+
+ return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(q_tail);
+
+static int q_head_show(struct seq_file *seq, void *v)
+{
+ struct ionic_queue *q = seq->private;
+
+ seq_printf(seq, "%d\n", q->head->index);
+
+ return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(q_head);
+
+static int cq_tail_show(struct seq_file *seq, void *v)
+{
+ struct ionic_cq *cq = seq->private;
+
+ seq_printf(seq, "%d\n", cq->tail->index);
+
+ return 0;
+}
+DEFINE_SHOW_ATTRIBUTE(cq_tail);
+
+static const struct debugfs_reg32 intr_ctrl_regs[] = {
+ { .name = "coal_init", .offset = 0, },
+ { .name = "mask", .offset = 4, },
+ { .name = "credits", .offset = 8, },
+ { .name = "mask_on_assert", .offset = 12, },
+ { .name = "coal_timer", .offset = 16, },
+};
+
+void ionic_debugfs_add_qcq(struct ionic_lif *lif, struct ionic_qcq *qcq)
+{
+ struct dentry *q_dentry, *cq_dentry, *intr_dentry;
+ struct ionic_dev *idev = &lif->ionic->idev;
+ struct debugfs_regset32 *intr_ctrl_regset;
+ struct ionic_intr_info *intr = &qcq->intr;
+ struct debugfs_blob_wrapper *desc_blob;
+ struct device *dev = lif->ionic->dev;
+ struct ionic_queue *q = &qcq->q;
+ struct ionic_cq *cq = &qcq->cq;
+
+ qcq->dentry = debugfs_create_dir(q->name, lif->dentry);
+
+ debugfs_create_x32("total_size", 0400, qcq->dentry, &qcq->total_size);
+ debugfs_create_x64("base_pa", 0400, qcq->dentry, &qcq->base_pa);
+
+ q_dentry = debugfs_create_dir("q", qcq->dentry);
+
+ debugfs_create_u32("index", 0400, q_dentry, &q->index);
+ debugfs_create_x64("base_pa", 0400, q_dentry, &q->base_pa);
+ if (qcq->flags & IONIC_QCQ_F_SG) {
+ debugfs_create_x64("sg_base_pa", 0400, q_dentry,
+ &q->sg_base_pa);
+ debugfs_create_u32("sg_desc_size", 0400, q_dentry,
+ &q->sg_desc_size);
+ }
+ debugfs_create_u32("num_descs", 0400, q_dentry, &q->num_descs);
+ debugfs_create_u32("desc_size", 0400, q_dentry, &q->desc_size);
+ debugfs_create_u32("pid", 0400, q_dentry, &q->pid);
+ debugfs_create_u32("qid", 0400, q_dentry, &q->hw_index);
+ debugfs_create_u32("qtype", 0400, q_dentry, &q->hw_type);
+ debugfs_create_u64("drop", 0400, q_dentry, &q->drop);
+ debugfs_create_u64("stop", 0400, q_dentry, &q->stop);
+ debugfs_create_u64("wake", 0400, q_dentry, &q->wake);
+
+ debugfs_create_file("tail", 0400, q_dentry, q, &q_tail_fops);
+ debugfs_create_file("head", 0400, q_dentry, q, &q_head_fops);
+
+ desc_blob = devm_kzalloc(dev, sizeof(*desc_blob), GFP_KERNEL);
+ if (!desc_blob)
+ return;
+ desc_blob->data = q->base;
+ desc_blob->size = (unsigned long)q->num_descs * q->desc_size;
+ debugfs_create_blob("desc_blob", 0400, q_dentry, desc_blob);
+
+ if (qcq->flags & IONIC_QCQ_F_SG) {
+ desc_blob = devm_kzalloc(dev, sizeof(*desc_blob), GFP_KERNEL);
+ if (!desc_blob)
+ return;
+ desc_blob->data = q->sg_base;
+ desc_blob->size = (unsigned long)q->num_descs * q->sg_desc_size;
+ debugfs_create_blob("sg_desc_blob", 0400, q_dentry,
+ desc_blob);
+ }
+
+ cq_dentry = debugfs_create_dir("cq", qcq->dentry);
+
+ debugfs_create_x64("base_pa", 0400, cq_dentry, &cq->base_pa);
+ debugfs_create_u32("num_descs", 0400, cq_dentry, &cq->num_descs);
+ debugfs_create_u32("desc_size", 0400, cq_dentry, &cq->desc_size);
+ debugfs_create_u8("done_color", 0400, cq_dentry,
+ (u8 *)&cq->done_color);
+
+ debugfs_create_file("tail", 0400, cq_dentry, cq, &cq_tail_fops);
+
+ desc_blob = devm_kzalloc(dev, sizeof(*desc_blob), GFP_KERNEL);
+ if (!desc_blob)
+ return;
+ desc_blob->data = cq->base;
+ desc_blob->size = (unsigned long)cq->num_descs * cq->desc_size;
+ debugfs_create_blob("desc_blob", 0400, cq_dentry, desc_blob);
+
+ if (qcq->flags & IONIC_QCQ_F_INTR) {
+ intr_dentry = debugfs_create_dir("intr", qcq->dentry);
+
+ debugfs_create_u32("index", 0400, intr_dentry,
+ &intr->index);
+ debugfs_create_u32("vector", 0400, intr_dentry,
+ &intr->vector);
+
+ intr_ctrl_regset = devm_kzalloc(dev, sizeof(*intr_ctrl_regset),
+ GFP_KERNEL);
+ if (!intr_ctrl_regset)
+ return;
+ intr_ctrl_regset->regs = intr_ctrl_regs;
+ intr_ctrl_regset->nregs = ARRAY_SIZE(intr_ctrl_regs);
+ intr_ctrl_regset->base = &idev->intr_ctrl[intr->index];
+
+ debugfs_create_regset32("intr_ctrl", 0400, intr_dentry,
+ intr_ctrl_regset);
+ }
+}
+
static int netdev_show(struct seq_file *seq, void *v)
{
struct net_device *netdev = seq->private;
@@ -94,4 +225,11 @@ void ionic_debugfs_del_lif(struct ionic_lif *lif)
debugfs_remove_recursive(lif->dentry);
lif->dentry = NULL;
}
+
+void ionic_debugfs_del_qcq(struct ionic_qcq *qcq)
+{
+ debugfs_remove_recursive(qcq->dentry);
+ qcq->dentry = NULL;
+}
+
#endif
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.h b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.h
index f742acf56adf..c44ebde170b6 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_debugfs.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_debugfs.h
@@ -15,7 +15,9 @@ void ionic_debugfs_del_dev(struct ionic *ionic);
void ionic_debugfs_add_ident(struct ionic *ionic);
void ionic_debugfs_add_sizes(struct ionic *ionic);
void ionic_debugfs_add_lif(struct ionic_lif *lif);
+void ionic_debugfs_add_qcq(struct ionic_lif *lif, struct ionic_qcq *qcq);
void ionic_debugfs_del_lif(struct ionic_lif *lif);
+void ionic_debugfs_del_qcq(struct ionic_qcq *qcq);
#else
static inline void ionic_debugfs_create(void) { }
static inline void ionic_debugfs_destroy(void) { }
@@ -24,7 +26,9 @@ static inline void ionic_debugfs_del_dev(struct ionic *ionic) { }
static inline void ionic_debugfs_add_ident(struct ionic *ionic) { }
static inline void ionic_debugfs_add_sizes(struct ionic *ionic) { }
static inline void ionic_debugfs_add_lif(struct ionic_lif *lif) { }
+static inline void ionic_debugfs_add_qcq(struct ionic_lif *lif, struct ionic_qcq *qcq) { }
static inline void ionic_debugfs_del_lif(struct ionic_lif *lif) { }
+static inline void ionic_debugfs_del_qcq(struct ionic_qcq *qcq) { }
#endif
#endif /* _IONIC_DEBUGFS_H_ */
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_dev.c b/drivers/net/ethernet/pensando/ionic/ionic_dev.c
index dbdde548848e..d168a6435322 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_dev.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_dev.c
@@ -262,7 +262,239 @@ void ionic_dev_cmd_lif_reset(struct ionic_dev *idev, u16 lif_index)
ionic_dev_cmd_go(idev, &cmd);
}
+void ionic_dev_cmd_adminq_init(struct ionic_dev *idev, struct ionic_qcq *qcq,
+ u16 lif_index, u16 intr_index)
+{
+ struct ionic_queue *q = &qcq->q;
+ struct ionic_cq *cq = &qcq->cq;
+
+ union ionic_dev_cmd cmd = {
+ .q_init.opcode = IONIC_CMD_Q_INIT,
+ .q_init.lif_index = cpu_to_le16(lif_index),
+ .q_init.type = q->type,
+ .q_init.index = cpu_to_le32(q->index),
+ .q_init.flags = cpu_to_le16(IONIC_QINIT_F_IRQ |
+ IONIC_QINIT_F_ENA),
+ .q_init.pid = cpu_to_le16(q->pid),
+ .q_init.intr_index = cpu_to_le16(intr_index),
+ .q_init.ring_size = ilog2(q->num_descs),
+ .q_init.ring_base = cpu_to_le64(q->base_pa),
+ .q_init.cq_ring_base = cpu_to_le64(cq->base_pa),
+ };
+
+ ionic_dev_cmd_go(idev, &cmd);
+}
+
int ionic_db_page_num(struct ionic_lif *lif, int pid)
{
return (lif->hw_index * lif->dbid_count) + pid;
}
+
+int ionic_cq_init(struct ionic_lif *lif, struct ionic_cq *cq,
+ struct ionic_intr_info *intr,
+ unsigned int num_descs, size_t desc_size)
+{
+ struct ionic_cq_info *cur;
+ unsigned int ring_size;
+ unsigned int i;
+
+ if (desc_size == 0 || !is_power_of_2(num_descs))
+ return -EINVAL;
+
+ ring_size = ilog2(num_descs);
+ if (ring_size < 2 || ring_size > 16)
+ return -EINVAL;
+
+ cq->lif = lif;
+ cq->bound_intr = intr;
+ cq->num_descs = num_descs;
+ cq->desc_size = desc_size;
+ cq->tail = cq->info;
+ cq->done_color = 1;
+
+ cur = cq->info;
+
+ for (i = 0; i < num_descs; i++) {
+ if (i + 1 == num_descs) {
+ cur->next = cq->info;
+ cur->last = true;
+ } else {
+ cur->next = cur + 1;
+ }
+ cur->index = i;
+ cur++;
+ }
+
+ return 0;
+}
+
+void ionic_cq_map(struct ionic_cq *cq, void *base, dma_addr_t base_pa)
+{
+ struct ionic_cq_info *cur;
+ unsigned int i;
+
+ cq->base = base;
+ cq->base_pa = base_pa;
+
+ for (i = 0, cur = cq->info; i < cq->num_descs; i++, cur++)
+ cur->cq_desc = base + (i * cq->desc_size);
+}
+
+void ionic_cq_bind(struct ionic_cq *cq, struct ionic_queue *q)
+{
+ cq->bound_q = q;
+}
+
+unsigned int ionic_cq_service(struct ionic_cq *cq, unsigned int work_to_do,
+ ionic_cq_cb cb, ionic_cq_done_cb done_cb,
+ void *done_arg)
+{
+ unsigned int work_done = 0;
+
+ if (work_to_do == 0)
+ return 0;
+
+ while (cb(cq, cq->tail)) {
+ if (cq->tail->last)
+ cq->done_color = !cq->done_color;
+ cq->tail = cq->tail->next;
+ DEBUG_STATS_CQE_CNT(cq);
+
+ if (++work_done >= work_to_do)
+ break;
+ }
+
+ if (work_done && done_cb)
+ done_cb(done_arg);
+
+ return work_done;
+}
+
+int ionic_q_init(struct ionic_lif *lif, struct ionic_dev *idev,
+ struct ionic_queue *q, unsigned int index, const char *name,
+ unsigned int num_descs, size_t desc_size,
+ size_t sg_desc_size, unsigned int pid)
+{
+ struct ionic_desc_info *cur;
+ unsigned int ring_size;
+ unsigned int i;
+
+ if (desc_size == 0 || !is_power_of_2(num_descs))
+ return -EINVAL;
+
+ ring_size = ilog2(num_descs);
+ if (ring_size < 2 || ring_size > 16)
+ return -EINVAL;
+
+ q->lif = lif;
+ q->idev = idev;
+ q->index = index;
+ q->num_descs = num_descs;
+ q->desc_size = desc_size;
+ q->sg_desc_size = sg_desc_size;
+ q->tail = q->info;
+ q->head = q->tail;
+ q->pid = pid;
+
+ snprintf(q->name, sizeof(q->name), "L%d-%s%u", lif->index, name, index);
+
+ cur = q->info;
+
+ for (i = 0; i < num_descs; i++) {
+ if (i + 1 == num_descs)
+ cur->next = q->info;
+ else
+ cur->next = cur + 1;
+ cur->index = i;
+ cur->left = num_descs - i;
+ cur++;
+ }
+
+ return 0;
+}
+
+void ionic_q_map(struct ionic_queue *q, void *base, dma_addr_t base_pa)
+{
+ struct ionic_desc_info *cur;
+ unsigned int i;
+
+ q->base = base;
+ q->base_pa = base_pa;
+
+ for (i = 0, cur = q->info; i < q->num_descs; i++, cur++)
+ cur->desc = base + (i * q->desc_size);
+}
+
+void ionic_q_sg_map(struct ionic_queue *q, void *base, dma_addr_t base_pa)
+{
+ struct ionic_desc_info *cur;
+ unsigned int i;
+
+ q->sg_base = base;
+ q->sg_base_pa = base_pa;
+
+ for (i = 0, cur = q->info; i < q->num_descs; i++, cur++)
+ cur->sg_desc = base + (i * q->sg_desc_size);
+}
+
+void ionic_q_post(struct ionic_queue *q, bool ring_doorbell, ionic_desc_cb cb,
+ void *cb_arg)
+{
+ struct device *dev = q->lif->ionic->dev;
+ struct ionic_lif *lif = q->lif;
+
+ q->head->cb = cb;
+ q->head->cb_arg = cb_arg;
+ q->head = q->head->next;
+
+ dev_dbg(dev, "lif=%d qname=%s qid=%d qtype=%d p_index=%d ringdb=%d\n",
+ q->lif->index, q->name, q->hw_type, q->hw_index,
+ q->head->index, ring_doorbell);
+
+ if (ring_doorbell)
+ ionic_dbell_ring(lif->kern_dbpage, q->hw_type,
+ q->dbval | q->head->index);
+}
+
+static bool ionic_q_is_posted(struct ionic_queue *q, unsigned int pos)
+{
+ unsigned int mask, tail, head;
+
+ mask = q->num_descs - 1;
+ tail = q->tail->index;
+ head = q->head->index;
+
+ return ((pos - tail) & mask) < ((head - tail) & mask);
+}
+
+void ionic_q_service(struct ionic_queue *q, struct ionic_cq_info *cq_info,
+ unsigned int stop_index)
+{
+ struct ionic_desc_info *desc_info;
+ ionic_desc_cb cb;
+ void *cb_arg;
+
+ /* check for empty queue */
+ if (q->tail->index == q->head->index)
+ return;
+
+ /* stop index must be for a descriptor that is not yet completed */
+ if (unlikely(!ionic_q_is_posted(q, stop_index)))
+ dev_err(q->lif->ionic->dev,
+ "ionic stop is not posted %s stop %u tail %u head %u\n",
+ q->name, stop_index, q->tail->index, q->head->index);
+
+ do {
+ desc_info = q->tail;
+ q->tail = desc_info->next;
+
+ cb = desc_info->cb;
+ cb_arg = desc_info->cb_arg;
+
+ desc_info->cb = NULL;
+ desc_info->cb_arg = NULL;
+
+ if (cb)
+ cb(q, desc_info, cq_info, cb_arg);
+ } while (desc_info->index != stop_index);
+}
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_dev.h b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
index 2252fa9ad0e3..86922dd26b72 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_dev.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
@@ -126,6 +126,59 @@ struct ionic_dev {
struct ionic_devinfo dev_info;
};
+struct ionic_cq_info {
+ void *cq_desc;
+ struct ionic_cq_info *next;
+ unsigned int index;
+ bool last;
+};
+
+struct ionic_queue;
+struct ionic_qcq;
+struct ionic_desc_info;
+
+typedef void (*ionic_desc_cb)(struct ionic_queue *q,
+ struct ionic_desc_info *desc_info,
+ struct ionic_cq_info *cq_info, void *cb_arg);
+
+struct ionic_desc_info {
+ void *desc;
+ void *sg_desc;
+ struct ionic_desc_info *next;
+ unsigned int index;
+ unsigned int left;
+ ionic_desc_cb cb;
+ void *cb_arg;
+};
+
+#define QUEUE_NAME_MAX_SZ 32
+
+struct ionic_queue {
+ u64 dbell_count;
+ u64 drop;
+ u64 stop;
+ u64 wake;
+ struct ionic_lif *lif;
+ struct ionic_desc_info *info;
+ struct ionic_desc_info *tail;
+ struct ionic_desc_info *head;
+ struct ionic_dev *idev;
+ unsigned int index;
+ unsigned int type;
+ unsigned int hw_index;
+ unsigned int hw_type;
+ u64 dbval;
+ void *base;
+ void *sg_base;
+ dma_addr_t base_pa;
+ dma_addr_t sg_base_pa;
+ unsigned int num_descs;
+ unsigned int desc_size;
+ unsigned int sg_desc_size;
+ unsigned int pid;
+ char name[QUEUE_NAME_MAX_SZ];
+};
+
#define INTR_INDEX_NOT_ASSIGNED -1
#define INTR_NAME_MAX_SZ 32
@@ -138,6 +191,20 @@ struct ionic_intr_info {
cpumask_t affinity_mask;
};
+struct ionic_cq {
+ void *base;
+ dma_addr_t base_pa;
+ struct ionic_lif *lif;
+ struct ionic_cq_info *info;
+ struct ionic_cq_info *tail;
+ struct ionic_queue *bound_q;
+ struct ionic_intr_info *bound_intr;
+ bool done_color;
+ unsigned int num_descs;
+ u64 compl_count;
+ unsigned int desc_size;
+};
+
struct ionic;
static inline void ionic_intr_init(struct ionic_dev *idev,
@@ -148,6 +215,23 @@ static inline void ionic_intr_init(struct ionic_dev *idev,
intr->index = index;
}
+static inline unsigned int ionic_q_space_avail(struct ionic_queue *q)
+{
+ unsigned int avail = q->tail->index;
+
+ if (q->head->index >= avail)
+ avail += q->head->left - 1;
+ else
+ avail -= q->head->index + 1;
+
+ return avail;
+}
+
+static inline bool ionic_q_has_space(struct ionic_queue *q, unsigned int want)
+{
+ return ionic_q_space_avail(q) >= want;
+}
+
void ionic_init_devinfo(struct ionic *ionic);
int ionic_dev_setup(struct ionic *ionic);
void ionic_dev_teardown(struct ionic *ionic);
@@ -174,7 +258,32 @@ void ionic_dev_cmd_lif_identify(struct ionic_dev *idev, u8 type, u8 ver);
void ionic_dev_cmd_lif_init(struct ionic_dev *idev, u16 lif_index,
dma_addr_t addr);
void ionic_dev_cmd_lif_reset(struct ionic_dev *idev, u16 lif_index);
+void ionic_dev_cmd_adminq_init(struct ionic_dev *idev, struct ionic_qcq *qcq,
+ u16 lif_index, u16 intr_index);
int ionic_db_page_num(struct ionic_lif *lif, int pid);
+int ionic_cq_init(struct ionic_lif *lif, struct ionic_cq *cq,
+ struct ionic_intr_info *intr,
+ unsigned int num_descs, size_t desc_size);
+void ionic_cq_map(struct ionic_cq *cq, void *base, dma_addr_t base_pa);
+void ionic_cq_bind(struct ionic_cq *cq, struct ionic_queue *q);
+typedef bool (*ionic_cq_cb)(struct ionic_cq *cq, struct ionic_cq_info *cq_info);
+typedef void (*ionic_cq_done_cb)(void *done_arg);
+unsigned int ionic_cq_service(struct ionic_cq *cq, unsigned int work_to_do,
+ ionic_cq_cb cb, ionic_cq_done_cb done_cb,
+ void *done_arg);
+
+int ionic_q_init(struct ionic_lif *lif, struct ionic_dev *idev,
+ struct ionic_queue *q, unsigned int index, const char *name,
+ unsigned int num_descs, size_t desc_size,
+ size_t sg_desc_size, unsigned int pid);
+void ionic_q_map(struct ionic_queue *q, void *base, dma_addr_t base_pa);
+void ionic_q_sg_map(struct ionic_queue *q, void *base, dma_addr_t base_pa);
+void ionic_q_post(struct ionic_queue *q, bool ring_doorbell, ionic_desc_cb cb,
+ void *cb_arg);
+void ionic_q_rewind(struct ionic_queue *q, struct ionic_desc_info *start);
+void ionic_q_service(struct ionic_queue *q, struct ionic_cq_info *cq_info,
+ unsigned int stop_index);
+
#endif /* _IONIC_DEV_H_ */
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.c b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
index e9dc97b968b5..6fc2b4bb4c58 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_lif.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
@@ -12,6 +12,284 @@
#include "ionic_lif.h"
#include "ionic_debugfs.h"
+static irqreturn_t ionic_isr(int irq, void *data)
+{
+ struct napi_struct *napi = data;
+
+ napi_schedule_irqoff(napi);
+
+ return IRQ_HANDLED;
+}
+
+static int ionic_request_irq(struct ionic_lif *lif, struct ionic_qcq *qcq)
+{
+ struct ionic_intr_info *intr = &qcq->intr;
+ struct device *dev = lif->ionic->dev;
+ struct ionic_queue *q = &qcq->q;
+ const char *name;
+
+ if (lif->registered)
+ name = lif->netdev->name;
+ else
+ name = dev_name(dev);
+
+ snprintf(intr->name, sizeof(intr->name),
+ "%s-%s-%s", IONIC_DRV_NAME, name, q->name);
+
+ return devm_request_irq(dev, intr->vector, ionic_isr,
+ 0, intr->name, &qcq->napi);
+}
+
+static int ionic_intr_alloc(struct ionic_lif *lif, struct ionic_intr_info *intr)
+{
+ struct ionic *ionic = lif->ionic;
+ int index;
+
+ index = find_first_zero_bit(ionic->intrs, ionic->nintrs);
+ if (index == ionic->nintrs) {
+ netdev_warn(lif->netdev, "%s: no intr, index=%d nintrs=%d\n",
+ __func__, index, ionic->nintrs);
+ return -ENOSPC;
+ }
+
+ set_bit(index, ionic->intrs);
+ ionic_intr_init(&ionic->idev, intr, index);
+
+ return 0;
+}
+
+static void ionic_intr_free(struct ionic_lif *lif, int index)
+{
+ if (index != INTR_INDEX_NOT_ASSIGNED && index < lif->ionic->nintrs)
+ clear_bit(index, lif->ionic->intrs);
+}
+
+static void ionic_lif_qcq_deinit(struct ionic_lif *lif, struct ionic_qcq *qcq)
+{
+ struct ionic_dev *idev = &lif->ionic->idev;
+ struct device *dev = lif->ionic->dev;
+
+ if (!qcq)
+ return;
+
+ ionic_debugfs_del_qcq(qcq);
+
+ if (!(qcq->flags & IONIC_QCQ_F_INITED))
+ return;
+
+ if (qcq->flags & IONIC_QCQ_F_INTR) {
+ ionic_intr_mask(idev->intr_ctrl, qcq->intr.index,
+ IONIC_INTR_MASK_SET);
+ devm_free_irq(dev, qcq->intr.vector, &qcq->napi);
+ netif_napi_del(&qcq->napi);
+ }
+
+ qcq->flags &= ~IONIC_QCQ_F_INITED;
+}
+
+static void ionic_qcq_free(struct ionic_lif *lif, struct ionic_qcq *qcq)
+{
+ struct device *dev = lif->ionic->dev;
+
+ if (!qcq)
+ return;
+
+ dma_free_coherent(dev, qcq->total_size, qcq->base, qcq->base_pa);
+ qcq->base = NULL;
+ qcq->base_pa = 0;
+
+ if (qcq->flags & IONIC_QCQ_F_INTR)
+ ionic_intr_free(lif, qcq->intr.index);
+
+ devm_kfree(dev, qcq->cq.info);
+ qcq->cq.info = NULL;
+ devm_kfree(dev, qcq->q.info);
+ qcq->q.info = NULL;
+ devm_kfree(dev, qcq);
+}
+
+static void ionic_qcqs_free(struct ionic_lif *lif)
+{
+ if (lif->adminqcq) {
+ ionic_qcq_free(lif, lif->adminqcq);
+ lif->adminqcq = NULL;
+ }
+}
+
+static int ionic_qcq_alloc(struct ionic_lif *lif, unsigned int type,
+ unsigned int index,
+ const char *name, unsigned int flags,
+ unsigned int num_descs, unsigned int desc_size,
+ unsigned int cq_desc_size,
+ unsigned int sg_desc_size,
+ unsigned int pid, struct ionic_qcq **qcq)
+{
+ struct ionic_dev *idev = &lif->ionic->idev;
+ u32 q_size, cq_size, sg_size, total_size;
+ struct device *dev = lif->ionic->dev;
+ void *q_base, *cq_base, *sg_base;
+ dma_addr_t cq_base_pa = 0;
+ dma_addr_t sg_base_pa = 0;
+ dma_addr_t q_base_pa = 0;
+ struct ionic_qcq *new;
+ int err;
+
+ *qcq = NULL;
+
+ q_size = num_descs * desc_size;
+ cq_size = num_descs * cq_desc_size;
+ sg_size = num_descs * sg_desc_size;
+
+ total_size = ALIGN(q_size, PAGE_SIZE) + ALIGN(cq_size, PAGE_SIZE);
+ /* Note: aligning q_size/cq_size is not enough due to cq_base
+ * address aligning as q_base could be not aligned to the page.
+ * Adding PAGE_SIZE.
+ */
+ total_size += PAGE_SIZE;
+ if (flags & IONIC_QCQ_F_SG) {
+ total_size += ALIGN(sg_size, PAGE_SIZE);
+ total_size += PAGE_SIZE;
+ }
+
+ new = devm_kzalloc(dev, sizeof(*new), GFP_KERNEL);
+ if (!new) {
+ netdev_err(lif->netdev, "Cannot allocate queue structure\n");
+ err = -ENOMEM;
+ goto err_out;
+ }
+
+ new->flags = flags;
+
+ new->q.info = devm_kzalloc(dev, sizeof(*new->q.info) * num_descs,
+ GFP_KERNEL);
+ if (!new->q.info) {
+ netdev_err(lif->netdev, "Cannot allocate queue info\n");
+ err = -ENOMEM;
+ goto err_out;
+ }
+
+ new->q.type = type;
+
+ err = ionic_q_init(lif, idev, &new->q, index, name, num_descs,
+ desc_size, sg_desc_size, pid);
+ if (err) {
+ netdev_err(lif->netdev, "Cannot initialize queue\n");
+ goto err_out;
+ }
+
+ if (flags & IONIC_QCQ_F_INTR) {
+ err = ionic_intr_alloc(lif, &new->intr);
+ if (err) {
+ netdev_warn(lif->netdev, "no intr for %s: %d\n",
+ name, err);
+ goto err_out;
+ }
+
+ err = ionic_bus_get_irq(lif->ionic, new->intr.index);
+ if (err < 0) {
+ netdev_warn(lif->netdev, "no vector for %s: %d\n",
+ name, err);
+ goto err_out_free_intr;
+ }
+ new->intr.vector = err;
+ ionic_intr_mask_assert(idev->intr_ctrl, new->intr.index,
+ IONIC_INTR_MASK_SET);
+
+ new->intr.cpu = new->intr.index % num_online_cpus();
+ if (cpu_online(new->intr.cpu))
+ cpumask_set_cpu(new->intr.cpu,
+ &new->intr.affinity_mask);
+ } else {
+ new->intr.index = INTR_INDEX_NOT_ASSIGNED;
+ }
+
+ new->cq.info = devm_kzalloc(dev, sizeof(*new->cq.info) * num_descs,
+ GFP_KERNEL);
+ if (!new->cq.info) {
+ netdev_err(lif->netdev, "Cannot allocate completion queue info\n");
+ err = -ENOMEM;
+ goto err_out_free_intr;
+ }
+
+ err = ionic_cq_init(lif, &new->cq, &new->intr, num_descs, cq_desc_size);
+ if (err) {
+ netdev_err(lif->netdev, "Cannot initialize completion queue\n");
+ goto err_out_free_intr;
+ }
+
+ new->base = dma_alloc_coherent(dev, total_size, &new->base_pa,
+ GFP_KERNEL);
+ if (!new->base) {
+ netdev_err(lif->netdev, "Cannot allocate queue DMA memory\n");
+ err = -ENOMEM;
+ goto err_out_free_intr;
+ }
+
+ new->total_size = total_size;
+
+ q_base = new->base;
+ q_base_pa = new->base_pa;
+
+ cq_base = (void *)ALIGN((uintptr_t)q_base + q_size, PAGE_SIZE);
+ cq_base_pa = ALIGN(q_base_pa + q_size, PAGE_SIZE);
+
+ if (flags & IONIC_QCQ_F_SG) {
+ sg_base = (void *)ALIGN((uintptr_t)cq_base + cq_size,
+ PAGE_SIZE);
+ sg_base_pa = ALIGN(cq_base_pa + cq_size, PAGE_SIZE);
+ ionic_q_sg_map(&new->q, sg_base, sg_base_pa);
+ }
+
+ ionic_q_map(&new->q, q_base, q_base_pa);
+ ionic_cq_map(&new->cq, cq_base, cq_base_pa);
+ ionic_cq_bind(&new->cq, &new->q);
+
+ *qcq = new;
+
+ return 0;
+
+err_out_free_intr:
+ ionic_intr_free(lif, new->intr.index);
+err_out:
+ dev_err(dev, "qcq alloc of %s%d failed %d\n", name, index, err);
+ return err;
+}
+
+static int ionic_qcqs_alloc(struct ionic_lif *lif)
+{
+ unsigned int flags;
+ int err;
+
+ flags = IONIC_QCQ_F_INTR;
+ err = ionic_qcq_alloc(lif, IONIC_QTYPE_ADMINQ, 0, "admin", flags,
+ IONIC_ADMINQ_LENGTH,
+ sizeof(struct ionic_admin_cmd),
+ sizeof(struct ionic_admin_comp),
+ 0, lif->kern_pid, &lif->adminqcq);
+ if (err)
+ return err;
+
+ return 0;
+}
+
+static bool ionic_adminq_service(struct ionic_cq *cq,
+ struct ionic_cq_info *cq_info)
+{
+ struct ionic_admin_comp *comp = cq_info->cq_desc;
+
+ if (!color_match(comp->color, cq->done_color))
+ return false;
+
+ ionic_q_service(cq->bound_q, cq_info, le16_to_cpu(comp->comp_index));
+
+ return true;
+}
+
+static int ionic_adminq_napi(struct napi_struct *napi, int budget)
+{
+ return ionic_napi(napi, budget, ionic_adminq_service, NULL, NULL);
+}
+
static struct ionic_lif *ionic_lif_alloc(struct ionic *ionic, unsigned int index)
{
struct device *dev = ionic->dev;
@@ -39,6 +317,8 @@ static struct ionic_lif *ionic_lif_alloc(struct ionic *ionic, unsigned int index
snprintf(lif->name, sizeof(lif->name), "lif%u", index);
+ spin_lock_init(&lif->adminq_lock);
+
/* allocate lif info */
lif->info_sz = ALIGN(sizeof(*lif->info), PAGE_SIZE);
lif->info = dma_alloc_coherent(dev, lif->info_sz,
@@ -49,10 +329,19 @@ static struct ionic_lif *ionic_lif_alloc(struct ionic *ionic, unsigned int index
goto err_out_free_netdev;
}
+ /* allocate queues */
+ err = ionic_qcqs_alloc(lif);
+ if (err)
+ goto err_out_free_lif_info;
+
list_add_tail(&lif->list, &ionic->lifs);
return lif;
+err_out_free_lif_info:
+ dma_free_coherent(dev, lif->info_sz, lif->info, lif->info_pa);
+ lif->info = NULL;
+ lif->info_pa = 0;
err_out_free_netdev:
free_netdev(lif->netdev);
lif = NULL;
@@ -87,6 +376,8 @@ static void ionic_lif_free(struct ionic_lif *lif)
{
struct device *dev = lif->ionic->dev;
+ /* free queues */
+ ionic_qcqs_free(lif);
ionic_lif_reset(lif);
/* free lif info */
@@ -125,6 +416,9 @@ static void ionic_lif_deinit(struct ionic_lif *lif)
clear_bit(IONIC_LIF_INITED, lif->state);
+ napi_disable(&lif->adminqcq->napi);
+ ionic_lif_qcq_deinit(lif, lif->adminqcq);
+
ionic_lif_reset(lif);
}
@@ -139,6 +433,59 @@ void ionic_lifs_deinit(struct ionic *ionic)
}
}
+static int ionic_lif_adminq_init(struct ionic_lif *lif)
+{
+ struct device *dev = lif->ionic->dev;
+ struct ionic_q_init_comp comp;
+ struct ionic_dev *idev;
+ struct ionic_qcq *qcq;
+ struct ionic_queue *q;
+ int err;
+
+ idev = &lif->ionic->idev;
+ qcq = lif->adminqcq;
+ q = &qcq->q;
+
+ mutex_lock(&lif->ionic->dev_cmd_lock);
+ ionic_dev_cmd_adminq_init(idev, qcq, lif->index, qcq->intr.index);
+ err = ionic_dev_cmd_wait(lif->ionic, DEVCMD_TIMEOUT);
+ ionic_dev_cmd_comp(idev, (union ionic_dev_cmd_comp *)&comp);
+ mutex_unlock(&lif->ionic->dev_cmd_lock);
+ if (err) {
+ netdev_err(lif->netdev, "adminq init failed %d\n", err);
+ return err;
+ }
+
+ q->hw_type = comp.hw_type;
+ q->hw_index = le32_to_cpu(comp.hw_index);
+ q->dbval = IONIC_DBELL_QID(q->hw_index);
+
+ dev_dbg(dev, "adminq->hw_type %d\n", q->hw_type);
+ dev_dbg(dev, "adminq->hw_index %d\n", q->hw_index);
+
+ netif_napi_add(lif->netdev, &qcq->napi, ionic_adminq_napi,
+ NAPI_POLL_WEIGHT);
+
+ err = ionic_request_irq(lif, qcq);
+ if (err) {
+ netdev_warn(lif->netdev, "adminq irq request failed %d\n", err);
+ netif_napi_del(&qcq->napi);
+ return err;
+ }
+
+ napi_enable(&qcq->napi);
+
+ if (qcq->flags & IONIC_QCQ_F_INTR)
+ ionic_intr_mask(idev->intr_ctrl, qcq->intr.index,
+ IONIC_INTR_MASK_CLEAR);
+
+ qcq->flags |= IONIC_QCQ_F_INITED;
+
+ ionic_debugfs_add_qcq(lif, qcq);
+
+ return 0;
+}
+
static int ionic_lif_init(struct ionic_lif *lif)
{
struct ionic_dev *idev = &lif->ionic->idev;
@@ -184,10 +531,19 @@ static int ionic_lif_init(struct ionic_lif *lif)
goto err_out_free_dbid;
}
+ err = ionic_lif_adminq_init(lif);
+ if (err)
+ goto err_out_adminq_deinit;
+
set_bit(IONIC_LIF_INITED, lif->state);
return 0;
+err_out_adminq_deinit:
+ ionic_lif_qcq_deinit(lif, lif->adminqcq);
+ ionic_lif_reset(lif);
+ ionic_bus_unmap_dbpage(lif->ionic, lif->kern_dbpage);
+ lif->kern_dbpage = NULL;
err_out_free_dbid:
kfree(lif->dbid_inuse);
lif->dbid_inuse = NULL;
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.h b/drivers/net/ethernet/pensando/ionic/ionic_lif.h
index ec8d06ad4192..d897048ae77f 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_lif.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.h
@@ -6,6 +6,55 @@
#include <linux/pci.h>
+#define IONIC_ADMINQ_LENGTH 16 /* must be a power of two */
+
+#define IONIC_MAX_NUM_NAPI_CNTR (NAPI_POLL_WEIGHT + 1)
+#define IONIC_MAX_NUM_SG_CNTR (IONIC_TX_MAX_SG_ELEMS + 1)
+
+struct ionic_tx_stats {
+ u64 pkts;
+ u64 bytes;
+};
+
+struct ionic_rx_stats {
+ u64 pkts;
+ u64 bytes;
+};
+
+#define IONIC_QCQ_F_INITED BIT(0)
+#define IONIC_QCQ_F_SG BIT(1)
+#define IONIC_QCQ_F_INTR BIT(2)
+
+struct ionic_napi_stats {
+ u64 poll_count;
+ u64 work_done_cntr[IONIC_MAX_NUM_NAPI_CNTR];
+};
+
+struct ionic_q_stats {
+ union {
+ struct ionic_tx_stats tx;
+ struct ionic_rx_stats rx;
+ };
+};
+
+struct ionic_qcq {
+ void *base;
+ dma_addr_t base_pa;
+ unsigned int total_size;
+ struct ionic_queue q;
+ struct ionic_cq cq;
+ struct ionic_intr_info intr;
+ struct napi_struct napi;
+ struct ionic_napi_stats napi_stats;
+ struct ionic_q_stats *stats;
+ unsigned int flags;
+ struct dentry *dentry;
+};
+
+#define q_to_qcq(q) container_of(q, struct ionic_qcq, q)
+#define napi_to_qcq(napi) container_of(napi, struct ionic_qcq, napi)
+#define napi_to_cq(napi) (&napi_to_qcq(napi)->cq)
+
enum ionic_lif_state_flags {
IONIC_LIF_INITED,
@@ -25,6 +74,8 @@ struct ionic_lif {
unsigned int hw_index;
unsigned int kern_pid;
u64 __iomem *kern_dbpage;
+ spinlock_t adminq_lock; /* lock for AdminQ operations */
+ struct ionic_qcq *adminqcq;
unsigned int neqs;
unsigned int nxqs;
@@ -46,4 +97,20 @@ int ionic_lif_identify(struct ionic *ionic, u8 lif_type,
union ionic_lif_identity *lif_ident);
int ionic_lifs_size(struct ionic *ionic);
+static inline void debug_stats_napi_poll(struct ionic_qcq *qcq,
+ unsigned int work_done)
+{
+ qcq->napi_stats.poll_count++;
+
+ if (work_done > (IONIC_MAX_NUM_NAPI_CNTR - 1))
+ work_done = IONIC_MAX_NUM_NAPI_CNTR - 1;
+
+ qcq->napi_stats.work_done_cntr[work_done]++;
+}
+
+#define DEBUG_STATS_CQE_CNT(cq) ((cq)->compl_count++)
+#define DEBUG_STATS_INTR_REARM(intr) ((intr)->rearm_count++)
+#define DEBUG_STATS_NAPI_POLL(qcq, work_done) \
+ debug_stats_napi_poll(qcq, work_done)
+
#endif /* _IONIC_LIF_H_ */
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_main.c b/drivers/net/ethernet/pensando/ionic/ionic_main.c
index d6d77abbc4af..56206fd3e3a5 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_main.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_main.c
@@ -169,6 +169,32 @@ static const char *ionic_opcode_to_str(enum ionic_cmd_opcode opcode)
}
}
+int ionic_napi(struct napi_struct *napi, int budget, ionic_cq_cb cb,
+ ionic_cq_done_cb done_cb, void *done_arg)
+{
+ struct ionic_qcq *qcq = napi_to_qcq(napi);
+ struct ionic_cq *cq = &qcq->cq;
+ u32 work_done, flags = 0;
+
+ work_done = ionic_cq_service(cq, budget, cb, done_cb, done_arg);
+
+ if (work_done < budget && napi_complete_done(napi, work_done)) {
+ flags |= IONIC_INTR_CRED_UNMASK;
+ DEBUG_STATS_INTR_REARM(cq->bound_intr);
+ }
+
+ if (work_done || flags) {
+ flags |= IONIC_INTR_CRED_RESET_COALESCE;
+ ionic_intr_credits(cq->lif->ionic->idev.intr_ctrl,
+ cq->bound_intr->index,
+ work_done, flags);
+ }
+
+ DEBUG_STATS_NAPI_POLL(qcq, work_done);
+
+ return work_done;
+}
+
int ionic_dev_cmd_wait(struct ionic *ionic, unsigned long max_seconds)
{
struct ionic_dev *idev = &ionic->idev;
--
2.17.1
^ permalink raw reply related
* [PATCH v7 net-next 06/19] ionic: Add interrupts and doorbells
From: Shannon Nelson @ 2019-09-03 22:28 UTC (permalink / raw)
To: snelson, netdev, davem
In-Reply-To: <20190903222821.46161-1-snelson@pensando.io>
The ionic interrupt model is based on interrupt control blocks
accessed through the PCI BAR. Doorbell registers are used by
the driver to signal to the NIC that requests are waiting on
the message queues. Interrupts are used by the NIC to signal
to the driver that answers are waiting on the completion queues.
Signed-off-by: Shannon Nelson <snelson@pensando.io>
---
drivers/net/ethernet/pensando/ionic/ionic.h | 3 ++
.../net/ethernet/pensando/ionic/ionic_bus.h | 2 +
.../ethernet/pensando/ionic/ionic_bus_pci.c | 12 ++++++
.../net/ethernet/pensando/ionic/ionic_dev.c | 6 +++
.../net/ethernet/pensando/ionic/ionic_dev.h | 22 +++++++++++
.../net/ethernet/pensando/ionic/ionic_lif.c | 39 +++++++++++++++++++
.../net/ethernet/pensando/ionic/ionic_lif.h | 4 ++
.../net/ethernet/pensando/ionic/ionic_regs.h | 3 ++
8 files changed, 91 insertions(+)
diff --git a/drivers/net/ethernet/pensando/ionic/ionic.h b/drivers/net/ethernet/pensando/ionic/ionic.h
index 723b2ba6874e..affd7a88b58b 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic.h
@@ -4,6 +4,8 @@
#ifndef _IONIC_H_
#define _IONIC_H_
+struct ionic_lif;
+
#include "ionic_if.h"
#include "ionic_dev.h"
#include "ionic_devlink.h"
@@ -39,6 +41,7 @@ struct ionic {
unsigned int nrxqs_per_lif;
DECLARE_BITMAP(lifbits, IONIC_LIFS_MAX);
unsigned int nintrs;
+ DECLARE_BITMAP(intrs, IONIC_INTR_CTRL_REGS_MAX);
};
int ionic_dev_cmd_wait(struct ionic *ionic, unsigned long max_wait);
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_bus.h b/drivers/net/ethernet/pensando/ionic/ionic_bus.h
index 3b1e2d0ebf8f..6b29e94f81d6 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_bus.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_bus.h
@@ -9,5 +9,7 @@ int ionic_bus_alloc_irq_vectors(struct ionic *ionic, unsigned int nintrs);
void ionic_bus_free_irq_vectors(struct ionic *ionic);
int ionic_bus_register_driver(void);
void ionic_bus_unregister_driver(void);
+void __iomem *ionic_bus_map_dbpage(struct ionic *ionic, int page_num);
+void ionic_bus_unmap_dbpage(struct ionic *ionic, void __iomem *page);
#endif /* _IONIC_BUS_H_ */
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c b/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c
index f0e0daee45bc..4f08d915c3d2 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_bus_pci.c
@@ -87,6 +87,18 @@ static void ionic_unmap_bars(struct ionic *ionic)
}
}
+void __iomem *ionic_bus_map_dbpage(struct ionic *ionic, int page_num)
+{
+ return pci_iomap_range(ionic->pdev,
+ ionic->bars[IONIC_PCI_BAR_DBELL].res_index,
+ (u64)page_num << PAGE_SHIFT, PAGE_SIZE);
+}
+
+void ionic_bus_unmap_dbpage(struct ionic *ionic, void __iomem *page)
+{
+ iounmap(page);
+}
+
static int ionic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
{
struct device *dev = &pdev->dev;
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_dev.c b/drivers/net/ethernet/pensando/ionic/ionic_dev.c
index 01e922fa9366..dbdde548848e 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_dev.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_dev.c
@@ -9,6 +9,7 @@
#include <linux/etherdevice.h>
#include "ionic.h"
#include "ionic_dev.h"
+#include "ionic_lif.h"
void ionic_init_devinfo(struct ionic *ionic)
{
@@ -260,3 +261,8 @@ void ionic_dev_cmd_lif_reset(struct ionic_dev *idev, u16 lif_index)
ionic_dev_cmd_go(idev, &cmd);
}
+
+int ionic_db_page_num(struct ionic_lif *lif, int pid)
+{
+ return (lif->hw_index * lif->dbid_count) + pid;
+}
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_dev.h b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
index e8d4fc888333..2252fa9ad0e3 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_dev.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_dev.h
@@ -126,8 +126,28 @@ struct ionic_dev {
struct ionic_devinfo dev_info;
};
+#define INTR_INDEX_NOT_ASSIGNED -1
+#define INTR_NAME_MAX_SZ 32
+
+struct ionic_intr_info {
+ char name[INTR_NAME_MAX_SZ];
+ unsigned int index;
+ unsigned int vector;
+ u64 rearm_count;
+ unsigned int cpu;
+ cpumask_t affinity_mask;
+};
+
struct ionic;
+static inline void ionic_intr_init(struct ionic_dev *idev,
+ struct ionic_intr_info *intr,
+ unsigned long index)
+{
+ ionic_intr_clean(idev->intr_ctrl, index);
+ intr->index = index;
+}
+
void ionic_init_devinfo(struct ionic *ionic);
int ionic_dev_setup(struct ionic *ionic);
void ionic_dev_teardown(struct ionic *ionic);
@@ -155,4 +175,6 @@ void ionic_dev_cmd_lif_init(struct ionic_dev *idev, u16 lif_index,
dma_addr_t addr);
void ionic_dev_cmd_lif_reset(struct ionic_dev *idev, u16 lif_index);
+int ionic_db_page_num(struct ionic_lif *lif, int pid);
+
#endif /* _IONIC_DEV_H_ */
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.c b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
index 5528043095d8..e9dc97b968b5 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_lif.c
+++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.c
@@ -94,6 +94,12 @@ static void ionic_lif_free(struct ionic_lif *lif)
lif->info = NULL;
lif->info_pa = 0;
+ /* unmap doorbell page */
+ ionic_bus_unmap_dbpage(lif->ionic, lif->kern_dbpage);
+ lif->kern_dbpage = NULL;
+ kfree(lif->dbid_inuse);
+ lif->dbid_inuse = NULL;
+
/* free netdev & lif */
ionic_debugfs_del_lif(lif);
list_del(&lif->list);
@@ -136,7 +142,9 @@ void ionic_lifs_deinit(struct ionic *ionic)
static int ionic_lif_init(struct ionic_lif *lif)
{
struct ionic_dev *idev = &lif->ionic->idev;
+ struct device *dev = lif->ionic->dev;
struct ionic_lif_init_comp comp;
+ int dbpage_num;
int err;
ionic_debugfs_add_lif(lif);
@@ -151,9 +159,40 @@ static int ionic_lif_init(struct ionic_lif *lif)
lif->hw_index = le16_to_cpu(comp.hw_index);
+ /* now that we have the hw_index we can figure out our doorbell page */
+ lif->dbid_count = le32_to_cpu(lif->ionic->ident.dev.ndbpgs_per_lif);
+ if (!lif->dbid_count) {
+ dev_err(dev, "No doorbell pages, aborting\n");
+ return -EINVAL;
+ }
+
+ lif->dbid_inuse = bitmap_alloc(lif->dbid_count, GFP_KERNEL);
+ if (!lif->dbid_inuse) {
+ dev_err(dev, "Failed alloc doorbell id bitmap, aborting\n");
+ return -ENOMEM;
+ }
+
+ /* first doorbell id reserved for kernel (dbid aka pid == zero) */
+ set_bit(0, lif->dbid_inuse);
+ lif->kern_pid = 0;
+
+ dbpage_num = ionic_db_page_num(lif, lif->kern_pid);
+ lif->kern_dbpage = ionic_bus_map_dbpage(lif->ionic, dbpage_num);
+ if (!lif->kern_dbpage) {
+ dev_err(dev, "Cannot map dbpage, aborting\n");
+ err = -ENOMEM;
+ goto err_out_free_dbid;
+ }
+
set_bit(IONIC_LIF_INITED, lif->state);
return 0;
+
+err_out_free_dbid:
+ kfree(lif->dbid_inuse);
+ lif->dbid_inuse = NULL;
+
+ return err;
}
int ionic_lifs_init(struct ionic *ionic)
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_lif.h b/drivers/net/ethernet/pensando/ionic/ionic_lif.h
index fff4fc287b89..ec8d06ad4192 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_lif.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_lif.h
@@ -23,6 +23,8 @@ struct ionic_lif {
bool registered;
unsigned int index;
unsigned int hw_index;
+ unsigned int kern_pid;
+ u64 __iomem *kern_dbpage;
unsigned int neqs;
unsigned int nxqs;
@@ -30,6 +32,8 @@ struct ionic_lif {
dma_addr_t info_pa;
u32 info_sz;
+ unsigned long *dbid_inuse;
+ unsigned int dbid_count;
struct dentry *dentry;
u32 flags;
};
diff --git a/drivers/net/ethernet/pensando/ionic/ionic_regs.h b/drivers/net/ethernet/pensando/ionic/ionic_regs.h
index 3523915061ed..03ee5a36472b 100644
--- a/drivers/net/ethernet/pensando/ionic/ionic_regs.h
+++ b/drivers/net/ethernet/pensando/ionic/ionic_regs.h
@@ -22,6 +22,9 @@ struct ionic_intr {
u32 rsvd[3];
};
+#define IONIC_INTR_CTRL_REGS_MAX 2048
+#define IONIC_INTR_CTRL_COAL_MAX 0x3F
+
/** enum ionic_intr_mask_vals - valid values for mask and mask_assert.
* @IONIC_INTR_MASK_CLEAR: unmask interrupt.
* @IONIC_INTR_MASK_SET: mask interrupt.
--
2.17.1
^ permalink raw reply related
* Re: [PATCH bpf-next 00/13] bpf: adding map batch processing support
From: Stanislav Fomichev @ 2019-09-03 22:30 UTC (permalink / raw)
To: Alexei Starovoitov
Cc: Yonghong Song, Jakub Kicinski, Brian Vazquez, Alexei Starovoitov,
bpf@vger.kernel.org, netdev@vger.kernel.org, Daniel Borkmann,
Kernel Team
In-Reply-To: <20190903210127.z6mhkryqg6qz62dq@ast-mbp.dhcp.thefacebook.com>
On 09/03, Alexei Starovoitov wrote:
> On Fri, Aug 30, 2019 at 02:18:09PM -0700, Stanislav Fomichev wrote:
> > > >
> > > > I personally like Jakub's/Quentin's proposal more. So if I get to choose
> > > > between this series and Jakub's filter+dump in BPF, I'd pick filter+dump
> > > > (pending per-cpu issue which we actually care about).
> > > >
> > > > But if we can have both, I don't have any objections; this patch
>
> I think we need to have both.
> imo Jakub's and Yonghong's approach are solving slightly different cases.
>
> filter+dump via program is better suited for LRU map walks where filter prog
> would do some non-trivial logic.
> Whereas plain 'delete all' or 'dump all' is much simpler to use without
> loading yet another prog just to dump it.
> bpf infra today isn't quite ready for this very short lived auxiliary progs.
> At prog load pages get read-only mapping, tlbs across cpus flushed,
> kallsyms populated, FDs allocated, etc.
> Loading the prog is a heavy operation. There was a chatter before to have
> built-in progs. This filter+dump could benefit from builtin 'allow all'
> or 'delete all' progs, but imo that complicates design and asks even
> more questions than it answers. Should this builtin progs show up
> in 'bpftool prog show' ? When do they load/unload? Same safety requirements
> as normal progs? etc.
> imo it's fine to have little bit overlap between apis.
> So I think we should proceed with both batching apis.
We don't need to load filter+dump every time we need a dump, right?
We, internally, want to have this 'batch dump' only for long running daemons
(I think the same applies to bcc), we can load this filter+dump once and
then have a sys_bpf() command to trigger it.
Also, related, if we add this batch dump, it doesn't mean that
everything should switch to it. For example, I feel like we
are perfectly fine if bpftool still uses get_next_key+lookup
since we use it only for debugging.
> Having said that I think both are suffering from the important issue pointed out
> by Brian: when kernel deletes an element get_next_key iterator over hash/lru
> map will produce duplicates.
> The amount of duplicates can be huge. When batched iterator is slow and
> bpf prog is doing a lot of update/delete, there could be 10x worth of duplicates,
> since walk will resume from the beginning.
> User space cannot be tasked to deal with it.
> I think this issue has to be solved in the kernel first and it may require
> different batching api.
>
> One idea is to use bucket spin_lock and batch process it bucket-at-a-time.
> From api pov the user space will tell kernel:
> - here is the buffer for N element. start dump from the beginning.
> - kernel will return <= N elements and an iterator.
> - user space will pass this opaque iterator back to get another batch
> For well behaved hash/lru map there will be zero or one elements per bucket.
> When there are 2+ the batching logic can process them together.
> If 'lookup' is requested the kernel can check whether user space provided
> enough space for these 2 elements. If not abort the batch earlier.
> get_next_key won't be used. Instead some sort of opaque iterator
> will be returned to user space, so next batch lookup can start from it.
> This iterator could be the index of the last dumped bucket.
> This idea won't work for pathological hash tables though.
> A lot of elements in a single bucket may be more than room for single batch.
> In such case iterator will get stuck, since num_of_elements_in_bucket > batch_buf_size.
> May be special error code can be used to solve that?
This all requires new per-map implementations unfortunately :-(
We were trying to see if we can somehow improve the existing bpf_map_ops
to be more friendly towards batching.
You also bring a valid point with regard to well behaved hash/lru,
we might be optimizing for the wrong case :-)
> I hope we can come up with other ideas to have a stable iterator over hash table.
> Let's use email to describe the ideas and upcoming LPC conference to
> sort out details and finalize the one to use.
^ permalink raw reply
* Re: [PATCH 0/4] ethtool: implement Energy Detect Powerdown support via phy-tunable
From: David Miller @ 2019-09-03 22:41 UTC (permalink / raw)
To: alexandru.ardelean; +Cc: netdev, linux-kernel, andrew, f.fainelli, hkallweit1
In-Reply-To: <20190903160626.7518-1-alexandru.ardelean@analog.com>
From: Alexandru Ardelean <alexandru.ardelean@analog.com>
Date: Tue, 3 Sep 2019 19:06:22 +0300
> First 2 patches implement the kernel support for controlling Energy Detect
> Powerdown support via phy-tunable, and the next 2 patches implement the
> ethtool user-space control.
You should do this as two separate patch series, one for the kernel
portion and one for the ethtool part.
^ permalink raw reply
* Re: [PATCH] Clock-independent TCP ISN generation
From: David Miller @ 2019-09-03 22:43 UTC (permalink / raw)
To: sirus.shahini; +Cc: eric.dumazet, shiraz.saleem, jgg, arnd, netdev, sirus
In-Reply-To: <e3bf138f-672e-cefa-5fe5-ea25af8d3d61@gmail.com>
From: Cyrus Sh <sirus.shahini@gmail.com>
Date: Tue, 3 Sep 2019 10:06:03 -0600
> On 9/3/19 9:59 AM, Eric Dumazet wrote:
>>
>> You could add a random delay to all SYN packets, if you believe your host has clock skews.
>
> In theory yes, but again do you know any practical example with tested
> applications and the list of the rules? I'm interested to see an actual example
> that somebody has carried out and observed its results.
That's ironic that you're asking for specific and "practical" examples
when you submitted a kernel patch that doesn't even compile.
^ permalink raw reply
* Re: [PATCH] Clock-independent TCP ISN generation
From: David Miller @ 2019-09-03 22:45 UTC (permalink / raw)
To: sirus.shahini; +Cc: eric.dumazet, shiraz.saleem, jgg, arnd, netdev, sirus
In-Reply-To: <e02c0aac-05c5-e0a4-9ae1-57685a0c3160@gmail.com>
From: Cyrus Sh <sirus.shahini@gmail.com>
Date: Tue, 3 Sep 2019 10:27:41 -0600
> It's up to you whether to want to keep using a problematic code that
> may endanger users or want to do something about it since we won't
> insist on having a patch accepted.
At least our problematic code, unlike your patch, compiles.
^ permalink raw reply
* Re: Proposal: r8152 firmware patching framework
From: David Miller @ 2019-09-03 22:50 UTC (permalink / raw)
To: pmalani
Cc: bambi.yeh, hayeswang, amber.chen, netdev, ryankao, jackc, albertk,
marcochen, nic_swsd, grundler
In-Reply-To: <CACeCKadhJz3fdR+0rm+O2E39EbJgmN5NipMT8GRNtorus8myEg@mail.gmail.com>
From: Prashant Malani <pmalani@chromium.org>
Date: Tue, 3 Sep 2019 14:32:01 -0700
> I've moved David to the TO list to hopefully get his suggestions and
> guidance about how to design this in a upstream-compatible way.
I am not an expert in this area so please do not solicit my opinion.
Thank you.
^ permalink raw reply
* [PATCH bpf-next] selftests/bpf: precision tracking tests
From: Alexei Starovoitov @ 2019-09-03 22:51 UTC (permalink / raw)
To: davem; +Cc: daniel, netdev, bpf, kernel-team
Add two tests to check that stack slot marking during backtracking
doesn't trigger 'spi > allocated_stack' warning.
One test is using BPF_ST insn. Another is using BPF_STX.
Signed-off-by: Alexei Starovoitov <ast@kernel.org>
---
these tests depend on the fix https://patchwork.ozlabs.org/patch/1157368/
---
.../testing/selftests/bpf/verifier/precise.c | 52 +++++++++++++++++++
1 file changed, 52 insertions(+)
diff --git a/tools/testing/selftests/bpf/verifier/precise.c b/tools/testing/selftests/bpf/verifier/precise.c
index a455a4a71f11..02151f8c940f 100644
--- a/tools/testing/selftests/bpf/verifier/precise.c
+++ b/tools/testing/selftests/bpf/verifier/precise.c
@@ -140,3 +140,55 @@
.errstr = "!read_ok",
.result = REJECT,
},
+{
+ "precise: ST insn causing spi > allocated_stack",
+ .insns = {
+ BPF_MOV64_REG(BPF_REG_3, BPF_REG_10),
+ BPF_JMP_IMM(BPF_JNE, BPF_REG_3, 123, 0),
+ BPF_ST_MEM(BPF_DW, BPF_REG_3, -8, 0),
+ BPF_LDX_MEM(BPF_DW, BPF_REG_4, BPF_REG_10, -8),
+ BPF_MOV64_IMM(BPF_REG_0, -1),
+ BPF_JMP_REG(BPF_JGT, BPF_REG_4, BPF_REG_0, 0),
+ BPF_EXIT_INSN(),
+ },
+ .prog_type = BPF_PROG_TYPE_XDP,
+ .flags = BPF_F_TEST_STATE_FREQ,
+ .errstr = "5: (2d) if r4 > r0 goto pc+0\
+ last_idx 5 first_idx 5\
+ parent didn't have regs=10 stack=0 marks\
+ last_idx 4 first_idx 2\
+ regs=10 stack=0 before 4\
+ regs=10 stack=0 before 3\
+ regs=0 stack=1 before 2\
+ last_idx 5 first_idx 5\
+ parent didn't have regs=1 stack=0 marks",
+ .result = VERBOSE_ACCEPT,
+ .retval = -1,
+},
+{
+ "precise: STX insn causing spi > allocated_stack",
+ .insns = {
+ BPF_RAW_INSN(BPF_JMP | BPF_CALL, 0, 0, 0, BPF_FUNC_get_prandom_u32),
+ BPF_MOV64_REG(BPF_REG_3, BPF_REG_10),
+ BPF_JMP_IMM(BPF_JNE, BPF_REG_3, 123, 0),
+ BPF_STX_MEM(BPF_DW, BPF_REG_3, BPF_REG_0, -8),
+ BPF_LDX_MEM(BPF_DW, BPF_REG_4, BPF_REG_10, -8),
+ BPF_MOV64_IMM(BPF_REG_0, -1),
+ BPF_JMP_REG(BPF_JGT, BPF_REG_4, BPF_REG_0, 0),
+ BPF_EXIT_INSN(),
+ },
+ .prog_type = BPF_PROG_TYPE_XDP,
+ .flags = BPF_F_TEST_STATE_FREQ,
+ .errstr = "last_idx 6 first_idx 6\
+ parent didn't have regs=10 stack=0 marks\
+ last_idx 5 first_idx 3\
+ regs=10 stack=0 before 5\
+ regs=10 stack=0 before 4\
+ regs=0 stack=1 before 3\
+ last_idx 6 first_idx 6\
+ parent didn't have regs=1 stack=0 marks\
+ last_idx 5 first_idx 3\
+ regs=1 stack=0 before 5",
+ .result = VERBOSE_ACCEPT,
+ .retval = -1,
+},
--
2.20.0
^ permalink raw reply related
* Re: [PATCH bpf-next 00/13] bpf: adding map batch processing support
From: Brian Vazquez @ 2019-09-03 23:07 UTC (permalink / raw)
To: Stanislav Fomichev
Cc: Alexei Starovoitov, Yonghong Song, Jakub Kicinski,
Alexei Starovoitov, bpf@vger.kernel.org, netdev@vger.kernel.org,
Daniel Borkmann, Kernel Team
In-Reply-To: <20190903223043.GC2101@mini-arch>
On Tue, Sep 3, 2019 at 3:30 PM Stanislav Fomichev <sdf@fomichev.me> wrote:
>
> On 09/03, Alexei Starovoitov wrote:
> > On Fri, Aug 30, 2019 at 02:18:09PM -0700, Stanislav Fomichev wrote:
> > > > >
> > > > > I personally like Jakub's/Quentin's proposal more. So if I get to choose
> > > > > between this series and Jakub's filter+dump in BPF, I'd pick filter+dump
> > > > > (pending per-cpu issue which we actually care about).
> > > > >
> > > > > But if we can have both, I don't have any objections; this patch
> >
> > I think we need to have both.
> > imo Jakub's and Yonghong's approach are solving slightly different cases.
> >
> > filter+dump via program is better suited for LRU map walks where filter prog
> > would do some non-trivial logic.
> > Whereas plain 'delete all' or 'dump all' is much simpler to use without
> > loading yet another prog just to dump it.
> > bpf infra today isn't quite ready for this very short lived auxiliary progs.
> > At prog load pages get read-only mapping, tlbs across cpus flushed,
> > kallsyms populated, FDs allocated, etc.
> > Loading the prog is a heavy operation. There was a chatter before to have
> > built-in progs. This filter+dump could benefit from builtin 'allow all'
> > or 'delete all' progs, but imo that complicates design and asks even
> > more questions than it answers. Should this builtin progs show up
> > in 'bpftool prog show' ? When do they load/unload? Same safety requirements
> > as normal progs? etc.
> > imo it's fine to have little bit overlap between apis.
> > So I think we should proceed with both batching apis.
> We don't need to load filter+dump every time we need a dump, right?
> We, internally, want to have this 'batch dump' only for long running daemons
> (I think the same applies to bcc), we can load this filter+dump once and
> then have a sys_bpf() command to trigger it.
>
> Also, related, if we add this batch dump, it doesn't mean that
> everything should switch to it. For example, I feel like we
> are perfectly fine if bpftool still uses get_next_key+lookup
> since we use it only for debugging.
>
> > Having said that I think both are suffering from the important issue pointed out
> > by Brian: when kernel deletes an element get_next_key iterator over hash/lru
> > map will produce duplicates.
> > The amount of duplicates can be huge. When batched iterator is slow and
> > bpf prog is doing a lot of update/delete, there could be 10x worth of duplicates,
> > since walk will resume from the beginning.
> > User space cannot be tasked to deal with it.
> > I think this issue has to be solved in the kernel first and it may require
> > different batching api.
We could also modify get_next_key behaviour _only_ when it's called
from a dumping function in which case we do know that we want to move
forward not backwards (basically if prev_key is not found, then
retrieve the first key in the next bucket).
That approach might miss entries that are in the same bucket where the
prev_key used to be, but missing entries is something that can always
happen (new additions in previous buckets), we can not control that
and as it has said before, if users care about consistency they can
use map-in-map.
> > One idea is to use bucket spin_lock and batch process it bucket-at-a-time.
> > From api pov the user space will tell kernel:
> > - here is the buffer for N element. start dump from the beginning.
> > - kernel will return <= N elements and an iterator.
> > - user space will pass this opaque iterator back to get another batch
> > For well behaved hash/lru map there will be zero or one elements per bucket.
> > When there are 2+ the batching logic can process them together.
> > If 'lookup' is requested the kernel can check whether user space provided
> > enough space for these 2 elements. If not abort the batch earlier.
> > get_next_key won't be used. Instead some sort of opaque iterator
> > will be returned to user space, so next batch lookup can start from it.
> > This iterator could be the index of the last dumped bucket.
> > This idea won't work for pathological hash tables though.
> > A lot of elements in a single bucket may be more than room for single batch.
> > In such case iterator will get stuck, since num_of_elements_in_bucket > batch_buf_size.
> > May be special error code can be used to solve that?
> This all requires new per-map implementations unfortunately :-(
> We were trying to see if we can somehow improve the existing bpf_map_ops
> to be more friendly towards batching.
Agreed that one of the motivations for current batching implementation
was to re use the existing code as much as we can. Although this might
be a good opportunity to do some per-map implementations as long as
they behave better than the current ones.
>
> You also bring a valid point with regard to well behaved hash/lru,
> we might be optimizing for the wrong case :-)
For pathological hash tables, wouldn't batching lookup_and_delete
help? In theory you are always requesting for the current first_key,
copying it and deleting it. And even if you retrieve the same key
during the batch (because another cpu added it again) it would contain
different info right?
> > I hope we can come up with other ideas to have a stable iterator over hash table.
> > Let's use email to describe the ideas and upcoming LPC conference to
> > sort out details and finalize the one to use.
^ permalink raw reply
* Re: [PATCH bpf-next 00/13] bpf: adding map batch processing support
From: Yonghong Song @ 2019-09-03 23:07 UTC (permalink / raw)
To: Stanislav Fomichev, Alexei Starovoitov
Cc: Jakub Kicinski, Brian Vazquez, Alexei Starovoitov,
bpf@vger.kernel.org, netdev@vger.kernel.org, Daniel Borkmann,
Kernel Team
In-Reply-To: <20190903223043.GC2101@mini-arch>
On 9/3/19 3:30 PM, Stanislav Fomichev wrote:
> On 09/03, Alexei Starovoitov wrote:
>> On Fri, Aug 30, 2019 at 02:18:09PM -0700, Stanislav Fomichev wrote:
>>>>>
>>>>> I personally like Jakub's/Quentin's proposal more. So if I get to choose
>>>>> between this series and Jakub's filter+dump in BPF, I'd pick filter+dump
>>>>> (pending per-cpu issue which we actually care about).
>>>>>
>>>>> But if we can have both, I don't have any objections; this patch
>>
>> I think we need to have both.
>> imo Jakub's and Yonghong's approach are solving slightly different cases.
>>
>> filter+dump via program is better suited for LRU map walks where filter prog
>> would do some non-trivial logic.
>> Whereas plain 'delete all' or 'dump all' is much simpler to use without
>> loading yet another prog just to dump it.
>> bpf infra today isn't quite ready for this very short lived auxiliary progs.
>> At prog load pages get read-only mapping, tlbs across cpus flushed,
>> kallsyms populated, FDs allocated, etc.
>> Loading the prog is a heavy operation. There was a chatter before to have
>> built-in progs. This filter+dump could benefit from builtin 'allow all'
>> or 'delete all' progs, but imo that complicates design and asks even
>> more questions than it answers. Should this builtin progs show up
>> in 'bpftool prog show' ? When do they load/unload? Same safety requirements
>> as normal progs? etc.
>> imo it's fine to have little bit overlap between apis.
>> So I think we should proceed with both batching apis.
> We don't need to load filter+dump every time we need a dump, right?
> We, internally, want to have this 'batch dump' only for long running daemons
> (I think the same applies to bcc), we can load this filter+dump once and
> then have a sys_bpf() command to trigger it.
>
> Also, related, if we add this batch dump, it doesn't mean that
> everything should switch to it. For example, I feel like we
> are perfectly fine if bpftool still uses get_next_key+lookup
> since we use it only for debugging.
>
>> Having said that I think both are suffering from the important issue pointed out
>> by Brian: when kernel deletes an element get_next_key iterator over hash/lru
>> map will produce duplicates.
>> The amount of duplicates can be huge. When batched iterator is slow and
>> bpf prog is doing a lot of update/delete, there could be 10x worth of duplicates,
>> since walk will resume from the beginning.
>> User space cannot be tasked to deal with it.
>> I think this issue has to be solved in the kernel first and it may require
>> different batching api.
>>
>> One idea is to use bucket spin_lock and batch process it bucket-at-a-time.
>> From api pov the user space will tell kernel:
>> - here is the buffer for N element. start dump from the beginning.
>> - kernel will return <= N elements and an iterator.
>> - user space will pass this opaque iterator back to get another batch
>> For well behaved hash/lru map there will be zero or one elements per bucket.
>> When there are 2+ the batching logic can process them together.
>> If 'lookup' is requested the kernel can check whether user space provided
>> enough space for these 2 elements. If not abort the batch earlier.
>> get_next_key won't be used. Instead some sort of opaque iterator
>> will be returned to user space, so next batch lookup can start from it.
>> This iterator could be the index of the last dumped bucket.
>> This idea won't work for pathological hash tables though.
>> A lot of elements in a single bucket may be more than room for single batch.
>> In such case iterator will get stuck, since num_of_elements_in_bucket > batch_buf_size.
>> May be special error code can be used to solve that?
> This all requires new per-map implementations unfortunately :-(
> We were trying to see if we can somehow improve the existing bpf_map_ops
> to be more friendly towards batching.
The below is a link to folly current hashmap implementation:
https://github.com/facebook/folly/blob/master/folly/concurrency/ConcurrentHashMap.h
It uses segment level batching and each segment can contain multiple
buckets. It support concurrent iterating vs. deletion.
I am yet to fully understand its implementation. But my guess is
that old elements are somehow kept long enough until all the
queries, read, etc. done and then garbage collection kicks in
and removes those old elements.
Kernel bucket-level locking should be okay. Indeed, per-map
implementation or tweaking existing per-map interfaces will be necessary.
>
> You also bring a valid point with regard to well behaved hash/lru,
> we might be optimizing for the wrong case :-)
>
>> I hope we can come up with other ideas to have a stable iterator over hash table.
>> Let's use email to describe the ideas and upcoming LPC conference to
>> sort out details and finalize the one to use.
^ permalink raw reply
* Re: [PATCH] Clock-independent TCP ISN generation
From: Cyrus Sh @ 2019-09-04 0:45 UTC (permalink / raw)
To: David Miller; +Cc: eric.dumazet, shiraz.saleem, jgg, arnd, netdev, sirus
In-Reply-To: <20190903.154553.508717744184330290.davem@davemloft.net>
On 9/3/19 4:45 PM, David Miller wrote:
> At least our problematic code, unlike your patch, compiles.
I obviously compiled and tested the code before sending along and this should be
easy to understand. Even I published the results in the link that I mentioned in
the initial message. Now I'm not sure what you're doing that you're unable to
compile the code. What compilation error message do you get? This patch has
been written for kernel 5.3-rc7.
Further the reason that I asked for a specific practical example was that I
wanted to actually test different techniques and see their effectiveness. (Again
should be easy to understand, I don't know why it's not for you and why it made
you upset)
^ permalink raw reply
* Re: [PATCH bpf-next 00/13] bpf: adding map batch processing support
From: Alexei Starovoitov @ 2019-09-04 1:35 UTC (permalink / raw)
To: Brian Vazquez
Cc: Stanislav Fomichev, Yonghong Song, Jakub Kicinski,
Alexei Starovoitov, bpf@vger.kernel.org, netdev@vger.kernel.org,
Daniel Borkmann, Kernel Team
In-Reply-To: <CAMzD94Qge5CtUZ+_0DsuQ_VLpmoADDZemFH1=WA-H8P0ax8qDQ@mail.gmail.com>
On Tue, Sep 03, 2019 at 04:07:17PM -0700, Brian Vazquez wrote:
>
> We could also modify get_next_key behaviour _only_ when it's called
> from a dumping function in which case we do know that we want to move
> forward not backwards (basically if prev_key is not found, then
> retrieve the first key in the next bucket).
>
> That approach might miss entries that are in the same bucket where the
> prev_key used to be, but missing entries is something that can always
> happen (new additions in previous buckets), we can not control that
> and as it has said before, if users care about consistency they can
> use map-in-map.
for dump-all case such miss of elements might be ok-ish,
but for delete-all it's probably not.
Imagine bpf prog is doing delete and bcc script is doing delete-all.
With 'go to the next bucket' logic there will be a case where
both kernel and user side are doing delete, but some elements are still left.
True that map-in-map is the answer, but if we're doing new get_next op
we probably should do it with less corner cases.
> > This all requires new per-map implementations unfortunately :-(
> > We were trying to see if we can somehow improve the existing bpf_map_ops
> > to be more friendly towards batching.
>
> Agreed that one of the motivations for current batching implementation
> was to re use the existing code as much as we can. Although this might
> be a good opportunity to do some per-map implementations as long as
> they behave better than the current ones.
I don't think non reuse of get_next is a big deal.
Adding another get_next_v2 to all maps is a low cost comparing
with advantages of having stable iterate api.
stable walk over hash map is imo more important feature of api than
perf improvement brought by batching.
^ permalink raw reply
* Re: [PATCH v2 bpf-next 2/3] bpf: implement CAP_BPF
From: Alexei Starovoitov @ 2019-09-04 1:39 UTC (permalink / raw)
To: nicolas.dichtel@6wind.com, Alexei Starovoitov, Daniel Borkmann
Cc: Alexei Starovoitov, luto@amacapital.net, davem@davemloft.net,
peterz@infradead.org, rostedt@goodmis.org, netdev@vger.kernel.org,
bpf@vger.kernel.org, Kernel Team, linux-api@vger.kernel.org
In-Reply-To: <59ac111e-7ce7-5e00-32c9-9b55482fe701@6wind.com>
On 8/30/19 8:19 AM, Nicolas Dichtel wrote:
> Le 29/08/2019 à 19:30, Alexei Starovoitov a écrit :
> [snip]
>> These are the links that showing that k8 can delegates caps.
>> Are you saying that you know of folks who specifically
>> delegate cap_sys_admin and cap_net_admin _only_ to a container to run bpf in there?
>>
> Yes, we need cap_sys_admin only to load bpf:
> tc filter add dev eth0 ingress matchall action bpf obj ./tc_test_kern.o sec test
>
> I'm not sure to understand why cap_net_admin is not enough to run the previous
> command (ie why load is forbidden).
because bpf syscall prog_load command requires cap_sys_admin in
the current implementation.
> I want to avoid sys_admin, thus cap_bpf will be ok. But we need to manage the
> backward compatibility.
re: backward compatibility...
do you know of any case where task is running under userid=nobody
with cap_sys_admin and cap_net_admin in order to do bpf ?
If not then what is the concern about compatibility?
^ permalink raw reply
* Re: [BACKPORT 4.14.y 4/8] net: sctp: fix warning "NULL check before some freeing functions is not needed"
From: Baolin Wang @ 2019-09-04 2:38 UTC (permalink / raw)
To: Greg KH
Cc: Marcelo Ricardo Leitner, # 3.4.x, vyasevich, Neil Horman,
David Miller, hariprasad.kelam, linux-sctp, Networking,
Arnd Bergmann, Orson Zhai, Vincent Guittot, LKML
In-Reply-To: <20190903183331.GB26562@kroah.com>
On Wed, 4 Sep 2019 at 02:33, Greg KH <greg@kroah.com> wrote:
>
> On Tue, Sep 03, 2019 at 11:52:06AM -0300, Marcelo Ricardo Leitner wrote:
> > On Tue, Sep 03, 2019 at 02:58:16PM +0800, Baolin Wang wrote:
> > > From: Hariprasad Kelam <hariprasad.kelam@gmail.com>
> > >
> > > This patch removes NULL checks before calling kfree.
> > >
> > > fixes below issues reported by coccicheck
> > > net/sctp/sm_make_chunk.c:2586:3-8: WARNING: NULL check before some
> > > freeing functions is not needed.
> > > net/sctp/sm_make_chunk.c:2652:3-8: WARNING: NULL check before some
> > > freeing functions is not needed.
> > > net/sctp/sm_make_chunk.c:2667:3-8: WARNING: NULL check before some
> > > freeing functions is not needed.
> > > net/sctp/sm_make_chunk.c:2684:3-8: WARNING: NULL check before some
> > > freeing functions is not needed.
> >
> > Hi. This doesn't seem the kind of patch that should be backported to
> > such old/stable releases. After all, it's just a cleanup.
>
> I agree, this does not seem necessary _unless_ it is needed for a later
> real fix.
It can remove warnings from our product kernel since this patch
(c4964bfaf433 sctp: Free cookie before we memdup a new one) was merged
into stable, we still need backport it to our product kernel manually.
But if you still think this is unnecessary, please ignore this patch.
Thanks for your comments.
--
Baolin Wang
Best Regards
^ permalink raw reply
* [PATCH 1/3] ixgbe: Use kzfree() rather than its implementation.
From: zhong jiang @ 2019-09-04 2:39 UTC (permalink / raw)
To: davem, anna.schumaker, trond.myklebust; +Cc: zhongjiang, netdev, linux-kernel
In-Reply-To: <1567564752-6430-1-git-send-email-zhongjiang@huawei.com>
Use kzfree() instead of memset() + kfree().
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
---
drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c | 9 +++------
1 file changed, 3 insertions(+), 6 deletions(-)
diff --git a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
index 31629fc..113f608 100644
--- a/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
+++ b/drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c
@@ -960,11 +960,9 @@ int ixgbe_ipsec_vf_add_sa(struct ixgbe_adapter *adapter, u32 *msgbuf, u32 vf)
return 0;
err_aead:
- memset(xs->aead, 0, sizeof(*xs->aead));
- kfree(xs->aead);
+ kzfree(xs->aead);
err_xs:
- memset(xs, 0, sizeof(*xs));
- kfree(xs);
+ kzfree(xs);
err_out:
msgbuf[1] = err;
return err;
@@ -1049,8 +1047,7 @@ int ixgbe_ipsec_vf_del_sa(struct ixgbe_adapter *adapter, u32 *msgbuf, u32 vf)
ixgbe_ipsec_del_sa(xs);
/* remove the xs that was made-up in the add request */
- memset(xs, 0, sizeof(*xs));
- kfree(xs);
+ kzfree(xs);
return 0;
}
--
1.7.12.4
^ permalink raw reply related
* [PATCH 0/3] net: Use kzfree() directly
From: zhong jiang @ 2019-09-04 2:39 UTC (permalink / raw)
To: davem, anna.schumaker, trond.myklebust; +Cc: zhongjiang, netdev, linux-kernel
With the help of Coccinelle. We find some place to replace.
@@
expression M, S;
@@
- memset(M, 0, S);
- kfree(M);
+ kzfree(M);
zhong jiang (3):
ixgbe: Use kzfree() rather than its implementation.
sunrpc: Use kzfree rather than its implementation.
net: mpoa: Use kzfree rather than its implementation.
drivers/net/ethernet/intel/ixgbe/ixgbe_ipsec.c | 9 +++------
net/atm/mpoa_caches.c | 6 ++----
net/sunrpc/auth_gss/gss_krb5_keys.c | 9 +++------
3 files changed, 8 insertions(+), 16 deletions(-)
--
1.7.12.4
^ permalink raw reply
* [PATCH 2/3] sunrpc: Use kzfree rather than its implementation.
From: zhong jiang @ 2019-09-04 2:39 UTC (permalink / raw)
To: davem, anna.schumaker, trond.myklebust; +Cc: zhongjiang, netdev, linux-kernel
In-Reply-To: <1567564752-6430-1-git-send-email-zhongjiang@huawei.com>
Use kzfree instead of memset() + kfree().
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
---
net/sunrpc/auth_gss/gss_krb5_keys.c | 9 +++------
1 file changed, 3 insertions(+), 6 deletions(-)
diff --git a/net/sunrpc/auth_gss/gss_krb5_keys.c b/net/sunrpc/auth_gss/gss_krb5_keys.c
index 550fdf1..3b7f721 100644
--- a/net/sunrpc/auth_gss/gss_krb5_keys.c
+++ b/net/sunrpc/auth_gss/gss_krb5_keys.c
@@ -228,14 +228,11 @@ u32 krb5_derive_key(const struct gss_krb5_enctype *gk5e,
ret = 0;
err_free_raw:
- memset(rawkey, 0, keybytes);
- kfree(rawkey);
+ kzfree(rawkey);
err_free_out:
- memset(outblockdata, 0, blocksize);
- kfree(outblockdata);
+ kzfree(outblockdata);
err_free_in:
- memset(inblockdata, 0, blocksize);
- kfree(inblockdata);
+ kzfree(inblockdata);
err_free_cipher:
crypto_free_sync_skcipher(cipher);
err_return:
--
1.7.12.4
^ permalink raw reply related
* [PATCH 3/3] net: mpoa: Use kzfree rather than its implementation.
From: zhong jiang @ 2019-09-04 2:39 UTC (permalink / raw)
To: davem, anna.schumaker, trond.myklebust; +Cc: zhongjiang, netdev, linux-kernel
In-Reply-To: <1567564752-6430-1-git-send-email-zhongjiang@huawei.com>
Use kzfree instead of memset() + kfree().
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
---
net/atm/mpoa_caches.c | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)
diff --git a/net/atm/mpoa_caches.c b/net/atm/mpoa_caches.c
index 4bb4183..3286f9d 100644
--- a/net/atm/mpoa_caches.c
+++ b/net/atm/mpoa_caches.c
@@ -180,8 +180,7 @@ static int cache_hit(in_cache_entry *entry, struct mpoa_client *mpc)
static void in_cache_put(in_cache_entry *entry)
{
if (refcount_dec_and_test(&entry->use)) {
- memset(entry, 0, sizeof(in_cache_entry));
- kfree(entry);
+ kzfree(entry);
}
}
@@ -416,8 +415,7 @@ static eg_cache_entry *eg_cache_get_by_src_ip(__be32 ipaddr,
static void eg_cache_put(eg_cache_entry *entry)
{
if (refcount_dec_and_test(&entry->use)) {
- memset(entry, 0, sizeof(eg_cache_entry));
- kfree(entry);
+ kzfree(entry);
}
}
--
1.7.12.4
^ permalink raw reply related
* Re: [RFC v3] vhost: introduce mdev based hardware vhost backend
From: Tiwei Bie @ 2019-09-04 2:48 UTC (permalink / raw)
To: Michael S. Tsirkin
Cc: jasowang, alex.williamson, maxime.coquelin, linux-kernel, kvm,
virtualization, netdev, dan.daly, cunming.liang, zhihong.wang,
lingshan.zhu
In-Reply-To: <20190903043704-mutt-send-email-mst@kernel.org>
On Tue, Sep 03, 2019 at 07:26:03AM -0400, Michael S. Tsirkin wrote:
> On Wed, Aug 28, 2019 at 01:37:12PM +0800, Tiwei Bie wrote:
> > Details about this can be found here:
> >
> > https://lwn.net/Articles/750770/
> >
> > What's new in this version
> > ==========================
> >
> > There are three choices based on the discussion [1] in RFC v2:
> >
> > > #1. We expose a VFIO device, so we can reuse the VFIO container/group
> > > based DMA API and potentially reuse a lot of VFIO code in QEMU.
> > >
> > > But in this case, we have two choices for the VFIO device interface
> > > (i.e. the interface on top of VFIO device fd):
> > >
> > > A) we may invent a new vhost protocol (as demonstrated by the code
> > > in this RFC) on VFIO device fd to make it work in VFIO's way,
> > > i.e. regions and irqs.
> > >
> > > B) Or as you proposed, instead of inventing a new vhost protocol,
> > > we can reuse most existing vhost ioctls on the VFIO device fd
> > > directly. There should be no conflicts between the VFIO ioctls
> > > (type is 0x3B) and VHOST ioctls (type is 0xAF) currently.
> > >
> > > #2. Instead of exposing a VFIO device, we may expose a VHOST device.
> > > And we will introduce a new mdev driver vhost-mdev to do this.
> > > It would be natural to reuse the existing kernel vhost interface
> > > (ioctls) on it as much as possible. But we will need to invent
> > > some APIs for DMA programming (reusing VHOST_SET_MEM_TABLE is a
> > > choice, but it's too heavy and doesn't support vIOMMU by itself).
> >
> > This version is more like a quick PoC to try Jason's proposal on
> > reusing vhost ioctls. And the second way (#1/B) in above three
> > choices was chosen in this version to demonstrate the idea quickly.
> >
> > Now the userspace API looks like this:
> >
> > - VFIO's container/group based IOMMU API is used to do the
> > DMA programming.
> >
> > - Vhost's existing ioctls are used to setup the device.
> >
> > And the device will report device_api as "vfio-vhost".
> >
> > Note that, there are dirty hacks in this version. If we decide to
> > go this way, some refactoring in vhost.c/vhost.h may be needed.
> >
> > PS. The direct mapping of the notify registers isn't implemented
> > in this version.
> >
> > [1] https://lkml.org/lkml/2019/7/9/101
> >
> > Signed-off-by: Tiwei Bie <tiwei.bie@intel.com>
>
> ....
>
> > +long vhost_mdev_ioctl(struct mdev_device *mdev, unsigned int cmd,
> > + unsigned long arg)
> > +{
> > + void __user *argp = (void __user *)arg;
> > + struct vhost_mdev *vdpa;
> > + unsigned long minsz;
> > + int ret = 0;
> > +
> > + if (!mdev)
> > + return -EINVAL;
> > +
> > + vdpa = mdev_get_drvdata(mdev);
> > + if (!vdpa)
> > + return -ENODEV;
> > +
> > + switch (cmd) {
> > + case VFIO_DEVICE_GET_INFO:
> > + {
> > + struct vfio_device_info info;
> > +
> > + minsz = offsetofend(struct vfio_device_info, num_irqs);
> > +
> > + if (copy_from_user(&info, (void __user *)arg, minsz)) {
> > + ret = -EFAULT;
> > + break;
> > + }
> > +
> > + if (info.argsz < minsz) {
> > + ret = -EINVAL;
> > + break;
> > + }
> > +
> > + info.flags = VFIO_DEVICE_FLAGS_VHOST;
> > + info.num_regions = 0;
> > + info.num_irqs = 0;
> > +
> > + if (copy_to_user((void __user *)arg, &info, minsz)) {
> > + ret = -EFAULT;
> > + break;
> > + }
> > +
> > + break;
> > + }
> > + case VFIO_DEVICE_GET_REGION_INFO:
> > + case VFIO_DEVICE_GET_IRQ_INFO:
> > + case VFIO_DEVICE_SET_IRQS:
> > + case VFIO_DEVICE_RESET:
> > + ret = -EINVAL;
> > + break;
> > +
> > + case VHOST_MDEV_SET_STATE:
> > + ret = vhost_set_state(vdpa, argp);
> > + break;
> > + case VHOST_GET_FEATURES:
> > + ret = vhost_get_features(vdpa, argp);
> > + break;
> > + case VHOST_SET_FEATURES:
> > + ret = vhost_set_features(vdpa, argp);
> > + break;
> > + case VHOST_GET_VRING_BASE:
> > + ret = vhost_get_vring_base(vdpa, argp);
> > + break;
> > + default:
> > + ret = vhost_dev_ioctl(&vdpa->dev, cmd, argp);
> > + if (ret == -ENOIOCTLCMD)
> > + ret = vhost_vring_ioctl(&vdpa->dev, cmd, argp);
> > + }
> > +
> > + return ret;
> > +}
> > +EXPORT_SYMBOL(vhost_mdev_ioctl);
>
>
> I don't have a problem with this approach. A small question:
> would it make sense to have two fds: send vhost ioctls
> on one and vfio ioctls on another?
> We can then pass vfio fd to the vhost fd with a
> SET_BACKEND ioctl.
>
> What do you think?
I like this idea! I will give it a try.
So we can introduce /dev/vhost-mdev to have the vhost fd, and let
userspace pass vfio fd to the vhost fd with a SET_BACKEND ioctl.
Thanks a lot!
Tiwei
>
> --
> MST
^ permalink raw reply
* [PATCH] net: hsr: remove an redundant null check before kfree_skb
From: zhong jiang @ 2019-09-04 3:09 UTC (permalink / raw)
To: davem, arvid.brodin; +Cc: zhongjiang, netdev, linux-kernel
kfree_skb has taken the null pointer into account. Hence just remove
the null check before kfree_skb.
Signed-off-by: zhong jiang <zhongjiang@huawei.com>
---
net/hsr/hsr_forward.c | 6 ++----
1 file changed, 2 insertions(+), 4 deletions(-)
diff --git a/net/hsr/hsr_forward.c b/net/hsr/hsr_forward.c
index ddd9605..0c9e5b0 100644
--- a/net/hsr/hsr_forward.c
+++ b/net/hsr/hsr_forward.c
@@ -367,10 +367,8 @@ void hsr_forward_skb(struct sk_buff *skb, struct hsr_port *port)
port->dev->stats.tx_bytes += skb->len;
}
- if (frame.skb_hsr)
- kfree_skb(frame.skb_hsr);
- if (frame.skb_std)
- kfree_skb(frame.skb_std);
+ kfree_skb(frame.skb_hsr);
+ kfree_skb(frame.skb_std);
return;
out_drop:
--
1.7.12.4
^ permalink raw reply related
* Re: [PATCH] net-ipv6: fix excessive RTF_ADDRCONF flag on ::1/128 local route (and others)
From: Lorenzo Colitti @ 2019-09-04 3:17 UTC (permalink / raw)
To: David Ahern; +Cc: Maciej Żenczykowski, David S . Miller, Linux NetDev
In-Reply-To: <60b98521-cf3a-1130-896d-2947fc4d5290@gmail.com>
On Wed, Sep 4, 2019 at 4:45 AM David Ahern <dsahern@gmail.com> wrote:
>
> exactly. It was shortsighted of me to add the ADDRCONF flag and removing
> it reverts back to the previous behavior.
>
> When I enable radvd, I do see the flag set when it should be and not for
> other addresses. I believe the patch is correct.
Ah, wait, I was confused. Sorry.
What I was saying is that RTF_ADDRCONF flag should be set on the local
table /128 route for the IPv6 addresses configured by RAs. The way
those are configured is ndisc_router_discovery -> addrconf_prefix_rcv
-> addrconf_prefix_rcv_add_addr -> ipv6_add_addr ->
addrconf_f6i_alloc. Because in this patch, addrconf_f6i_alloc
unconditionally clears RTF_ADDRCONF, I didn't see how the flag could
be set. But I now realize that that was not happening before David's
commit, either: in 5.1 those routes were not created with RTF_ADDRCONF
either.
In other words: before 5.1, the /128 routes in the local table to IPv6
addresses created by SLAAC did not have RTF_ADDRCONF. After David's
commit, they did, because *all* /128 routes to IPv6 addresses had
RTF_ADDRCONF (correct for SLAAC adresses, but definitely incorrect for
manual addresses, loopback, etc.). If this commit is applied, we'll go
back to the 5.1 state. In the future it would be good to ensure that
at least the /128 routes created by SLAAC do have RTF_ADDRCONF, but no
need to do so in this commit.
^ permalink raw reply
* [PATCHv2 net-next] ipmr: remove cache_resolve_queue_len
From: Hangbin Liu @ 2019-09-04 3:34 UTC (permalink / raw)
To: netdev
Cc: Phil Karn, Sukumar Gopalakrishnan, David S . Miller,
Alexey Kuznetsov, Hideaki YOSHIFUJI, Hangbin Liu
In-Reply-To: <20190903084359.13310-1-liuhangbin@gmail.com>
This is a re-post of previous patch wrote by David Miller[1].
Phil Karn reported[2] that on busy networks with lots of unresolved
multicast routing entries, the creation of new multicast group routes
can be extremely slow and unreliable.
The reason is we hard-coded multicast route entries with unresolved source
addresses(cache_resolve_queue_len) to 10. If some multicast route never
resolves and the unresolved source addresses increased, there will
be no ability to create new multicast route cache.
To resolve this issue, we need either add a sysctl entry to make the
cache_resolve_queue_len configurable, or just remove cache_resolve_queue_len
directly, as we already have the socket receive queue limits of mrouted
socket, pointed by David.
From my side, I'd perfer to remove the cache_resolve_queue_len instead
of creating two more(IPv4 and IPv6 version) sysctl entry.
[1] https://lkml.org/lkml/2018/7/22/11
[2] https://lkml.org/lkml/2018/7/21/343
v2: hold the mfc_unres_lock while walking the unresolved list in
queue_count(), as Nikolay Aleksandrov remind.
Reported-by: Phil Karn <karn@ka9q.net>
Signed-off-by: Hangbin Liu <liuhangbin@gmail.com>
---
include/linux/mroute_base.h | 2 --
net/ipv4/ipmr.c | 30 +++++++++++++++++++++---------
net/ipv6/ip6mr.c | 10 +++-------
3 files changed, 24 insertions(+), 18 deletions(-)
diff --git a/include/linux/mroute_base.h b/include/linux/mroute_base.h
index 34de06b426ef..50fb89533029 100644
--- a/include/linux/mroute_base.h
+++ b/include/linux/mroute_base.h
@@ -235,7 +235,6 @@ struct mr_table_ops {
* @mfc_hash: Hash table of all resolved routes for easy lookup
* @mfc_cache_list: list of resovled routes for possible traversal
* @maxvif: Identifier of highest value vif currently in use
- * @cache_resolve_queue_len: current size of unresolved queue
* @mroute_do_assert: Whether to inform userspace on wrong ingress
* @mroute_do_pim: Whether to receive IGMP PIMv1
* @mroute_reg_vif_num: PIM-device vif index
@@ -252,7 +251,6 @@ struct mr_table {
struct rhltable mfc_hash;
struct list_head mfc_cache_list;
int maxvif;
- atomic_t cache_resolve_queue_len;
bool mroute_do_assert;
bool mroute_do_pim;
bool mroute_do_wrvifwhole;
diff --git a/net/ipv4/ipmr.c b/net/ipv4/ipmr.c
index c07bc82cbbe9..4d67c64d94a4 100644
--- a/net/ipv4/ipmr.c
+++ b/net/ipv4/ipmr.c
@@ -744,8 +744,6 @@ static void ipmr_destroy_unres(struct mr_table *mrt, struct mfc_cache *c)
struct sk_buff *skb;
struct nlmsgerr *e;
- atomic_dec(&mrt->cache_resolve_queue_len);
-
while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved))) {
if (ip_hdr(skb)->version == 0) {
struct nlmsghdr *nlh = skb_pull(skb,
@@ -1133,9 +1131,11 @@ static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
}
if (!found) {
+ bool was_empty;
+
/* Create a new entry if allowable */
- if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
- (c = ipmr_cache_alloc_unres()) == NULL) {
+ c = ipmr_cache_alloc_unres();
+ if (!c) {
spin_unlock_bh(&mfc_unres_lock);
kfree_skb(skb);
@@ -1161,11 +1161,11 @@ static int ipmr_cache_unresolved(struct mr_table *mrt, vifi_t vifi,
return err;
}
- atomic_inc(&mrt->cache_resolve_queue_len);
+ was_empty = list_empty(&mrt->mfc_unres_queue);
list_add(&c->_c.list, &mrt->mfc_unres_queue);
mroute_netlink_event(mrt, c, RTM_NEWROUTE);
- if (atomic_read(&mrt->cache_resolve_queue_len) == 1)
+ if (was_empty)
mod_timer(&mrt->ipmr_expire_timer,
c->_c.mfc_un.unres.expires);
}
@@ -1272,7 +1272,6 @@ static int ipmr_mfc_add(struct net *net, struct mr_table *mrt,
if (uc->mfc_origin == c->mfc_origin &&
uc->mfc_mcastgrp == c->mfc_mcastgrp) {
list_del(&_uc->list);
- atomic_dec(&mrt->cache_resolve_queue_len);
found = true;
break;
}
@@ -1328,7 +1327,7 @@ static void mroute_clean_tables(struct mr_table *mrt, int flags)
}
if (flags & MRT_FLUSH_MFC) {
- if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
+ if (!list_empty(&mrt->mfc_unres_queue)) {
spin_lock_bh(&mfc_unres_lock);
list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
list_del(&c->list);
@@ -2750,9 +2749,22 @@ static int ipmr_rtm_route(struct sk_buff *skb, struct nlmsghdr *nlh,
return ipmr_mfc_delete(tbl, &mfcc, parent);
}
+static int queue_count(struct mr_table *mrt)
+{
+ struct list_head *pos;
+ int count = 0;
+
+ spin_lock_bh(&mfc_unres_lock);
+ list_for_each(pos, &mrt->mfc_unres_queue)
+ count++;
+ spin_unlock_bh(&mfc_unres_lock);
+
+ return count;
+}
+
static bool ipmr_fill_table(struct mr_table *mrt, struct sk_buff *skb)
{
- u32 queue_len = atomic_read(&mrt->cache_resolve_queue_len);
+ u32 queue_len = queue_count(mrt);
if (nla_put_u32(skb, IPMRA_TABLE_ID, mrt->id) ||
nla_put_u32(skb, IPMRA_TABLE_CACHE_RES_QUEUE_LEN, queue_len) ||
diff --git a/net/ipv6/ip6mr.c b/net/ipv6/ip6mr.c
index e80d36c5073d..d02f0f903559 100644
--- a/net/ipv6/ip6mr.c
+++ b/net/ipv6/ip6mr.c
@@ -768,8 +768,6 @@ static void ip6mr_destroy_unres(struct mr_table *mrt, struct mfc6_cache *c)
struct net *net = read_pnet(&mrt->net);
struct sk_buff *skb;
- atomic_dec(&mrt->cache_resolve_queue_len);
-
while ((skb = skb_dequeue(&c->_c.mfc_un.unres.unresolved)) != NULL) {
if (ipv6_hdr(skb)->version == 0) {
struct nlmsghdr *nlh = skb_pull(skb,
@@ -1148,8 +1146,8 @@ static int ip6mr_cache_unresolved(struct mr_table *mrt, mifi_t mifi,
* Create a new entry if allowable
*/
- if (atomic_read(&mrt->cache_resolve_queue_len) >= 10 ||
- (c = ip6mr_cache_alloc_unres()) == NULL) {
+ c = ip6mr_cache_alloc_unres();
+ if (!c) {
spin_unlock_bh(&mfc_unres_lock);
kfree_skb(skb);
@@ -1176,7 +1174,6 @@ static int ip6mr_cache_unresolved(struct mr_table *mrt, mifi_t mifi,
return err;
}
- atomic_inc(&mrt->cache_resolve_queue_len);
list_add(&c->_c.list, &mrt->mfc_unres_queue);
mr6_netlink_event(mrt, c, RTM_NEWROUTE);
@@ -1468,7 +1465,6 @@ static int ip6mr_mfc_add(struct net *net, struct mr_table *mrt,
if (ipv6_addr_equal(&uc->mf6c_origin, &c->mf6c_origin) &&
ipv6_addr_equal(&uc->mf6c_mcastgrp, &c->mf6c_mcastgrp)) {
list_del(&_uc->list);
- atomic_dec(&mrt->cache_resolve_queue_len);
found = true;
break;
}
@@ -1526,7 +1522,7 @@ static void mroute_clean_tables(struct mr_table *mrt, int flags)
}
if (flags & MRT6_FLUSH_MFC) {
- if (atomic_read(&mrt->cache_resolve_queue_len) != 0) {
+ if (!list_empty(&mrt->mfc_unres_queue)) {
spin_lock_bh(&mfc_unres_lock);
list_for_each_entry_safe(c, tmp, &mrt->mfc_unres_queue, list) {
list_del(&c->list);
--
2.19.2
^ permalink raw reply related
* [PATCH 0/3] net: remove an redundant continue
From: zhong jiang @ 2019-09-04 3:46 UTC (permalink / raw)
To: davem, kvalo, pkshih; +Cc: zhongjiang, netdev, linux-kernel
With the help of Coccinelle. we find some place to replace.
@@
for (...;...;...) {
...
if (...) {
...
- continue;
}
}
zhong jiang (3):
rtlwifi: Remove an unnecessary continue in
_rtl8723be_phy_config_bb_with_pgheaderfile
nfp: Drop unnecessary continue in nfp_net_pf_alloc_vnics
ath10k: Drop unnecessary continue in ath10k_mac_update_vif_chan
drivers/net/ethernet/netronome/nfp/nfp_net_main.c | 4 +---
drivers/net/wireless/ath/ath10k/mac.c | 4 +---
drivers/net/wireless/realtek/rtlwifi/rtl8723be/phy.c | 1 -
3 files changed, 2 insertions(+), 7 deletions(-)
--
1.7.12.4
^ permalink raw reply
page: next (older) | prev (newer) | latest
- recent:[subjects (threaded)|topics (new)|topics (active)]
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox