* [PATCH net-next 01/15] sfc: Store flags from MC_CMD_DRV_ATTACH for later use
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
@ 2013-12-14 23:01 ` Ben Hutchings
2013-12-14 23:01 ` [PATCH net-next 02/15] sfc: Do not use MAC address as clock name Ben Hutchings
` (14 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:01 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/mcdi.c | 18 ++++++++++++++++--
drivers/net/ethernet/sfc/mcdi.h | 2 ++
2 files changed, 18 insertions(+), 2 deletions(-)
diff --git a/drivers/net/ethernet/sfc/mcdi.c b/drivers/net/ethernet/sfc/mcdi.c
index d317dfdee845..d608838f6729 100644
--- a/drivers/net/ethernet/sfc/mcdi.c
+++ b/drivers/net/ethernet/sfc/mcdi.c
@@ -1132,13 +1132,27 @@ static int efx_mcdi_drv_attach(struct efx_nic *efx, bool driver_operating,
goto fail;
}
+ if (driver_operating) {
+ if (outlen >= MC_CMD_DRV_ATTACH_EXT_OUT_LEN) {
+ efx->mcdi->fn_flags =
+ MCDI_DWORD(outbuf,
+ DRV_ATTACH_EXT_OUT_FUNC_FLAGS);
+ } else {
+ /* Synthesise flags for Siena */
+ efx->mcdi->fn_flags =
+ 1 << MC_CMD_DRV_ATTACH_EXT_OUT_FLAG_LINKCTRL |
+ 1 << MC_CMD_DRV_ATTACH_EXT_OUT_FLAG_TRUSTED |
+ (efx_port_num(efx) == 0) <<
+ MC_CMD_DRV_ATTACH_EXT_OUT_FLAG_PRIMARY;
+ }
+ }
+
/* We currently assume we have control of the external link
* and are completely trusted by firmware. Abort probing
* if that's not true for this function.
*/
if (driver_operating &&
- outlen >= MC_CMD_DRV_ATTACH_EXT_OUT_LEN &&
- (MCDI_DWORD(outbuf, DRV_ATTACH_EXT_OUT_FUNC_FLAGS) &
+ (efx->mcdi->fn_flags &
(1 << MC_CMD_DRV_ATTACH_EXT_OUT_FLAG_LINKCTRL |
1 << MC_CMD_DRV_ATTACH_EXT_OUT_FLAG_TRUSTED)) !=
(1 << MC_CMD_DRV_ATTACH_EXT_OUT_FLAG_LINKCTRL |
diff --git a/drivers/net/ethernet/sfc/mcdi.h b/drivers/net/ethernet/sfc/mcdi.h
index 171f5f58f84a..52931aebf3c3 100644
--- a/drivers/net/ethernet/sfc/mcdi.h
+++ b/drivers/net/ethernet/sfc/mcdi.h
@@ -94,12 +94,14 @@ struct efx_mcdi_mtd_partition {
* struct efx_mcdi_data - extra state for NICs that implement MCDI
* @iface: Interface/protocol state
* @hwmon: Hardware monitor state
+ * @fn_flags: Flags for this function, as returned by %MC_CMD_DRV_ATTACH.
*/
struct efx_mcdi_data {
struct efx_mcdi_iface iface;
#ifdef CONFIG_SFC_MCDI_MON
struct efx_mcdi_mon hwmon;
#endif
+ u32 fn_flags;
};
#ifdef CONFIG_SFC_MCDI_MON
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 02/15] sfc: Do not use MAC address as clock name
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
2013-12-14 23:01 ` [PATCH net-next 01/15] sfc: Store flags from MC_CMD_DRV_ATTACH for later use Ben Hutchings
@ 2013-12-14 23:01 ` Ben Hutchings
2013-12-14 23:01 ` [PATCH net-next 03/15] sfc: Initialise efx_ptp_data::phc_clock_info from a static template Ben Hutchings
` (13 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:01 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
We'll be sharing clocks between multiple functions with their own MAC
addresses. The name field is now documented as 'A short "friendly
name" to identify the clock ...' and '... not meant to be a unique
id.' So use the name 'sfc'.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ptp.c | 5 ++---
1 files changed, 2 insertions(+), 3 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index aa4876edaac6..5b434597ac63 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -965,9 +965,8 @@ int efx_ptp_probe(struct efx_nic *efx, struct efx_channel *channel)
ptp->evt_overflow = false;
ptp->phc_clock_info.owner = THIS_MODULE;
- snprintf(ptp->phc_clock_info.name,
- sizeof(ptp->phc_clock_info.name),
- "%pm", efx->net_dev->perm_addr);
+ strlcpy(ptp->phc_clock_info.name, "sfc",
+ sizeof(ptp->phc_clock_info.name));
ptp->phc_clock_info.max_adj = MAX_PPB;
ptp->phc_clock_info.n_alarm = 0;
ptp->phc_clock_info.n_ext_ts = 0;
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 03/15] sfc: Initialise efx_ptp_data::phc_clock_info from a static template
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
2013-12-14 23:01 ` [PATCH net-next 01/15] sfc: Store flags from MC_CMD_DRV_ATTACH for later use Ben Hutchings
2013-12-14 23:01 ` [PATCH net-next 02/15] sfc: Do not use MAC address as clock name Ben Hutchings
@ 2013-12-14 23:01 ` Ben Hutchings
2013-12-14 23:01 ` [PATCH net-next 04/15] sfc: Remove kernel-doc for efx_ptp_data fields not present in this version Ben Hutchings
` (12 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:01 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ptp.c | 30 ++++++++++++++++--------------
1 files changed, 16 insertions(+), 14 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index 5b434597ac63..96008202cbc3 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -926,6 +926,21 @@ static void efx_ptp_worker(struct work_struct *work)
efx_ptp_process_rx(efx, skb);
}
+static const struct ptp_clock_info efx_phc_clock_info = {
+ .owner = THIS_MODULE,
+ .name = "sfc",
+ .max_adj = MAX_PPB,
+ .n_alarm = 0,
+ .n_ext_ts = 0,
+ .n_per_out = 0,
+ .pps = 1,
+ .adjfreq = efx_phc_adjfreq,
+ .adjtime = efx_phc_adjtime,
+ .gettime = efx_phc_gettime,
+ .settime = efx_phc_settime,
+ .enable = efx_phc_enable,
+};
+
/* Initialise PTP state. */
int efx_ptp_probe(struct efx_nic *efx, struct efx_channel *channel)
{
@@ -964,20 +979,7 @@ int efx_ptp_probe(struct efx_nic *efx, struct efx_channel *channel)
list_add(&ptp->rx_evts[pos].link, &ptp->evt_free_list);
ptp->evt_overflow = false;
- ptp->phc_clock_info.owner = THIS_MODULE;
- strlcpy(ptp->phc_clock_info.name, "sfc",
- sizeof(ptp->phc_clock_info.name));
- ptp->phc_clock_info.max_adj = MAX_PPB;
- ptp->phc_clock_info.n_alarm = 0;
- ptp->phc_clock_info.n_ext_ts = 0;
- ptp->phc_clock_info.n_per_out = 0;
- ptp->phc_clock_info.pps = 1;
- ptp->phc_clock_info.adjfreq = efx_phc_adjfreq;
- ptp->phc_clock_info.adjtime = efx_phc_adjtime;
- ptp->phc_clock_info.gettime = efx_phc_gettime;
- ptp->phc_clock_info.settime = efx_phc_settime;
- ptp->phc_clock_info.enable = efx_phc_enable;
-
+ ptp->phc_clock_info = efx_phc_clock_info;
ptp->phc_clock = ptp_clock_register(&ptp->phc_clock_info,
&efx->pci_dev->dev);
if (IS_ERR(ptp->phc_clock)) {
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 04/15] sfc: Remove kernel-doc for efx_ptp_data fields not present in this version
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (2 preceding siblings ...)
2013-12-14 23:01 ` [PATCH net-next 03/15] sfc: Initialise efx_ptp_data::phc_clock_info from a static template Ben Hutchings
@ 2013-12-14 23:01 ` Ben Hutchings
2013-12-14 23:02 ` [PATCH net-next 05/15] sfc: PTP - tidy up unused/useless variables Ben Hutchings
` (11 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:01 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ptp.c | 15 ---------------
1 files changed, 0 insertions(+), 15 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index 96008202cbc3..be7ed4e1c96c 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -248,21 +248,6 @@ struct efx_ptp_timeset {
* @nic_ts_enabled: Flag indicating if NIC generated TS events are handled
* @txbuf: Buffer for use when transmitting (PTP) packets to MC (avoids
* allocations in main data path).
- * @debug_ptp_dir: PTP debugfs directory
- * @missed_rx_sync: Number of packets received without syncrhonisation.
- * @good_syncs: Number of successful synchronisations.
- * @no_time_syncs: Number of synchronisations with no good times.
- * @bad_sync_durations: Number of synchronisations with bad durations.
- * @bad_syncs: Number of failed synchronisations.
- * @last_sync_time: Number of nanoseconds for last synchronisation.
- * @sync_timeouts: Number of synchronisation timeouts
- * @fast_syncs: Number of synchronisations requiring short delay
- * @min_sync_delta: Minimum time between event and synchronisation
- * @max_sync_delta: Maximum time between event and synchronisation
- * @average_sync_delta: Average time between event and synchronisation.
- * Modified moving average.
- * @last_sync_delta: Last time between event and synchronisation
- * @mc_stats: Context value for MC statistics
* @timeset: Last set of synchronisation statistics.
*/
struct efx_ptp_data {
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 05/15] sfc: PTP - tidy up unused/useless variables
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (3 preceding siblings ...)
2013-12-14 23:01 ` [PATCH net-next 04/15] sfc: Remove kernel-doc for efx_ptp_data fields not present in this version Ben Hutchings
@ 2013-12-14 23:02 ` Ben Hutchings
2013-12-14 23:02 ` [PATCH net-next 06/15] sfc: Tidy up PTP synchronization code Ben Hutchings
` (10 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:02 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
From: Laurence Evans <levans@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ptp.c | 16 +---------------
1 files changed, 1 insertions(+), 15 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index be7ed4e1c96c..82d8e27e8a58 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -237,9 +237,6 @@ struct efx_ptp_timeset {
* @evt_code: Last event code
* @start: Address at which MC indicates ready for synchronisation
* @host_time_pps: Host time at last PPS
- * @last_sync_ns: Last number of nanoseconds between readings when synchronising
- * @base_sync_ns: Number of nanoseconds for last synchronisation.
- * @base_sync_valid: Whether base_sync_time is valid.
* @current_adjfreq: Current ppb adjustment.
* @phc_clock: Pointer to registered phc device
* @phc_clock_info: Registration structure for phc device
@@ -274,9 +271,6 @@ struct efx_ptp_data {
int evt_code;
struct efx_buffer start;
struct pps_event_time host_time_pps;
- unsigned last_sync_ns;
- unsigned base_sync_ns;
- bool base_sync_valid;
s64 current_adjfreq;
struct ptp_clock *phc_clock;
struct ptp_clock_info phc_clock_info;
@@ -482,18 +476,10 @@ efx_ptp_process_times(struct efx_nic *efx, MCDI_DECLARE_STRUCT_PTR(synch_buf),
if (ngood == 0) {
netif_warn(efx, drv, efx->net_dev,
- "PTP no suitable synchronisations %dns\n",
- ptp->base_sync_ns);
+ "PTP no suitable synchronisations\n");
return -EAGAIN;
}
- /* Average minimum this synchronisation */
- ptp->last_sync_ns = DIV_ROUND_UP(total, ngood);
- if (!ptp->base_sync_valid || (ptp->last_sync_ns < ptp->base_sync_ns)) {
- ptp->base_sync_valid = true;
- ptp->base_sync_ns = ptp->last_sync_ns;
- }
-
/* Calculate delay from actual PPS to last_time */
delta.tv_nsec =
ptp->timeset[last_good].nanoseconds +
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 06/15] sfc: Tidy up PTP synchronization code
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (4 preceding siblings ...)
2013-12-14 23:02 ` [PATCH net-next 05/15] sfc: PTP - tidy up unused/useless variables Ben Hutchings
@ 2013-12-14 23:02 ` Ben Hutchings
2013-12-15 3:30 ` David Miller
2013-12-14 23:03 ` [PATCH net-next 07/15] sfc: Add support for SFC9100 timestamp format Ben Hutchings
` (9 subsequent siblings)
15 siblings, 1 reply; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:02 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
From: Laurence Evans <levans@solarflare.com>
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ptp.c | 44 +++++++++++++++++++++------------------
1 files changed, 24 insertions(+), 20 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index 82d8e27e8a58..361d7cedd897 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -435,7 +435,6 @@ efx_ptp_process_times(struct efx_nic *efx, MCDI_DECLARE_STRUCT_PTR(synch_buf),
MCDI_VAR_ARRAY_LEN(response_length,
PTP_OUT_SYNCHRONIZE_TIMESET);
unsigned i;
- unsigned total;
unsigned ngood = 0;
unsigned last_good = 0;
struct efx_ptp_data *ptp = efx->ptp_data;
@@ -446,33 +445,38 @@ efx_ptp_process_times(struct efx_nic *efx, MCDI_DECLARE_STRUCT_PTR(synch_buf),
if (number_readings == 0)
return -EAGAIN;
- /* Read the set of results and increment stats for any results that
- * appera to be erroneous.
+ /* Read the set of results and find the last good host-MC
+ * synchronization result. The MC times when it finishes reading the
+ * host time so the corrected window time should be fairly constant
+ * for a given platform.
*/
for (i = 0; i < number_readings; i++) {
+ s32 window, corrected;
+
efx_ptp_read_timeset(
MCDI_ARRAY_STRUCT_PTR(synch_buf,
PTP_OUT_SYNCHRONIZE_TIMESET, i),
&ptp->timeset[i]);
- }
- /* Find the last good host-MC synchronization result. The MC times
- * when it finishes reading the host time so the corrected window time
- * should be fairly constant for a given platform.
- */
- total = 0;
- for (i = 0; i < number_readings; i++)
- if (ptp->timeset[i].window > ptp->timeset[i].waitns) {
- unsigned win;
-
- win = ptp->timeset[i].window - ptp->timeset[i].waitns;
- if (win >= MIN_SYNCHRONISATION_NS &&
- win < MAX_SYNCHRONISATION_NS) {
- total += ptp->timeset[i].window;
- ngood++;
- last_good = i;
- }
+ window = ptp->timeset[i].window;
+ corrected = window - ptp->timeset[i].waitns;
+
+ /* We expect the uncorrected synchronization window to be at
+ * least as large as the interval between host start and end
+ * times. If it is smaller than this then this is mostly likely
+ * to be a consequence of the host's time being adjusted.
+ * Check that the corrected sync window is in a reasonable
+ * range. If it is out of range it is likely to be because an
+ * interrupt or other delay occurred between reading the system
+ * time and writing it to MC memory.
+ */
+ if (window >= SYNCHRONISATION_GRANULARITY_NS &&
+ corrected < MAX_SYNCHRONISATION_NS &&
+ corrected >= MIN_SYNCHRONISATION_NS) {
+ ngood++;
+ last_good = i;
}
+ }
if (ngood == 0) {
netif_warn(efx, drv, efx->net_dev,
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 07/15] sfc: Add support for SFC9100 timestamp format
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (5 preceding siblings ...)
2013-12-14 23:02 ` [PATCH net-next 06/15] sfc: Tidy up PTP synchronization code Ben Hutchings
@ 2013-12-14 23:03 ` Ben Hutchings
2013-12-14 23:03 ` [PATCH net-next 08/15] sfc: split setup of hardware timestamping into NIC-type operation Ben Hutchings
` (8 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:03 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
From: Laurence Evans <levans@solarflare.com>
The clock minor tick on the SFC9100 family is 2^-27 s, not 1 ns.
There are also various pipeline delays which we need to correct for
when interpreting timestamps.
We query the firmware for the clock format and corrections at run-time.
[bwh: Combined and rebased several changes]
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ptp.c | 256 ++++++++++++++++++++++++++++++++++++----
1 files changed, 231 insertions(+), 25 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index 361d7cedd897..cf2ae11b13f1 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -62,7 +62,7 @@
#define SYNCHRONISATION_GRANULARITY_NS 200
/* Minimum permitted length of a (corrected) synchronisation time */
-#define MIN_SYNCHRONISATION_NS 120
+#define DEFAULT_MIN_SYNCHRONISATION_NS 120
/* Maximum permitted length of a (corrected) synchronisation time */
#define MAX_SYNCHRONISATION_NS 1000
@@ -195,20 +195,20 @@ struct efx_ptp_event_rx {
/**
* struct efx_ptp_timeset - Synchronisation between host and MC
* @host_start: Host time immediately before hardware timestamp taken
- * @seconds: Hardware timestamp, seconds
- * @nanoseconds: Hardware timestamp, nanoseconds
+ * @major: Hardware timestamp, major
+ * @minor: Hardware timestamp, minor
* @host_end: Host time immediately after hardware timestamp taken
- * @waitns: Number of nanoseconds between hardware timestamp being read and
+ * @wait: Number of NIC clock ticks between hardware timestamp being read and
* host end time being seen
* @window: Difference of host_end and host_start
* @valid: Whether this timeset is valid
*/
struct efx_ptp_timeset {
u32 host_start;
- u32 seconds;
- u32 nanoseconds;
+ u32 major;
+ u32 minor;
u32 host_end;
- u32 waitns;
+ u32 wait;
u32 window; /* Derived: end - start, allowing for wrap */
};
@@ -232,6 +232,14 @@ struct efx_ptp_timeset {
* @config: Current timestamp configuration
* @enabled: PTP operation enabled
* @mode: Mode in which PTP operating (PTP version)
+ * @time_format: Time format supported by this NIC
+ * @ns_to_nic_time: Function to convert from scalar nanoseconds to NIC time
+ * @nic_to_kernel_time: Function to convert from NIC to kernel time
+ * @min_synchronisation_ns: Minimum acceptable corrected sync window
+ * @ts_corrections.tx: Required driver correction of transmit timestamps
+ * @ts_corrections.rx: Required driver correction of receive timestamps
+ * @ts_corrections.pps_out: PPS output error (information only)
+ * @ts_corrections.pps_in: Required driver correction of PPS input timestamps
* @evt_frags: Partly assembled PTP events
* @evt_frag_idx: Current fragment number
* @evt_code: Last event code
@@ -266,6 +274,17 @@ struct efx_ptp_data {
struct hwtstamp_config config;
bool enabled;
unsigned int mode;
+ unsigned int time_format;
+ void (*ns_to_nic_time)(s64 ns, u32 *nic_major, u32 *nic_minor);
+ ktime_t (*nic_to_kernel_time)(u32 nic_major, u32 nic_minor,
+ s32 correction);
+ unsigned int min_synchronisation_ns;
+ struct {
+ s32 tx;
+ s32 rx;
+ s32 pps_out;
+ s32 pps_in;
+ } ts_corrections;
efx_qword_t evt_frags[MAX_EVENT_FRAGS];
int evt_frag_idx;
int evt_code;
@@ -290,6 +309,167 @@ static int efx_phc_settime(struct ptp_clock_info *ptp,
static int efx_phc_enable(struct ptp_clock_info *ptp,
struct ptp_clock_request *request, int on);
+/* For Siena platforms NIC time is s and ns */
+static void efx_ptp_ns_to_s_ns(s64 ns, u32 *nic_major, u32 *nic_minor)
+{
+ struct timespec ts = ns_to_timespec(ns);
+ *nic_major = ts.tv_sec;
+ *nic_minor = ts.tv_nsec;
+}
+
+static ktime_t efx_ptp_s_ns_to_ktime(u32 nic_major, u32 nic_minor,
+ s32 correction)
+{
+ ktime_t kt = ktime_set(nic_major, nic_minor);
+ if (correction >= 0)
+ kt = ktime_add_ns(kt, (u64)correction);
+ else
+ kt = ktime_sub_ns(kt, (u64)-correction);
+ return kt;
+}
+
+/* To convert from s27 format to ns we multiply then divide by a power of 2.
+ * For the conversion from ns to s27, the operation is also converted to a
+ * multiply and shift.
+ */
+#define S27_TO_NS_SHIFT (27)
+#define NS_TO_S27_MULT (((1ULL << 63) + NSEC_PER_SEC / 2) / NSEC_PER_SEC)
+#define NS_TO_S27_SHIFT (63 - S27_TO_NS_SHIFT)
+#define S27_MINOR_MAX (1 << S27_TO_NS_SHIFT)
+
+/* For Huntington platforms NIC time is in seconds and fractions of a second
+ * where the minor register only uses 27 bits in units of 2^-27s.
+ */
+static void efx_ptp_ns_to_s27(s64 ns, u32 *nic_major, u32 *nic_minor)
+{
+ struct timespec ts = ns_to_timespec(ns);
+ u32 maj = ts.tv_sec;
+ u32 min = (u32)(((u64)ts.tv_nsec * NS_TO_S27_MULT +
+ (1ULL << (NS_TO_S27_SHIFT - 1))) >> NS_TO_S27_SHIFT);
+
+ /* The conversion can result in the minor value exceeding the maximum.
+ * In this case, round up to the next second.
+ */
+ if (min >= S27_MINOR_MAX) {
+ min -= S27_MINOR_MAX;
+ maj++;
+ }
+
+ *nic_major = maj;
+ *nic_minor = min;
+}
+
+static ktime_t efx_ptp_s27_to_ktime(u32 nic_major, u32 nic_minor,
+ s32 correction)
+{
+ u32 ns;
+
+ /* Apply the correction and deal with carry */
+ nic_minor += correction;
+ if ((s32)nic_minor < 0) {
+ nic_minor += S27_MINOR_MAX;
+ nic_major--;
+ } else if (nic_minor >= S27_MINOR_MAX) {
+ nic_minor -= S27_MINOR_MAX;
+ nic_major++;
+ }
+
+ ns = (u32)(((u64)nic_minor * NSEC_PER_SEC +
+ (1ULL << (S27_TO_NS_SHIFT - 1))) >> S27_TO_NS_SHIFT);
+
+ return ktime_set(nic_major, ns);
+}
+
+/* Get PTP attributes and set up time conversions */
+static int efx_ptp_get_attributes(struct efx_nic *efx)
+{
+ MCDI_DECLARE_BUF(inbuf, MC_CMD_PTP_IN_GET_ATTRIBUTES_LEN);
+ MCDI_DECLARE_BUF(outbuf, MC_CMD_PTP_OUT_GET_ATTRIBUTES_LEN);
+ struct efx_ptp_data *ptp = efx->ptp_data;
+ int rc;
+ u32 fmt;
+ size_t out_len;
+
+ /* Get the PTP attributes. If the NIC doesn't support the operation we
+ * use the default format for compatibility with older NICs i.e.
+ * seconds and nanoseconds.
+ */
+ MCDI_SET_DWORD(inbuf, PTP_IN_OP, MC_CMD_PTP_OP_GET_ATTRIBUTES);
+ MCDI_SET_DWORD(inbuf, PTP_IN_PERIPH_ID, 0);
+ rc = efx_mcdi_rpc(efx, MC_CMD_PTP, inbuf, sizeof(inbuf),
+ outbuf, sizeof(outbuf), &out_len);
+ if (rc == 0)
+ fmt = MCDI_DWORD(outbuf, PTP_OUT_GET_ATTRIBUTES_TIME_FORMAT);
+ else if (rc == -EINVAL)
+ fmt = MC_CMD_PTP_OUT_GET_ATTRIBUTES_SECONDS_NANOSECONDS;
+ else
+ return rc;
+
+ if (fmt == MC_CMD_PTP_OUT_GET_ATTRIBUTES_SECONDS_27FRACTION) {
+ ptp->ns_to_nic_time = efx_ptp_ns_to_s27;
+ ptp->nic_to_kernel_time = efx_ptp_s27_to_ktime;
+ } else if (fmt == MC_CMD_PTP_OUT_GET_ATTRIBUTES_SECONDS_NANOSECONDS) {
+ ptp->ns_to_nic_time = efx_ptp_ns_to_s_ns;
+ ptp->nic_to_kernel_time = efx_ptp_s_ns_to_ktime;
+ } else {
+ return -ERANGE;
+ }
+
+ ptp->time_format = fmt;
+
+ /* MC_CMD_PTP_OP_GET_ATTRIBUTES is an extended version of an older
+ * operation MC_CMD_PTP_OP_GET_TIME_FORMAT that also returns a value
+ * to use for the minimum acceptable corrected synchronization window.
+ * If we have the extra information store it. For older firmware that
+ * does not implement the extended command use the default value.
+ */
+ if (rc == 0 && out_len >= MC_CMD_PTP_OUT_GET_ATTRIBUTES_LEN)
+ ptp->min_synchronisation_ns =
+ MCDI_DWORD(outbuf,
+ PTP_OUT_GET_ATTRIBUTES_SYNC_WINDOW_MIN);
+ else
+ ptp->min_synchronisation_ns = DEFAULT_MIN_SYNCHRONISATION_NS;
+
+ return 0;
+}
+
+/* Get PTP timestamp corrections */
+static int efx_ptp_get_timestamp_corrections(struct efx_nic *efx)
+{
+ MCDI_DECLARE_BUF(inbuf, MC_CMD_PTP_IN_GET_TIMESTAMP_CORRECTIONS_LEN);
+ MCDI_DECLARE_BUF(outbuf, MC_CMD_PTP_OUT_GET_TIMESTAMP_CORRECTIONS_LEN);
+ int rc;
+
+ /* Get the timestamp corrections from the NIC. If this operation is
+ * not supported (older NICs) then no correction is required.
+ */
+ MCDI_SET_DWORD(inbuf, PTP_IN_OP,
+ MC_CMD_PTP_OP_GET_TIMESTAMP_CORRECTIONS);
+ MCDI_SET_DWORD(inbuf, PTP_IN_PERIPH_ID, 0);
+
+ rc = efx_mcdi_rpc(efx, MC_CMD_PTP, inbuf, sizeof(inbuf),
+ outbuf, sizeof(outbuf), NULL);
+ if (rc == 0) {
+ efx->ptp_data->ts_corrections.tx = MCDI_DWORD(outbuf,
+ PTP_OUT_GET_TIMESTAMP_CORRECTIONS_TRANSMIT);
+ efx->ptp_data->ts_corrections.rx = MCDI_DWORD(outbuf,
+ PTP_OUT_GET_TIMESTAMP_CORRECTIONS_RECEIVE);
+ efx->ptp_data->ts_corrections.pps_out = MCDI_DWORD(outbuf,
+ PTP_OUT_GET_TIMESTAMP_CORRECTIONS_PPS_OUT);
+ efx->ptp_data->ts_corrections.pps_in = MCDI_DWORD(outbuf,
+ PTP_OUT_GET_TIMESTAMP_CORRECTIONS_PPS_IN);
+ } else if (rc == -EINVAL) {
+ efx->ptp_data->ts_corrections.tx = 0;
+ efx->ptp_data->ts_corrections.rx = 0;
+ efx->ptp_data->ts_corrections.pps_out = 0;
+ efx->ptp_data->ts_corrections.pps_in = 0;
+ } else {
+ return rc;
+ }
+
+ return 0;
+}
+
/* Enable MCDI PTP support. */
static int efx_ptp_enable(struct efx_nic *efx)
{
@@ -402,11 +582,10 @@ static void efx_ptp_read_timeset(MCDI_DECLARE_STRUCT_PTR(data),
unsigned start_ns, end_ns;
timeset->host_start = MCDI_DWORD(data, PTP_OUT_SYNCHRONIZE_HOSTSTART);
- timeset->seconds = MCDI_DWORD(data, PTP_OUT_SYNCHRONIZE_SECONDS);
- timeset->nanoseconds = MCDI_DWORD(data,
- PTP_OUT_SYNCHRONIZE_NANOSECONDS);
+ timeset->major = MCDI_DWORD(data, PTP_OUT_SYNCHRONIZE_MAJOR);
+ timeset->minor = MCDI_DWORD(data, PTP_OUT_SYNCHRONIZE_MINOR);
timeset->host_end = MCDI_DWORD(data, PTP_OUT_SYNCHRONIZE_HOSTEND),
- timeset->waitns = MCDI_DWORD(data, PTP_OUT_SYNCHRONIZE_WAITNS);
+ timeset->wait = MCDI_DWORD(data, PTP_OUT_SYNCHRONIZE_WAITNS);
/* Ignore seconds */
start_ns = timeset->host_start & MC_NANOSECOND_MASK;
@@ -441,6 +620,7 @@ efx_ptp_process_times(struct efx_nic *efx, MCDI_DECLARE_STRUCT_PTR(synch_buf),
u32 last_sec;
u32 start_sec;
struct timespec delta;
+ ktime_t mc_time;
if (number_readings == 0)
return -EAGAIN;
@@ -452,14 +632,17 @@ efx_ptp_process_times(struct efx_nic *efx, MCDI_DECLARE_STRUCT_PTR(synch_buf),
*/
for (i = 0; i < number_readings; i++) {
s32 window, corrected;
+ struct timespec wait;
efx_ptp_read_timeset(
MCDI_ARRAY_STRUCT_PTR(synch_buf,
PTP_OUT_SYNCHRONIZE_TIMESET, i),
&ptp->timeset[i]);
+ wait = ktime_to_timespec(
+ ptp->nic_to_kernel_time(0, ptp->timeset[i].wait, 0));
window = ptp->timeset[i].window;
- corrected = window - ptp->timeset[i].waitns;
+ corrected = window - wait.tv_nsec;
/* We expect the uncorrected synchronization window to be at
* least as large as the interval between host start and end
@@ -472,7 +655,7 @@ efx_ptp_process_times(struct efx_nic *efx, MCDI_DECLARE_STRUCT_PTR(synch_buf),
*/
if (window >= SYNCHRONISATION_GRANULARITY_NS &&
corrected < MAX_SYNCHRONISATION_NS &&
- corrected >= MIN_SYNCHRONISATION_NS) {
+ corrected >= ptp->min_synchronisation_ns) {
ngood++;
last_good = i;
}
@@ -484,9 +667,15 @@ efx_ptp_process_times(struct efx_nic *efx, MCDI_DECLARE_STRUCT_PTR(synch_buf),
return -EAGAIN;
}
+ /* Convert the NIC time into kernel time. No correction is required-
+ * this time is the output of a firmware process.
+ */
+ mc_time = ptp->nic_to_kernel_time(ptp->timeset[last_good].major,
+ ptp->timeset[last_good].minor, 0);
+
/* Calculate delay from actual PPS to last_time */
- delta.tv_nsec =
- ptp->timeset[last_good].nanoseconds +
+ delta = ktime_to_timespec(mc_time);
+ delta.tv_nsec +=
last_time->ts_real.tv_nsec -
(ptp->timeset[last_good].host_start & MC_NANOSECOND_MASK);
@@ -596,9 +785,10 @@ static int efx_ptp_xmit_skb(struct efx_nic *efx, struct sk_buff *skb)
goto fail;
memset(×tamps, 0, sizeof(timestamps));
- timestamps.hwtstamp = ktime_set(
- MCDI_DWORD(txtime, PTP_OUT_TRANSMIT_SECONDS),
- MCDI_DWORD(txtime, PTP_OUT_TRANSMIT_NANOSECONDS));
+ timestamps.hwtstamp = ptp_data->nic_to_kernel_time(
+ MCDI_DWORD(txtime, PTP_OUT_TRANSMIT_MAJOR),
+ MCDI_DWORD(txtime, PTP_OUT_TRANSMIT_MINOR),
+ ptp_data->ts_corrections.tx);
skb_tstamp_tx(skb, ×tamps);
@@ -954,6 +1144,16 @@ int efx_ptp_probe(struct efx_nic *efx, struct efx_channel *channel)
list_add(&ptp->rx_evts[pos].link, &ptp->evt_free_list);
ptp->evt_overflow = false;
+ /* Get the NIC PTP attributes and set up time conversions */
+ rc = efx_ptp_get_attributes(efx);
+ if (rc < 0)
+ goto fail3;
+
+ /* Get the timestamp corrections */
+ rc = efx_ptp_get_timestamp_corrections(efx);
+ if (rc < 0)
+ goto fail3;
+
ptp->phc_clock_info = efx_phc_clock_info;
ptp->phc_clock = ptp_clock_register(&ptp->phc_clock_info,
&efx->pci_dev->dev);
@@ -1358,9 +1558,10 @@ static void ptp_event_rx(struct efx_nic *efx, struct efx_ptp_data *ptp)
MCDI_EVENT_SRC) << 8) |
(EFX_QWORD_FIELD(ptp->evt_frags[0],
MCDI_EVENT_SRC) << 16));
- evt->hwtimestamp = ktime_set(
+ evt->hwtimestamp = efx->ptp_data->nic_to_kernel_time(
EFX_QWORD_FIELD(ptp->evt_frags[0], MCDI_EVENT_DATA),
- EFX_QWORD_FIELD(ptp->evt_frags[1], MCDI_EVENT_DATA));
+ EFX_QWORD_FIELD(ptp->evt_frags[1], MCDI_EVENT_DATA),
+ ptp->ts_corrections.rx);
evt->expiry = jiffies + msecs_to_jiffies(PKT_EVENT_LIFETIME_MS);
list_add_tail(&evt->link, &ptp->evt_list);
@@ -1470,18 +1671,20 @@ static int efx_phc_adjfreq(struct ptp_clock_info *ptp, s32 delta)
static int efx_phc_adjtime(struct ptp_clock_info *ptp, s64 delta)
{
+ u32 nic_major, nic_minor;
struct efx_ptp_data *ptp_data = container_of(ptp,
struct efx_ptp_data,
phc_clock_info);
struct efx_nic *efx = ptp_data->efx;
- struct timespec delta_ts = ns_to_timespec(delta);
MCDI_DECLARE_BUF(inbuf, MC_CMD_PTP_IN_ADJUST_LEN);
+ efx->ptp_data->ns_to_nic_time(delta, &nic_major, &nic_minor);
+
MCDI_SET_DWORD(inbuf, PTP_IN_OP, MC_CMD_PTP_OP_ADJUST);
MCDI_SET_DWORD(inbuf, PTP_IN_PERIPH_ID, 0);
MCDI_SET_QWORD(inbuf, PTP_IN_ADJUST_FREQ, ptp_data->current_adjfreq);
- MCDI_SET_DWORD(inbuf, PTP_IN_ADJUST_SECONDS, (u32)delta_ts.tv_sec);
- MCDI_SET_DWORD(inbuf, PTP_IN_ADJUST_NANOSECONDS, (u32)delta_ts.tv_nsec);
+ MCDI_SET_DWORD(inbuf, PTP_IN_ADJUST_MAJOR, nic_major);
+ MCDI_SET_DWORD(inbuf, PTP_IN_ADJUST_MINOR, nic_minor);
return efx_mcdi_rpc(efx, MC_CMD_PTP, inbuf, sizeof(inbuf),
NULL, 0, NULL);
}
@@ -1495,6 +1698,7 @@ static int efx_phc_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
MCDI_DECLARE_BUF(inbuf, MC_CMD_PTP_IN_READ_NIC_TIME_LEN);
MCDI_DECLARE_BUF(outbuf, MC_CMD_PTP_OUT_READ_NIC_TIME_LEN);
int rc;
+ ktime_t kt;
MCDI_SET_DWORD(inbuf, PTP_IN_OP, MC_CMD_PTP_OP_READ_NIC_TIME);
MCDI_SET_DWORD(inbuf, PTP_IN_PERIPH_ID, 0);
@@ -1504,8 +1708,10 @@ static int efx_phc_gettime(struct ptp_clock_info *ptp, struct timespec *ts)
if (rc != 0)
return rc;
- ts->tv_sec = MCDI_DWORD(outbuf, PTP_OUT_READ_NIC_TIME_SECONDS);
- ts->tv_nsec = MCDI_DWORD(outbuf, PTP_OUT_READ_NIC_TIME_NANOSECONDS);
+ kt = ptp_data->nic_to_kernel_time(
+ MCDI_DWORD(outbuf, PTP_OUT_READ_NIC_TIME_MAJOR),
+ MCDI_DWORD(outbuf, PTP_OUT_READ_NIC_TIME_MINOR), 0);
+ *ts = ktime_to_timespec(kt);
return 0;
}
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 08/15] sfc: split setup of hardware timestamping into NIC-type operation
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (6 preceding siblings ...)
2013-12-14 23:03 ` [PATCH net-next 07/15] sfc: Add support for SFC9100 timestamp format Ben Hutchings
@ 2013-12-14 23:03 ` Ben Hutchings
2013-12-14 23:04 ` [PATCH net-next 09/15] sfc: Copy RX prefix into skb head area in efx_rx_mk_skb() Ben Hutchings
` (7 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:03 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
From: Daniel Pieczko <dpieczko@solarflare.com>
I added efx_ptp_get_mode() to avoid moving the definition for
efx_ptp_data, since the current PTP mode is needed for
siena.c:siena_set_ptp_hwtstamp.
[bwh: Also move the rx_filters mask, and add kernel-doc]
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/net_driver.h | 9 ++++
drivers/net/ethernet/sfc/nic.h | 3 +
drivers/net/ethernet/sfc/ptp.c | 69 +++++----------------------------
drivers/net/ethernet/sfc/siena.c | 69 ++++++++++++++++++++++++++------
4 files changed, 78 insertions(+), 72 deletions(-)
diff --git a/drivers/net/ethernet/sfc/net_driver.h b/drivers/net/ethernet/sfc/net_driver.h
index f47bac78b92c..01ffbe849683 100644
--- a/drivers/net/ethernet/sfc/net_driver.h
+++ b/drivers/net/ethernet/sfc/net_driver.h
@@ -91,6 +91,7 @@
/* Forward declare Precision Time Protocol (PTP) support structure. */
struct efx_ptp_data;
+struct hwtstamp_config;
struct efx_self_tests;
@@ -1042,6 +1043,10 @@ struct efx_mtd_partition {
* @mtd_sync: Wait for write-back to complete on MTD partition. This
* also notifies the driver that a writer has finished using this
* partition.
+ * @ptp_write_host_time: Send host time to MC as part of sync protocol
+ * @ptp_set_ts_config: Set hardware timestamp configuration. The flags
+ * and tx_type will already have been validated but this operation
+ * must validate and update rx_filter.
* @revision: Hardware architecture revision
* @txd_ptr_tbl_base: TX descriptor ring base address
* @rxd_ptr_tbl_base: RX descriptor ring base address
@@ -1060,6 +1065,7 @@ struct efx_mtd_partition {
* @offload_features: net_device feature flags for protocol offload
* features implemented in hardware
* @mcdi_max_ver: Maximum MCDI version supported
+ * @hwtstamp_filters: Mask of hardware timestamp filter types supported
*/
struct efx_nic_type {
unsigned int (*mem_map_size)(struct efx_nic *efx);
@@ -1161,6 +1167,8 @@ struct efx_nic_type {
int (*mtd_sync)(struct mtd_info *mtd);
#endif
void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
+ int (*ptp_set_ts_config)(struct efx_nic *efx,
+ struct hwtstamp_config *init);
int revision;
unsigned int txd_ptr_tbl_base;
@@ -1179,6 +1187,7 @@ struct efx_nic_type {
netdev_features_t offload_features;
int mcdi_max_ver;
unsigned int max_rx_ip_filters;
+ u32 hwtstamp_filters;
};
/**************************************************************************
diff --git a/drivers/net/ethernet/sfc/nic.h b/drivers/net/ethernet/sfc/nic.h
index 25178536f053..999ef285662d 100644
--- a/drivers/net/ethernet/sfc/nic.h
+++ b/drivers/net/ethernet/sfc/nic.h
@@ -561,6 +561,9 @@ int efx_ptp_set_ts_config(struct efx_nic *efx, struct ifreq *ifr);
int efx_ptp_get_ts_config(struct efx_nic *efx, struct ifreq *ifr);
void efx_ptp_get_ts_info(struct efx_nic *efx, struct ethtool_ts_info *ts_info);
bool efx_ptp_is_ptp_tx(struct efx_nic *efx, struct sk_buff *skb);
+int efx_ptp_get_mode(struct efx_nic *efx);
+int efx_ptp_change_mode(struct efx_nic *efx, bool enable_wanted,
+ unsigned int new_mode);
int efx_ptp_tx(struct efx_nic *efx, struct sk_buff *skb);
void efx_ptp_event(struct efx_nic *efx, efx_qword_t *ev);
void efx_ptp_start_datapath(struct efx_nic *efx);
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index cf2ae11b13f1..1d1a6f7325da 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -1361,8 +1361,13 @@ int efx_ptp_tx(struct efx_nic *efx, struct sk_buff *skb)
return NETDEV_TX_OK;
}
-static int efx_ptp_change_mode(struct efx_nic *efx, bool enable_wanted,
- unsigned int new_mode)
+int efx_ptp_get_mode(struct efx_nic *efx)
+{
+ return efx->ptp_data->mode;
+}
+
+int efx_ptp_change_mode(struct efx_nic *efx, bool enable_wanted,
+ unsigned int new_mode)
{
if ((enable_wanted != efx->ptp_data->enabled) ||
(enable_wanted && (efx->ptp_data->mode != new_mode))) {
@@ -1406,8 +1411,6 @@ static int efx_ptp_change_mode(struct efx_nic *efx, bool enable_wanted,
static int efx_ptp_ts_init(struct efx_nic *efx, struct hwtstamp_config *init)
{
- bool enable_wanted = false;
- unsigned int new_mode;
int rc;
if (init->flags)
@@ -1417,57 +1420,11 @@ static int efx_ptp_ts_init(struct efx_nic *efx, struct hwtstamp_config *init)
(init->tx_type != HWTSTAMP_TX_ON))
return -ERANGE;
- new_mode = efx->ptp_data->mode;
- /* Determine whether any PTP HW operations are required */
- switch (init->rx_filter) {
- case HWTSTAMP_FILTER_NONE:
- break;
- case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
- case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
- case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
- init->rx_filter = HWTSTAMP_FILTER_PTP_V1_L4_EVENT;
- new_mode = MC_CMD_PTP_MODE_V1;
- enable_wanted = true;
- break;
- case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
- case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
- case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
- /* Although these three are accepted only IPV4 packets will be
- * timestamped
- */
- init->rx_filter = HWTSTAMP_FILTER_PTP_V2_L4_EVENT;
- new_mode = MC_CMD_PTP_MODE_V2_ENHANCED;
- enable_wanted = true;
- break;
- case HWTSTAMP_FILTER_PTP_V2_EVENT:
- case HWTSTAMP_FILTER_PTP_V2_SYNC:
- case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
- case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
- case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
- case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
- /* Non-IP + IPv6 timestamping not supported */
- return -ERANGE;
- break;
- default:
- return -ERANGE;
- }
-
- if (init->tx_type != HWTSTAMP_TX_OFF)
- enable_wanted = true;
-
- /* Old versions of the firmware do not support the improved
- * UUID filtering option (SF bug 33070). If the firmware does
- * not accept the enhanced mode, fall back to the standard PTP
- * v2 UUID filtering.
- */
- rc = efx_ptp_change_mode(efx, enable_wanted, new_mode);
- if ((rc != 0) && (new_mode == MC_CMD_PTP_MODE_V2_ENHANCED))
- rc = efx_ptp_change_mode(efx, enable_wanted, MC_CMD_PTP_MODE_V2);
- if (rc != 0)
+ rc = efx->type->ptp_set_ts_config(efx, init);
+ if (rc)
return rc;
efx->ptp_data->config = *init;
-
return 0;
}
@@ -1483,13 +1440,7 @@ void efx_ptp_get_ts_info(struct efx_nic *efx, struct ethtool_ts_info *ts_info)
SOF_TIMESTAMPING_RAW_HARDWARE);
ts_info->phc_index = ptp_clock_index(ptp->phc_clock);
ts_info->tx_types = 1 << HWTSTAMP_TX_OFF | 1 << HWTSTAMP_TX_ON;
- ts_info->rx_filters = (1 << HWTSTAMP_FILTER_NONE |
- 1 << HWTSTAMP_FILTER_PTP_V1_L4_EVENT |
- 1 << HWTSTAMP_FILTER_PTP_V1_L4_SYNC |
- 1 << HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ |
- 1 << HWTSTAMP_FILTER_PTP_V2_L4_EVENT |
- 1 << HWTSTAMP_FILTER_PTP_V2_L4_SYNC |
- 1 << HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ);
+ ts_info->rx_filters = ptp->efx->type->hwtstamp_filters;
}
int efx_ptp_set_ts_config(struct efx_nic *efx, struct ifreq *ifr)
diff --git a/drivers/net/ethernet/sfc/siena.c b/drivers/net/ethernet/sfc/siena.c
index 53b4ce67be0e..f65db356fe09 100644
--- a/drivers/net/ethernet/sfc/siena.c
+++ b/drivers/net/ethernet/sfc/siena.c
@@ -118,6 +118,54 @@ out:
/**************************************************************************
*
+ * PTP
+ *
+ **************************************************************************
+ */
+
+static void siena_ptp_write_host_time(struct efx_nic *efx, u32 host_time)
+{
+ _efx_writed(efx, cpu_to_le32(host_time),
+ FR_CZ_MC_TREG_SMEM + MC_SMEM_P0_PTP_TIME_OFST);
+}
+
+static int siena_ptp_set_ts_config(struct efx_nic *efx,
+ struct hwtstamp_config *init)
+{
+ int rc;
+
+ switch (init->rx_filter) {
+ case HWTSTAMP_FILTER_NONE:
+ /* if TX timestamping is still requested then leave PTP on */
+ return efx_ptp_change_mode(efx,
+ init->tx_type != HWTSTAMP_TX_OFF,
+ efx_ptp_get_mode(efx));
+ case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
+ case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
+ case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
+ init->rx_filter = HWTSTAMP_FILTER_PTP_V1_L4_EVENT;
+ return efx_ptp_change_mode(efx, true, MC_CMD_PTP_MODE_V1);
+ case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
+ init->rx_filter = HWTSTAMP_FILTER_PTP_V2_L4_EVENT;
+ rc = efx_ptp_change_mode(efx, true,
+ MC_CMD_PTP_MODE_V2_ENHANCED);
+ /* bug 33070 - old versions of the firmware do not support the
+ * improved UUID filtering option. Similarly old versions of the
+ * application do not expect it to be enabled. If the firmware
+ * does not accept the enhanced mode, fall back to the standard
+ * PTP v2 UUID filtering. */
+ if (rc != 0)
+ rc = efx_ptp_change_mode(efx, true, MC_CMD_PTP_MODE_V2);
+ return rc;
+ default:
+ return -ERANGE;
+ }
+}
+
+/**************************************************************************
+ *
* Device reset
*
**************************************************************************
@@ -839,19 +887,6 @@ fail:
/**************************************************************************
*
- * PTP
- *
- **************************************************************************
- */
-
-static void siena_ptp_write_host_time(struct efx_nic *efx, u32 host_time)
-{
- _efx_writed(efx, cpu_to_le32(host_time),
- FR_CZ_MC_TREG_SMEM + MC_SMEM_P0_PTP_TIME_OFST);
-}
-
-/**************************************************************************
- *
* Revision-dependent attributes used by efx.c and nic.c
*
**************************************************************************
@@ -942,6 +977,7 @@ const struct efx_nic_type siena_a0_nic_type = {
.mtd_sync = efx_mcdi_mtd_sync,
#endif
.ptp_write_host_time = siena_ptp_write_host_time,
+ .ptp_set_ts_config = siena_ptp_set_ts_config,
.revision = EFX_REV_SIENA_A0,
.txd_ptr_tbl_base = FR_BZ_TX_DESC_PTR_TBL,
@@ -960,4 +996,11 @@ const struct efx_nic_type siena_a0_nic_type = {
NETIF_F_RXHASH | NETIF_F_NTUPLE),
.mcdi_max_ver = 1,
.max_rx_ip_filters = FR_BZ_RX_FILTER_TBL0_ROWS,
+ .hwtstamp_filters = (1 << HWTSTAMP_FILTER_NONE |
+ 1 << HWTSTAMP_FILTER_PTP_V1_L4_EVENT |
+ 1 << HWTSTAMP_FILTER_PTP_V1_L4_SYNC |
+ 1 << HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ |
+ 1 << HWTSTAMP_FILTER_PTP_V2_L4_EVENT |
+ 1 << HWTSTAMP_FILTER_PTP_V2_L4_SYNC |
+ 1 << HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ),
};
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 09/15] sfc: Copy RX prefix into skb head area in efx_rx_mk_skb()
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (7 preceding siblings ...)
2013-12-14 23:03 ` [PATCH net-next 08/15] sfc: split setup of hardware timestamping into NIC-type operation Ben Hutchings
@ 2013-12-14 23:04 ` Ben Hutchings
2013-12-14 23:05 ` [PATCH net-next 10/15] sfc: Add RX packet timestamping for EF10 Ben Hutchings
` (6 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:04 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
We can potentially pull the entire packet contents into the head area
and then free the page it was in. In order to read an inline
timestamp safely, we need to copy the prefix into the head area as
well.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/net_driver.h | 9 ---------
drivers/net/ethernet/sfc/rx.c | 10 +++++++---
2 files changed, 7 insertions(+), 12 deletions(-)
diff --git a/drivers/net/ethernet/sfc/net_driver.h b/drivers/net/ethernet/sfc/net_driver.h
index 01ffbe849683..1a3e4972c68f 100644
--- a/drivers/net/ethernet/sfc/net_driver.h
+++ b/drivers/net/ethernet/sfc/net_driver.h
@@ -521,15 +521,6 @@ enum nic_state {
STATE_RECOVERY = 3, /* device recovering from PCI error */
};
-/*
- * Alignment of the skb->head which wraps a page-allocated RX buffer
- *
- * The skb allocated to wrap an rx_buffer can have this alignment. Since
- * the data is memcpy'd from the rx_buf, it does not need to be equal to
- * NET_IP_ALIGN.
- */
-#define EFX_PAGE_SKB_ALIGN 2
-
/* Forward declaration */
struct efx_nic;
diff --git a/drivers/net/ethernet/sfc/rx.c b/drivers/net/ethernet/sfc/rx.c
index 8671bc199a9d..d12abc543975 100644
--- a/drivers/net/ethernet/sfc/rx.c
+++ b/drivers/net/ethernet/sfc/rx.c
@@ -476,14 +476,18 @@ static struct sk_buff *efx_rx_mk_skb(struct efx_channel *channel,
struct sk_buff *skb;
/* Allocate an SKB to store the headers */
- skb = netdev_alloc_skb(efx->net_dev, hdr_len + EFX_PAGE_SKB_ALIGN);
+ skb = netdev_alloc_skb(efx->net_dev,
+ efx->rx_ip_align + efx->rx_prefix_size +
+ hdr_len);
if (unlikely(skb == NULL))
return NULL;
EFX_BUG_ON_PARANOID(rx_buf->len < hdr_len);
- skb_reserve(skb, EFX_PAGE_SKB_ALIGN);
- memcpy(__skb_put(skb, hdr_len), eh, hdr_len);
+ memcpy(skb->data + efx->rx_ip_align, eh - efx->rx_prefix_size,
+ efx->rx_prefix_size + hdr_len);
+ skb_reserve(skb, efx->rx_ip_align + efx->rx_prefix_size);
+ __skb_put(skb, hdr_len);
/* Append the remaining page(s) onto the frag list */
if (rx_buf->len > hdr_len) {
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 10/15] sfc: Add RX packet timestamping for EF10
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (8 preceding siblings ...)
2013-12-14 23:04 ` [PATCH net-next 09/15] sfc: Copy RX prefix into skb head area in efx_rx_mk_skb() Ben Hutchings
@ 2013-12-14 23:05 ` Ben Hutchings
2013-12-14 23:05 ` [PATCH net-next 11/15] sfc: Store VPD serial number at probe time Ben Hutchings
` (5 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:05 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
From: Jon Cooper <jcooper@solarflare.com>
The EF10 firmware can optionally insert RX timestamps in the packet
prefix. These only include the clock minor value. We must also
enable periodic time sync events on each event queue which provide
the high bits of the clock value.
[bwh: Combined and rebased several changes.
Added the above description and some sanity checks for inline vs
separate timestamps.
Changed efx_rx_skb_attach_timestamp() to read the packet prefix
from the skb head area.]
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ef10.c | 123 ++++++++++++++++++++++++++++++-
drivers/net/ethernet/sfc/efx.c | 2 +
drivers/net/ethernet/sfc/mcdi.c | 3 +
drivers/net/ethernet/sfc/net_driver.h | 22 ++++++
drivers/net/ethernet/sfc/nic.h | 9 ++
drivers/net/ethernet/sfc/ptp.c | 131 +++++++++++++++++++++++++++++---
drivers/net/ethernet/sfc/rx.c | 2 +
7 files changed, 277 insertions(+), 15 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ef10.c b/drivers/net/ethernet/sfc/ef10.c
index ca922b958a2e..a3d891764030 100644
--- a/drivers/net/ethernet/sfc/ef10.c
+++ b/drivers/net/ethernet/sfc/ef10.c
@@ -1469,8 +1469,9 @@ static void efx_ef10_rx_init(struct efx_rx_queue *rx_queue)
MCDI_SET_DWORD(inbuf, INIT_RXQ_IN_LABEL, efx_rx_queue_index(rx_queue));
MCDI_SET_DWORD(inbuf, INIT_RXQ_IN_INSTANCE,
efx_rx_queue_index(rx_queue));
- MCDI_POPULATE_DWORD_1(inbuf, INIT_RXQ_IN_FLAGS,
- INIT_RXQ_IN_FLAG_PREFIX, 1);
+ MCDI_POPULATE_DWORD_2(inbuf, INIT_RXQ_IN_FLAGS,
+ INIT_RXQ_IN_FLAG_PREFIX, 1,
+ INIT_RXQ_IN_FLAG_TIMESTAMP, 1);
MCDI_SET_DWORD(inbuf, INIT_RXQ_IN_OWNER_ID, 0);
MCDI_SET_DWORD(inbuf, INIT_RXQ_IN_PORT_ID, EVB_PORT_ID_ASSIGNED);
@@ -3406,6 +3407,119 @@ static void efx_ef10_ptp_write_host_time(struct efx_nic *efx, u32 host_time)
_efx_writed(efx, cpu_to_le32(host_time), ER_DZ_MC_DB_LWRD);
}
+static int efx_ef10_rx_enable_timestamping(struct efx_channel *channel,
+ bool temp)
+{
+ MCDI_DECLARE_BUF(inbuf, MC_CMD_PTP_IN_TIME_EVENT_SUBSCRIBE_LEN);
+ int rc;
+
+ if (channel->sync_events_state == SYNC_EVENTS_REQUESTED ||
+ channel->sync_events_state == SYNC_EVENTS_VALID ||
+ (temp && channel->sync_events_state == SYNC_EVENTS_DISABLED))
+ return 0;
+ channel->sync_events_state = SYNC_EVENTS_REQUESTED;
+
+ MCDI_SET_DWORD(inbuf, PTP_IN_OP, MC_CMD_PTP_OP_TIME_EVENT_SUBSCRIBE);
+ MCDI_SET_DWORD(inbuf, PTP_IN_PERIPH_ID, 0);
+ MCDI_SET_DWORD(inbuf, PTP_IN_TIME_EVENT_SUBSCRIBE_QUEUE,
+ channel->channel);
+
+ rc = efx_mcdi_rpc(channel->efx, MC_CMD_PTP,
+ inbuf, sizeof(inbuf), NULL, 0, NULL);
+
+ if (rc != 0)
+ channel->sync_events_state = temp ? SYNC_EVENTS_QUIESCENT :
+ SYNC_EVENTS_DISABLED;
+
+ return rc;
+}
+
+static int efx_ef10_rx_disable_timestamping(struct efx_channel *channel,
+ bool temp)
+{
+ MCDI_DECLARE_BUF(inbuf, MC_CMD_PTP_IN_TIME_EVENT_UNSUBSCRIBE_LEN);
+ int rc;
+
+ if (channel->sync_events_state == SYNC_EVENTS_DISABLED ||
+ (temp && channel->sync_events_state == SYNC_EVENTS_QUIESCENT))
+ return 0;
+ if (channel->sync_events_state == SYNC_EVENTS_QUIESCENT) {
+ channel->sync_events_state = SYNC_EVENTS_DISABLED;
+ return 0;
+ }
+ channel->sync_events_state = temp ? SYNC_EVENTS_QUIESCENT :
+ SYNC_EVENTS_DISABLED;
+
+ MCDI_SET_DWORD(inbuf, PTP_IN_OP, MC_CMD_PTP_OP_TIME_EVENT_UNSUBSCRIBE);
+ MCDI_SET_DWORD(inbuf, PTP_IN_PERIPH_ID, 0);
+ MCDI_SET_DWORD(inbuf, PTP_IN_TIME_EVENT_UNSUBSCRIBE_CONTROL,
+ MC_CMD_PTP_IN_TIME_EVENT_UNSUBSCRIBE_SINGLE);
+ MCDI_SET_DWORD(inbuf, PTP_IN_TIME_EVENT_UNSUBSCRIBE_QUEUE,
+ channel->channel);
+
+ rc = efx_mcdi_rpc(channel->efx, MC_CMD_PTP,
+ inbuf, sizeof(inbuf), NULL, 0, NULL);
+
+ return rc;
+}
+
+static int efx_ef10_ptp_set_ts_sync_events(struct efx_nic *efx, bool en,
+ bool temp)
+{
+ int (*set)(struct efx_channel *channel, bool temp);
+ struct efx_channel *channel;
+
+ set = en ?
+ efx_ef10_rx_enable_timestamping :
+ efx_ef10_rx_disable_timestamping;
+
+ efx_for_each_channel(channel, efx) {
+ int rc = set(channel, temp);
+ if (en && rc != 0) {
+ efx_ef10_ptp_set_ts_sync_events(efx, false, temp);
+ return rc;
+ }
+ }
+
+ return 0;
+}
+
+static int efx_ef10_ptp_set_ts_config(struct efx_nic *efx,
+ struct hwtstamp_config *init)
+{
+ int rc;
+
+ switch (init->rx_filter) {
+ case HWTSTAMP_FILTER_NONE:
+ efx_ef10_ptp_set_ts_sync_events(efx, false, false);
+ /* if TX timestamping is still requested then leave PTP on */
+ return efx_ptp_change_mode(efx,
+ init->tx_type != HWTSTAMP_TX_OFF, 0);
+ case HWTSTAMP_FILTER_ALL:
+ case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
+ case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
+ case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
+ case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
+ case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
+ case HWTSTAMP_FILTER_PTP_V2_EVENT:
+ case HWTSTAMP_FILTER_PTP_V2_SYNC:
+ case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
+ init->rx_filter = HWTSTAMP_FILTER_ALL;
+ rc = efx_ptp_change_mode(efx, true, 0);
+ if (!rc)
+ rc = efx_ef10_ptp_set_ts_sync_events(efx, true, false);
+ if (rc)
+ efx_ptp_change_mode(efx, false, 0);
+ return rc;
+ default:
+ return -ERANGE;
+ }
+}
+
const struct efx_nic_type efx_hunt_a0_nic_type = {
.mem_map_size = efx_ef10_mem_map_size,
.probe = efx_ef10_probe,
@@ -3484,11 +3598,14 @@ const struct efx_nic_type efx_hunt_a0_nic_type = {
.mtd_sync = efx_mcdi_mtd_sync,
#endif
.ptp_write_host_time = efx_ef10_ptp_write_host_time,
+ .ptp_set_ts_sync_events = efx_ef10_ptp_set_ts_sync_events,
+ .ptp_set_ts_config = efx_ef10_ptp_set_ts_config,
.revision = EFX_REV_HUNT_A0,
.max_dma_mask = DMA_BIT_MASK(ESF_DZ_TX_KER_BUF_ADDR_WIDTH),
.rx_prefix_size = ES_DZ_RX_PREFIX_SIZE,
.rx_hash_offset = ES_DZ_RX_PREFIX_HASH_OFST,
+ .rx_ts_offset = ES_DZ_RX_PREFIX_TSTAMP_OFST,
.can_rx_scatter = true,
.always_rx_scatter = true,
.max_interrupt_mode = EFX_INT_MODE_MSIX,
@@ -3497,4 +3614,6 @@ const struct efx_nic_type efx_hunt_a0_nic_type = {
NETIF_F_RXHASH | NETIF_F_NTUPLE),
.mcdi_max_ver = 2,
.max_rx_ip_filters = HUNT_FILTER_TBL_ROWS,
+ .hwtstamp_filters = 1 << HWTSTAMP_FILTER_NONE |
+ 1 << HWTSTAMP_FILTER_ALL,
};
diff --git a/drivers/net/ethernet/sfc/efx.c b/drivers/net/ethernet/sfc/efx.c
index 893cd78c3a6c..b9929d9ed69d 100644
--- a/drivers/net/ethernet/sfc/efx.c
+++ b/drivers/net/ethernet/sfc/efx.c
@@ -2586,6 +2586,8 @@ static int efx_init_struct(struct efx_nic *efx,
NET_IP_ALIGN ? (efx->rx_prefix_size + NET_IP_ALIGN) % 4 : 0;
efx->rx_packet_hash_offset =
efx->type->rx_hash_offset - efx->type->rx_prefix_size;
+ efx->rx_packet_ts_offset =
+ efx->type->rx_ts_offset - efx->type->rx_prefix_size;
spin_lock_init(&efx->stats_lock);
mutex_init(&efx->mac_lock);
efx->phy_op = &efx_dummy_phy_operations;
diff --git a/drivers/net/ethernet/sfc/mcdi.c b/drivers/net/ethernet/sfc/mcdi.c
index d608838f6729..540f57915d6f 100644
--- a/drivers/net/ethernet/sfc/mcdi.c
+++ b/drivers/net/ethernet/sfc/mcdi.c
@@ -1018,6 +1018,9 @@ void efx_mcdi_process_event(struct efx_channel *channel,
case MCDI_EVENT_CODE_PTP_PPS:
efx_ptp_event(efx, event);
break;
+ case MCDI_EVENT_CODE_PTP_TIME:
+ efx_time_sync_event(channel, event);
+ break;
case MCDI_EVENT_CODE_TX_FLUSH:
case MCDI_EVENT_CODE_RX_FLUSH:
/* Two flush events will be sent: one to the same event
diff --git a/drivers/net/ethernet/sfc/net_driver.h b/drivers/net/ethernet/sfc/net_driver.h
index 1a3e4972c68f..0d0e8eb88332 100644
--- a/drivers/net/ethernet/sfc/net_driver.h
+++ b/drivers/net/ethernet/sfc/net_driver.h
@@ -369,6 +369,13 @@ enum efx_rx_alloc_method {
RX_ALLOC_METHOD_PAGE = 2,
};
+enum efx_sync_events_state {
+ SYNC_EVENTS_DISABLED = 0,
+ SYNC_EVENTS_QUIESCENT,
+ SYNC_EVENTS_REQUESTED,
+ SYNC_EVENTS_VALID,
+};
+
/**
* struct efx_channel - An Efx channel
*
@@ -408,6 +415,9 @@ enum efx_rx_alloc_method {
* by __efx_rx_packet(), if @rx_pkt_n_frags != 0
* @rx_queue: RX queue for this channel
* @tx_queue: TX queues for this channel
+ * @sync_events_state: Current state of sync events on this channel
+ * @sync_timestamp_major: Major part of the last ptp sync event
+ * @sync_timestamp_minor: Minor part of the last ptp sync event
*/
struct efx_channel {
struct efx_nic *efx;
@@ -446,6 +456,10 @@ struct efx_channel {
struct efx_rx_queue rx_queue;
struct efx_tx_queue tx_queue[EFX_TXQ_TYPES];
+
+ enum efx_sync_events_state sync_events_state;
+ u32 sync_timestamp_major;
+ u32 sync_timestamp_minor;
};
/**
@@ -686,6 +700,8 @@ struct vfdi_status;
* (valid only if @rx_prefix_size != 0; always negative)
* @rx_packet_len_offset: Offset of RX packet length from start of packet data
* (valid only for NICs that set %EFX_RX_PKT_PREFIX_LEN; always negative)
+ * @rx_packet_ts_offset: Offset of timestamp from start of packet data
+ * (valid only if channel->sync_timestamps_enabled; always negative)
* @rx_hash_key: Toeplitz hash key for RSS
* @rx_indir_table: Indirection table for RSS
* @rx_scatter: Scatter mode enabled for receives
@@ -820,6 +836,7 @@ struct efx_nic {
unsigned int rx_prefix_size;
int rx_packet_hash_offset;
int rx_packet_len_offset;
+ int rx_packet_ts_offset;
u8 rx_hash_key[40];
u32 rx_indir_table[128];
bool rx_scatter;
@@ -1035,6 +1052,8 @@ struct efx_mtd_partition {
* also notifies the driver that a writer has finished using this
* partition.
* @ptp_write_host_time: Send host time to MC as part of sync protocol
+ * @ptp_set_ts_sync_events: Enable or disable sync events for inline RX
+ * timestamping, possibly only temporarily for the purposes of a reset.
* @ptp_set_ts_config: Set hardware timestamp configuration. The flags
* and tx_type will already have been validated but this operation
* must validate and update rx_filter.
@@ -1047,6 +1066,7 @@ struct efx_mtd_partition {
* @max_dma_mask: Maximum possible DMA mask
* @rx_prefix_size: Size of RX prefix before packet data
* @rx_hash_offset: Offset of RX flow hash within prefix
+ * @rx_ts_offset: Offset of timestamp within prefix
* @rx_buffer_padding: Size of padding at end of RX packet
* @can_rx_scatter: NIC is able to scatter packets to multiple buffers
* @always_rx_scatter: NIC will always scatter packets to multiple buffers
@@ -1158,6 +1178,7 @@ struct efx_nic_type {
int (*mtd_sync)(struct mtd_info *mtd);
#endif
void (*ptp_write_host_time)(struct efx_nic *efx, u32 host_time);
+ int (*ptp_set_ts_sync_events)(struct efx_nic *efx, bool en, bool temp);
int (*ptp_set_ts_config)(struct efx_nic *efx,
struct hwtstamp_config *init);
@@ -1170,6 +1191,7 @@ struct efx_nic_type {
u64 max_dma_mask;
unsigned int rx_prefix_size;
unsigned int rx_hash_offset;
+ unsigned int rx_ts_offset;
unsigned int rx_buffer_padding;
bool can_rx_scatter;
bool always_rx_scatter;
diff --git a/drivers/net/ethernet/sfc/nic.h b/drivers/net/ethernet/sfc/nic.h
index 999ef285662d..5d9e2dc121f7 100644
--- a/drivers/net/ethernet/sfc/nic.h
+++ b/drivers/net/ethernet/sfc/nic.h
@@ -566,6 +566,15 @@ int efx_ptp_change_mode(struct efx_nic *efx, bool enable_wanted,
unsigned int new_mode);
int efx_ptp_tx(struct efx_nic *efx, struct sk_buff *skb);
void efx_ptp_event(struct efx_nic *efx, efx_qword_t *ev);
+void efx_time_sync_event(struct efx_channel *channel, efx_qword_t *ev);
+void __efx_rx_skb_attach_timestamp(struct efx_channel *channel,
+ struct sk_buff *skb);
+static inline void efx_rx_skb_attach_timestamp(struct efx_channel *channel,
+ struct sk_buff *skb)
+{
+ if (channel->sync_events_state == SYNC_EVENTS_VALID)
+ __efx_rx_skb_attach_timestamp(channel, skb);
+}
void efx_ptp_start_datapath(struct efx_nic *efx);
void efx_ptp_stop_datapath(struct efx_nic *efx);
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index 1d1a6f7325da..843e98dfb1b2 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -216,6 +216,8 @@ struct efx_ptp_timeset {
* struct efx_ptp_data - Precision Time Protocol (PTP) state
* @efx: The NIC context
* @channel: The PTP channel (Siena only)
+ * @rx_ts_inline: Flag for whether RX timestamps are inline (else they are
+ * separate events)
* @rxq: Receive queue (awaiting timestamps)
* @txq: Transmit queue
* @evt_list: List of MC receive events awaiting packets
@@ -258,6 +260,7 @@ struct efx_ptp_timeset {
struct efx_ptp_data {
struct efx_nic *efx;
struct efx_channel *channel;
+ bool rx_ts_inline;
struct sk_buff_head rxq;
struct sk_buff_head txq;
struct list_head evt_list;
@@ -317,8 +320,8 @@ static void efx_ptp_ns_to_s_ns(s64 ns, u32 *nic_major, u32 *nic_minor)
*nic_minor = ts.tv_nsec;
}
-static ktime_t efx_ptp_s_ns_to_ktime(u32 nic_major, u32 nic_minor,
- s32 correction)
+static ktime_t efx_ptp_s_ns_to_ktime_correction(u32 nic_major, u32 nic_minor,
+ s32 correction)
{
ktime_t kt = ktime_set(nic_major, nic_minor);
if (correction >= 0)
@@ -359,11 +362,16 @@ static void efx_ptp_ns_to_s27(s64 ns, u32 *nic_major, u32 *nic_minor)
*nic_minor = min;
}
-static ktime_t efx_ptp_s27_to_ktime(u32 nic_major, u32 nic_minor,
- s32 correction)
+static inline ktime_t efx_ptp_s27_to_ktime(u32 nic_major, u32 nic_minor)
{
- u32 ns;
+ u32 ns = (u32)(((u64)nic_minor * NSEC_PER_SEC +
+ (1ULL << (S27_TO_NS_SHIFT - 1))) >> S27_TO_NS_SHIFT);
+ return ktime_set(nic_major, ns);
+}
+static ktime_t efx_ptp_s27_to_ktime_correction(u32 nic_major, u32 nic_minor,
+ s32 correction)
+{
/* Apply the correction and deal with carry */
nic_minor += correction;
if ((s32)nic_minor < 0) {
@@ -374,10 +382,7 @@ static ktime_t efx_ptp_s27_to_ktime(u32 nic_major, u32 nic_minor,
nic_major++;
}
- ns = (u32)(((u64)nic_minor * NSEC_PER_SEC +
- (1ULL << (S27_TO_NS_SHIFT - 1))) >> S27_TO_NS_SHIFT);
-
- return ktime_set(nic_major, ns);
+ return efx_ptp_s27_to_ktime(nic_major, nic_minor);
}
/* Get PTP attributes and set up time conversions */
@@ -407,10 +412,10 @@ static int efx_ptp_get_attributes(struct efx_nic *efx)
if (fmt == MC_CMD_PTP_OUT_GET_ATTRIBUTES_SECONDS_27FRACTION) {
ptp->ns_to_nic_time = efx_ptp_ns_to_s27;
- ptp->nic_to_kernel_time = efx_ptp_s27_to_ktime;
+ ptp->nic_to_kernel_time = efx_ptp_s27_to_ktime_correction;
} else if (fmt == MC_CMD_PTP_OUT_GET_ATTRIBUTES_SECONDS_NANOSECONDS) {
ptp->ns_to_nic_time = efx_ptp_ns_to_s_ns;
- ptp->nic_to_kernel_time = efx_ptp_s_ns_to_ktime;
+ ptp->nic_to_kernel_time = efx_ptp_s_ns_to_ktime_correction;
} else {
return -ERANGE;
}
@@ -806,6 +811,9 @@ static void efx_ptp_drop_time_expired_events(struct efx_nic *efx)
struct list_head *cursor;
struct list_head *next;
+ if (ptp->rx_ts_inline)
+ return;
+
/* Drop time-expired events */
spin_lock_bh(&ptp->evt_lock);
if (!list_empty(&ptp->evt_list)) {
@@ -839,6 +847,8 @@ static enum ptp_packet_state efx_ptp_match_rx(struct efx_nic *efx,
struct efx_ptp_match *match;
enum ptp_packet_state rc = PTP_PACKET_STATE_UNMATCHED;
+ WARN_ON_ONCE(ptp->rx_ts_inline);
+
spin_lock_bh(&ptp->evt_lock);
evts_waiting = !list_empty(&ptp->evt_list);
spin_unlock_bh(&ptp->evt_lock);
@@ -1061,8 +1071,6 @@ static void efx_ptp_pps_worker(struct work_struct *work)
ptp_clock_event(ptp->phc_clock, &ptp_evt);
}
-/* Process any pending transmissions and timestamp any received packets.
- */
static void efx_ptp_worker(struct work_struct *work)
{
struct efx_ptp_data *ptp_data =
@@ -1120,6 +1128,7 @@ int efx_ptp_probe(struct efx_nic *efx, struct efx_channel *channel)
ptp->efx = efx;
ptp->channel = channel;
+ ptp->rx_ts_inline = efx_nic_rev(efx) >= EFX_REV_HUNT_A0;
rc = efx_nic_alloc_buffer(efx, &ptp->start, sizeof(int), GFP_KERNEL);
if (rc != 0)
@@ -1491,6 +1500,9 @@ static void ptp_event_rx(struct efx_nic *efx, struct efx_ptp_data *ptp)
{
struct efx_ptp_event_rx *evt = NULL;
+ if (WARN_ON_ONCE(ptp->rx_ts_inline))
+ return;
+
if (ptp->evt_frag_idx != 3) {
ptp_event_failure(efx, 3);
return;
@@ -1587,6 +1599,93 @@ void efx_ptp_event(struct efx_nic *efx, efx_qword_t *ev)
}
}
+void efx_time_sync_event(struct efx_channel *channel, efx_qword_t *ev)
+{
+ channel->sync_timestamp_major = MCDI_EVENT_FIELD(*ev, PTP_TIME_MAJOR);
+ channel->sync_timestamp_minor =
+ MCDI_EVENT_FIELD(*ev, PTP_TIME_MINOR_26_19) << 19;
+ /* if sync events have been disabled then we want to silently ignore
+ * this event, so throw away result.
+ */
+ (void) cmpxchg(&channel->sync_events_state, SYNC_EVENTS_REQUESTED,
+ SYNC_EVENTS_VALID);
+}
+
+/* make some assumptions about the time representation rather than abstract it,
+ * since we currently only support one type of inline timestamping and only on
+ * EF10.
+ */
+#define MINOR_TICKS_PER_SECOND 0x8000000
+/* Fuzz factor for sync events to be out of order with RX events */
+#define FUZZ (MINOR_TICKS_PER_SECOND / 10)
+#define EXPECTED_SYNC_EVENTS_PER_SECOND 4
+
+static inline u32 efx_rx_buf_timestamp_minor(struct efx_nic *efx, const u8 *eh)
+{
+#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS)
+ return __le32_to_cpup((const __le32 *)(eh + efx->rx_packet_ts_offset));
+#else
+ const u8 *data = eh + efx->rx_packet_ts_offset;
+ return (u32)data[0] |
+ (u32)data[1] << 8 |
+ (u32)data[2] << 16 |
+ (u32)data[3] << 24;
+#endif
+}
+
+void __efx_rx_skb_attach_timestamp(struct efx_channel *channel,
+ struct sk_buff *skb)
+{
+ struct efx_nic *efx = channel->efx;
+ u32 pkt_timestamp_major, pkt_timestamp_minor;
+ u32 diff, carry;
+ struct skb_shared_hwtstamps *timestamps;
+
+ pkt_timestamp_minor = (efx_rx_buf_timestamp_minor(efx,
+ skb_mac_header(skb)) +
+ (u32) efx->ptp_data->ts_corrections.rx) &
+ (MINOR_TICKS_PER_SECOND - 1);
+
+ /* get the difference between the packet and sync timestamps,
+ * modulo one second
+ */
+ diff = (pkt_timestamp_minor - channel->sync_timestamp_minor) &
+ (MINOR_TICKS_PER_SECOND - 1);
+ /* do we roll over a second boundary and need to carry the one? */
+ carry = channel->sync_timestamp_minor + diff > MINOR_TICKS_PER_SECOND ?
+ 1 : 0;
+
+ if (diff <= MINOR_TICKS_PER_SECOND / EXPECTED_SYNC_EVENTS_PER_SECOND +
+ FUZZ) {
+ /* packet is ahead of the sync event by a quarter of a second or
+ * less (allowing for fuzz)
+ */
+ pkt_timestamp_major = channel->sync_timestamp_major + carry;
+ } else if (diff >= MINOR_TICKS_PER_SECOND - FUZZ) {
+ /* packet is behind the sync event but within the fuzz factor.
+ * This means the RX packet and sync event crossed as they were
+ * placed on the event queue, which can sometimes happen.
+ */
+ pkt_timestamp_major = channel->sync_timestamp_major - 1 + carry;
+ } else {
+ /* it's outside tolerance in both directions. this might be
+ * indicative of us missing sync events for some reason, so
+ * we'll call it an error rather than risk giving a bogus
+ * timestamp.
+ */
+ netif_vdbg(efx, drv, efx->net_dev,
+ "packet timestamp %x too far from sync event %x:%x\n",
+ pkt_timestamp_minor, channel->sync_timestamp_major,
+ channel->sync_timestamp_minor);
+ return;
+ }
+
+ /* attach the timestamps to the skb */
+ timestamps = skb_hwtstamps(skb);
+ timestamps->hwtstamp =
+ efx_ptp_s27_to_ktime(pkt_timestamp_major, pkt_timestamp_minor);
+}
+
static int efx_phc_adjfreq(struct ptp_clock_info *ptp, s32 delta)
{
struct efx_ptp_data *ptp_data = container_of(ptp,
@@ -1728,9 +1827,15 @@ void efx_ptp_start_datapath(struct efx_nic *efx)
{
if (efx_ptp_restart(efx))
netif_err(efx, drv, efx->net_dev, "Failed to restart PTP.\n");
+ /* re-enable timestamping if it was previously enabled */
+ if (efx->type->ptp_set_ts_sync_events)
+ efx->type->ptp_set_ts_sync_events(efx, true, true);
}
void efx_ptp_stop_datapath(struct efx_nic *efx)
{
+ /* temporarily disable timestamping */
+ if (efx->type->ptp_set_ts_sync_events)
+ efx->type->ptp_set_ts_sync_events(efx, false, true);
efx_ptp_stop(efx);
}
diff --git a/drivers/net/ethernet/sfc/rx.c b/drivers/net/ethernet/sfc/rx.c
index d12abc543975..1fde9b8ac456 100644
--- a/drivers/net/ethernet/sfc/rx.c
+++ b/drivers/net/ethernet/sfc/rx.c
@@ -624,6 +624,8 @@ static void efx_rx_deliver(struct efx_channel *channel, u8 *eh,
if (likely(rx_buf->flags & EFX_RX_PKT_CSUMMED))
skb->ip_summed = CHECKSUM_UNNECESSARY;
+ efx_rx_skb_attach_timestamp(channel, skb);
+
if (channel->type->receive_skb)
if (channel->type->receive_skb(channel, skb))
return;
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 11/15] sfc: Store VPD serial number at probe time
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (9 preceding siblings ...)
2013-12-14 23:05 ` [PATCH net-next 10/15] sfc: Add RX packet timestamping for EF10 Ben Hutchings
@ 2013-12-14 23:05 ` Ben Hutchings
2013-12-14 23:06 ` [PATCH net-next 12/15] sfc: Associate primary and secondary functions of controller Ben Hutchings
` (4 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:05 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
Original version by Stuart Hodgson.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/efx.c | 38 ++++++++++++++++++++++++++------
drivers/net/ethernet/sfc/net_driver.h | 3 ++
2 files changed, 34 insertions(+), 7 deletions(-)
diff --git a/drivers/net/ethernet/sfc/efx.c b/drivers/net/ethernet/sfc/efx.c
index b9929d9ed69d..5745a9fd034e 100644
--- a/drivers/net/ethernet/sfc/efx.c
+++ b/drivers/net/ethernet/sfc/efx.c
@@ -2628,6 +2628,8 @@ static void efx_fini_struct(struct efx_nic *efx)
for (i = 0; i < EFX_MAX_CHANNELS; i++)
kfree(efx->channel[i]);
+ kfree(efx->vpd_sn);
+
if (efx->workqueue) {
destroy_workqueue(efx->workqueue);
efx->workqueue = NULL;
@@ -2699,12 +2701,12 @@ static void efx_pci_remove(struct pci_dev *pci_dev)
* always appear within the first 512 bytes.
*/
#define SFC_VPD_LEN 512
-static void efx_print_product_vpd(struct efx_nic *efx)
+static void efx_probe_vpd_strings(struct efx_nic *efx)
{
struct pci_dev *dev = efx->pci_dev;
char vpd_data[SFC_VPD_LEN];
ssize_t vpd_size;
- int i, j;
+ int ro_start, ro_size, i, j;
/* Get the vpd data from the device */
vpd_size = pci_read_vpd(dev, 0, sizeof(vpd_data), vpd_data);
@@ -2714,14 +2716,15 @@ static void efx_print_product_vpd(struct efx_nic *efx)
}
/* Get the Read only section */
- i = pci_vpd_find_tag(vpd_data, 0, vpd_size, PCI_VPD_LRDT_RO_DATA);
- if (i < 0) {
+ ro_start = pci_vpd_find_tag(vpd_data, 0, vpd_size, PCI_VPD_LRDT_RO_DATA);
+ if (ro_start < 0) {
netif_err(efx, drv, efx->net_dev, "VPD Read-only not found\n");
return;
}
- j = pci_vpd_lrdt_size(&vpd_data[i]);
- i += PCI_VPD_LRDT_TAG_SIZE;
+ ro_size = pci_vpd_lrdt_size(&vpd_data[ro_start]);
+ j = ro_size;
+ i = ro_start + PCI_VPD_LRDT_TAG_SIZE;
if (i + j > vpd_size)
j = vpd_size - i;
@@ -2741,6 +2744,27 @@ static void efx_print_product_vpd(struct efx_nic *efx)
netif_info(efx, drv, efx->net_dev,
"Part Number : %.*s\n", j, &vpd_data[i]);
+
+ i = ro_start + PCI_VPD_LRDT_TAG_SIZE;
+ j = ro_size;
+ i = pci_vpd_find_info_keyword(vpd_data, i, j, "SN");
+ if (i < 0) {
+ netif_err(efx, drv, efx->net_dev, "Serial number not found\n");
+ return;
+ }
+
+ j = pci_vpd_info_field_size(&vpd_data[i]);
+ i += PCI_VPD_INFO_FLD_HDR_SIZE;
+ if (i + j > vpd_size) {
+ netif_err(efx, drv, efx->net_dev, "Incomplete serial number\n");
+ return;
+ }
+
+ efx->vpd_sn = kmalloc(j + 1, GFP_KERNEL);
+ if (!efx->vpd_sn)
+ return;
+
+ snprintf(efx->vpd_sn, j + 1, "%s", &vpd_data[i]);
}
@@ -2837,7 +2861,7 @@ static int efx_pci_probe(struct pci_dev *pci_dev,
netif_info(efx, probe, efx->net_dev,
"Solarflare NIC detected\n");
- efx_print_product_vpd(efx);
+ efx_probe_vpd_strings(efx);
/* Set up basic I/O (BAR mappings etc) */
rc = efx_init_io(efx);
diff --git a/drivers/net/ethernet/sfc/net_driver.h b/drivers/net/ethernet/sfc/net_driver.h
index 0d0e8eb88332..be8a616a1649 100644
--- a/drivers/net/ethernet/sfc/net_driver.h
+++ b/drivers/net/ethernet/sfc/net_driver.h
@@ -771,6 +771,7 @@ struct vfdi_status;
* @local_lock: Mutex protecting %local_addr_list and %local_page_list.
* @peer_work: Work item to broadcast peer addresses to VMs.
* @ptp_data: PTP state data
+ * @vpd_sn: Serial number read from VPD
* @monitor_work: Hardware monitor workitem
* @biu_lock: BIU (bus interface unit) lock
* @last_irq_cpu: Last CPU to handle a possible test interrupt. This
@@ -920,6 +921,8 @@ struct efx_nic {
struct efx_ptp_data *ptp_data;
+ char *vpd_sn;
+
/* The following fields may be written more often */
struct delayed_work monitor_work ____cacheline_aligned_in_smp;
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 12/15] sfc: Associate primary and secondary functions of controller
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (10 preceding siblings ...)
2013-12-14 23:05 ` [PATCH net-next 11/15] sfc: Store VPD serial number at probe time Ben Hutchings
@ 2013-12-14 23:06 ` Ben Hutchings
2013-12-14 23:06 ` [PATCH net-next 13/15] sfc: Enable PTP clock and timestamping for all functions on EF10 Ben Hutchings
` (3 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:06 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
The primary function of an EF10 controller will share its clock
device with other functions in the same domain (which we call
secondary functions). To this end, we need to associate functions
on the same controller.
We do not control probe order, so allow primary and secondary
functions to appear in any order. Maintain global lists of all
primary functions and of unassociated secondary functions,
and a list of secondary functions on each primary function.
Use the VPD serial number to tell whether functions are part of the
same controller. VPD will not be readable by virtual functions, so
this may need to be revisited later.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/efx.c | 77 +++++++++++++++++++++++++++++++++
drivers/net/ethernet/sfc/falcon.c | 2 +
drivers/net/ethernet/sfc/mcdi.c | 4 ++
drivers/net/ethernet/sfc/net_driver.h | 10 ++++
4 files changed, 93 insertions(+), 0 deletions(-)
diff --git a/drivers/net/ethernet/sfc/efx.c b/drivers/net/ethernet/sfc/efx.c
index 5745a9fd034e..d6a927021523 100644
--- a/drivers/net/ethernet/sfc/efx.c
+++ b/drivers/net/ethernet/sfc/efx.c
@@ -1117,6 +1117,77 @@ static void efx_remove_port(struct efx_nic *efx)
*
**************************************************************************/
+static LIST_HEAD(efx_primary_list);
+static LIST_HEAD(efx_unassociated_list);
+
+static bool efx_same_controller(struct efx_nic *left, struct efx_nic *right)
+{
+ return left->type == right->type &&
+ left->vpd_sn && right->vpd_sn &&
+ !strcmp(left->vpd_sn, right->vpd_sn);
+}
+
+static void efx_associate(struct efx_nic *efx)
+{
+ struct efx_nic *other, *next;
+
+ if (efx->primary == efx) {
+ /* Adding primary function; look for secondaries */
+
+ netif_dbg(efx, probe, efx->net_dev, "adding to primary list\n");
+ list_add_tail(&efx->node, &efx_primary_list);
+
+ list_for_each_entry_safe(other, next, &efx_unassociated_list,
+ node) {
+ if (efx_same_controller(efx, other)) {
+ list_del(&other->node);
+ netif_dbg(other, probe, other->net_dev,
+ "moving to secondary list of %s %s\n",
+ pci_name(efx->pci_dev),
+ efx->net_dev->name);
+ list_add_tail(&other->node,
+ &efx->secondary_list);
+ other->primary = efx;
+ }
+ }
+ } else {
+ /* Adding secondary function; look for primary */
+
+ list_for_each_entry(other, &efx_primary_list, node) {
+ if (efx_same_controller(efx, other)) {
+ netif_dbg(efx, probe, efx->net_dev,
+ "adding to secondary list of %s %s\n",
+ pci_name(other->pci_dev),
+ other->net_dev->name);
+ list_add_tail(&efx->node,
+ &other->secondary_list);
+ efx->primary = other;
+ return;
+ }
+ }
+
+ netif_dbg(efx, probe, efx->net_dev,
+ "adding to unassociated list\n");
+ list_add_tail(&efx->node, &efx_unassociated_list);
+ }
+}
+
+static void efx_dissociate(struct efx_nic *efx)
+{
+ struct efx_nic *other, *next;
+
+ list_del(&efx->node);
+ efx->primary = NULL;
+
+ list_for_each_entry_safe(other, next, &efx->secondary_list, node) {
+ list_del(&other->node);
+ netif_dbg(other, probe, other->net_dev,
+ "moving to unassociated list\n");
+ list_add_tail(&other->node, &efx_unassociated_list);
+ other->primary = NULL;
+ }
+}
+
/* This configures the PCI device to enable I/O and DMA. */
static int efx_init_io(struct efx_nic *efx)
{
@@ -2214,6 +2285,8 @@ static int efx_register_netdev(struct efx_nic *efx)
efx_init_tx_queue_core_txq(tx_queue);
}
+ efx_associate(efx);
+
rtnl_unlock();
rc = device_create_file(&efx->pci_dev->dev, &dev_attr_phy_type);
@@ -2227,6 +2300,7 @@ static int efx_register_netdev(struct efx_nic *efx)
fail_registered:
rtnl_lock();
+ efx_dissociate(efx);
unregister_netdevice(net_dev);
fail_locked:
efx->state = STATE_UNINIT;
@@ -2568,6 +2642,8 @@ static int efx_init_struct(struct efx_nic *efx,
int i;
/* Initialise common structures */
+ INIT_LIST_HEAD(&efx->node);
+ INIT_LIST_HEAD(&efx->secondary_list);
spin_lock_init(&efx->biu_lock);
#ifdef CONFIG_SFC_MTD
INIT_LIST_HEAD(&efx->mtd_list);
@@ -2674,6 +2750,7 @@ static void efx_pci_remove(struct pci_dev *pci_dev)
/* Mark the NIC as fini, then stop the interface */
rtnl_lock();
+ efx_dissociate(efx);
dev_close(efx->net_dev);
efx_disable_interrupts(efx);
rtnl_unlock();
diff --git a/drivers/net/ethernet/sfc/falcon.c b/drivers/net/ethernet/sfc/falcon.c
index 4a9e05c82e2a..76699f4e6e04 100644
--- a/drivers/net/ethernet/sfc/falcon.c
+++ b/drivers/net/ethernet/sfc/falcon.c
@@ -2247,6 +2247,8 @@ static int falcon_probe_nic(struct efx_nic *efx)
struct falcon_board *board;
int rc;
+ efx->primary = efx; /* only one usable function per controller */
+
/* Allocate storage for hardware specific data */
nic_data = kzalloc(sizeof(*nic_data), GFP_KERNEL);
if (!nic_data)
diff --git a/drivers/net/ethernet/sfc/mcdi.c b/drivers/net/ethernet/sfc/mcdi.c
index 540f57915d6f..0d5d7b5325e8 100644
--- a/drivers/net/ethernet/sfc/mcdi.c
+++ b/drivers/net/ethernet/sfc/mcdi.c
@@ -102,6 +102,10 @@ int efx_mcdi_init(struct efx_nic *efx)
netif_err(efx, probe, efx->net_dev,
"Host already registered with MCPU\n");
+ if (efx->mcdi->fn_flags &
+ (1 << MC_CMD_DRV_ATTACH_EXT_OUT_FLAG_PRIMARY))
+ efx->primary = efx;
+
return 0;
}
diff --git a/drivers/net/ethernet/sfc/net_driver.h b/drivers/net/ethernet/sfc/net_driver.h
index be8a616a1649..653b8782c956 100644
--- a/drivers/net/ethernet/sfc/net_driver.h
+++ b/drivers/net/ethernet/sfc/net_driver.h
@@ -657,6 +657,13 @@ struct vfdi_status;
* struct efx_nic - an Efx NIC
* @name: Device name (net device name or bus id before net device registered)
* @pci_dev: The PCI device
+ * @node: List node for maintaning primary/secondary function lists
+ * @primary: &struct efx_nic instance for the primary function of this
+ * controller. May be the same structure, and may be %NULL if no
+ * primary function is bound. Serialised by rtnl_lock.
+ * @secondary_list: List of &struct efx_nic instances for the secondary PCI
+ * functions of the controller, if this is for the primary function.
+ * Serialised by rtnl_lock.
* @type: Controller type attributes
* @legacy_irq: IRQ number
* @workqueue: Workqueue for port reconfigures and the HW monitor.
@@ -786,6 +793,9 @@ struct efx_nic {
/* The following fields should be written very rarely */
char name[IFNAMSIZ];
+ struct list_head node;
+ struct efx_nic *primary;
+ struct list_head secondary_list;
struct pci_dev *pci_dev;
unsigned int port_num;
const struct efx_nic_type *type;
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 13/15] sfc: Enable PTP clock and timestamping for all functions on EF10
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (11 preceding siblings ...)
2013-12-14 23:06 ` [PATCH net-next 12/15] sfc: Associate primary and secondary functions of controller Ben Hutchings
@ 2013-12-14 23:06 ` Ben Hutchings
2013-12-14 23:07 ` [PATCH net-next 14/15] sfc: Don't clear timestamps in efx_ptp_rx() Ben Hutchings
` (2 subsequent siblings)
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:06 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
The SFC9100 family has only one clock per controller, shared by all
functions. Therefore only create a clock device under the primary
function, and make all other functions refer to the primary's clock
device.
Since PTP functionality is limited to port 0 and PF 0 on the earlier
SFN[56]322F boards, and we also set the primary flag for that
function, we can make the creation of a clock device conditional only
on this flag.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ef10.c | 5 +++-
drivers/net/ethernet/sfc/ptp.c | 42 ++++++++++++++++++++++++--------------
2 files changed, 30 insertions(+), 17 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ef10.c b/drivers/net/ethernet/sfc/ef10.c
index a3d891764030..dfff8e7a55e2 100644
--- a/drivers/net/ethernet/sfc/ef10.c
+++ b/drivers/net/ethernet/sfc/ef10.c
@@ -264,6 +264,8 @@ static int efx_ef10_probe(struct efx_nic *efx)
if (rc)
goto fail3;
+ efx_ptp_probe(efx, NULL);
+
return 0;
fail3:
@@ -472,9 +474,10 @@ static void efx_ef10_remove(struct efx_nic *efx)
struct efx_ef10_nic_data *nic_data = efx->nic_data;
int rc;
+ efx_ptp_remove(efx);
+
efx_mcdi_mon_remove(efx);
- /* This needs to be after efx_ptp_remove_channel() with no filters */
efx_ef10_rx_free_indir_table(efx);
if (nic_data->wc_membase)
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index 843e98dfb1b2..fe3c6d04fce8 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -248,7 +248,7 @@ struct efx_ptp_timeset {
* @start: Address at which MC indicates ready for synchronisation
* @host_time_pps: Host time at last PPS
* @current_adjfreq: Current ppb adjustment.
- * @phc_clock: Pointer to registered phc device
+ * @phc_clock: Pointer to registered phc device (if primary function)
* @phc_clock_info: Registration structure for phc device
* @pps_work: pps work task for handling pps events
* @pps_workwq: pps work queue
@@ -1163,19 +1163,22 @@ int efx_ptp_probe(struct efx_nic *efx, struct efx_channel *channel)
if (rc < 0)
goto fail3;
- ptp->phc_clock_info = efx_phc_clock_info;
- ptp->phc_clock = ptp_clock_register(&ptp->phc_clock_info,
- &efx->pci_dev->dev);
- if (IS_ERR(ptp->phc_clock)) {
- rc = PTR_ERR(ptp->phc_clock);
- goto fail3;
- }
+ if (efx->mcdi->fn_flags &
+ (1 << MC_CMD_DRV_ATTACH_EXT_OUT_FLAG_PRIMARY)) {
+ ptp->phc_clock_info = efx_phc_clock_info;
+ ptp->phc_clock = ptp_clock_register(&ptp->phc_clock_info,
+ &efx->pci_dev->dev);
+ if (IS_ERR(ptp->phc_clock)) {
+ rc = PTR_ERR(ptp->phc_clock);
+ goto fail3;
+ }
- INIT_WORK(&ptp->pps_work, efx_ptp_pps_worker);
- ptp->pps_workwq = create_singlethread_workqueue("sfc_pps");
- if (!ptp->pps_workwq) {
- rc = -ENOMEM;
- goto fail4;
+ INIT_WORK(&ptp->pps_work, efx_ptp_pps_worker);
+ ptp->pps_workwq = create_singlethread_workqueue("sfc_pps");
+ if (!ptp->pps_workwq) {
+ rc = -ENOMEM;
+ goto fail4;
+ }
}
ptp->nic_ts_enabled = false;
@@ -1224,10 +1227,12 @@ void efx_ptp_remove(struct efx_nic *efx)
skb_queue_purge(&efx->ptp_data->rxq);
skb_queue_purge(&efx->ptp_data->txq);
- ptp_clock_unregister(efx->ptp_data->phc_clock);
+ if (efx->ptp_data->phc_clock) {
+ destroy_workqueue(efx->ptp_data->pps_workwq);
+ ptp_clock_unregister(efx->ptp_data->phc_clock);
+ }
destroy_workqueue(efx->ptp_data->workwq);
- destroy_workqueue(efx->ptp_data->pps_workwq);
efx_nic_free_buffer(efx, &efx->ptp_data->start);
kfree(efx->ptp_data);
@@ -1440,6 +1445,9 @@ static int efx_ptp_ts_init(struct efx_nic *efx, struct hwtstamp_config *init)
void efx_ptp_get_ts_info(struct efx_nic *efx, struct ethtool_ts_info *ts_info)
{
struct efx_ptp_data *ptp = efx->ptp_data;
+ struct efx_nic *primary = efx->primary;
+
+ ASSERT_RTNL();
if (!ptp)
return;
@@ -1447,7 +1455,9 @@ void efx_ptp_get_ts_info(struct efx_nic *efx, struct ethtool_ts_info *ts_info)
ts_info->so_timestamping |= (SOF_TIMESTAMPING_TX_HARDWARE |
SOF_TIMESTAMPING_RX_HARDWARE |
SOF_TIMESTAMPING_RAW_HARDWARE);
- ts_info->phc_index = ptp_clock_index(ptp->phc_clock);
+ if (primary && primary->ptp_data && primary->ptp_data->phc_clock)
+ ts_info->phc_index =
+ ptp_clock_index(primary->ptp_data->phc_clock);
ts_info->tx_types = 1 << HWTSTAMP_TX_OFF | 1 << HWTSTAMP_TX_ON;
ts_info->rx_filters = ptp->efx->type->hwtstamp_filters;
}
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 14/15] sfc: Don't clear timestamps in efx_ptp_rx()
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (12 preceding siblings ...)
2013-12-14 23:06 ` [PATCH net-next 13/15] sfc: Enable PTP clock and timestamping for all functions on EF10 Ben Hutchings
@ 2013-12-14 23:07 ` Ben Hutchings
2013-12-14 23:07 ` [PATCH net-next 15/15] sfc: Remove unnecessary condition for processing the TX timestamp queue Ben Hutchings
2013-12-15 3:34 ` Pull request: sfc-next 2013-12-14 David Miller
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:07 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
A freshly allocated skb starts with timestamps clear.
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ptp.c | 6 ------
1 files changed, 0 insertions(+), 6 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index fe3c6d04fce8..06c6420f40d7 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -1323,14 +1323,8 @@ static bool efx_ptp_rx(struct efx_channel *channel, struct sk_buff *skb)
/* Does this packet require timestamping? */
if (ntohs(*(__be16 *)&skb->data[PTP_DPORT_OFFSET]) == PTP_EVENT_PORT) {
- struct skb_shared_hwtstamps *timestamps;
-
match->state = PTP_PACKET_STATE_UNMATCHED;
- /* Clear all timestamps held: filled in later */
- timestamps = skb_hwtstamps(skb);
- memset(timestamps, 0, sizeof(*timestamps));
-
/* We expect the sequence number to be in the same position in
* the packet for PTP V1 and V2
*/
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* [PATCH net-next 15/15] sfc: Remove unnecessary condition for processing the TX timestamp queue
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (13 preceding siblings ...)
2013-12-14 23:07 ` [PATCH net-next 14/15] sfc: Don't clear timestamps in efx_ptp_rx() Ben Hutchings
@ 2013-12-14 23:07 ` Ben Hutchings
2013-12-15 3:34 ` Pull request: sfc-next 2013-12-14 David Miller
15 siblings, 0 replies; 18+ messages in thread
From: Ben Hutchings @ 2013-12-14 23:07 UTC (permalink / raw)
To: David Miller; +Cc: netdev, linux-net-drivers, Richard Cochran
Signed-off-by: Ben Hutchings <bhutchings@solarflare.com>
---
drivers/net/ethernet/sfc/ptp.c | 16 ++++------------
1 files changed, 4 insertions(+), 12 deletions(-)
diff --git a/drivers/net/ethernet/sfc/ptp.c b/drivers/net/ethernet/sfc/ptp.c
index 06c6420f40d7..084e2d44790f 100644
--- a/drivers/net/ethernet/sfc/ptp.c
+++ b/drivers/net/ethernet/sfc/ptp.c
@@ -890,13 +890,10 @@ static enum ptp_packet_state efx_ptp_match_rx(struct efx_nic *efx,
/* Process any queued receive events and corresponding packets
*
* q is returned with all the packets that are ready for delivery.
- * true is returned if at least one of those packets requires
- * synchronisation.
*/
-static bool efx_ptp_process_events(struct efx_nic *efx, struct sk_buff_head *q)
+static void efx_ptp_process_events(struct efx_nic *efx, struct sk_buff_head *q)
{
struct efx_ptp_data *ptp = efx->ptp_data;
- bool rc = false;
struct sk_buff *skb;
while ((skb = skb_dequeue(&ptp->rxq))) {
@@ -907,7 +904,6 @@ static bool efx_ptp_process_events(struct efx_nic *efx, struct sk_buff_head *q)
__skb_queue_tail(q, skb);
} else if (efx_ptp_match_rx(efx, skb) ==
PTP_PACKET_STATE_MATCHED) {
- rc = true;
__skb_queue_tail(q, skb);
} else if (time_after(jiffies, match->expiry)) {
match->state = PTP_PACKET_STATE_TIMED_OUT;
@@ -921,8 +917,6 @@ static bool efx_ptp_process_events(struct efx_nic *efx, struct sk_buff_head *q)
break;
}
}
-
- return rc;
}
/* Complete processing of a received packet */
@@ -1088,12 +1082,10 @@ static void efx_ptp_worker(struct work_struct *work)
efx_ptp_drop_time_expired_events(efx);
__skb_queue_head_init(&tempq);
- if (efx_ptp_process_events(efx, &tempq) ||
- !skb_queue_empty(&ptp_data->txq)) {
+ efx_ptp_process_events(efx, &tempq);
- while ((skb = skb_dequeue(&ptp_data->txq)))
- efx_ptp_xmit_skb(efx, skb);
- }
+ while ((skb = skb_dequeue(&ptp_data->txq)))
+ efx_ptp_xmit_skb(efx, skb);
while ((skb = __skb_dequeue(&tempq)))
efx_ptp_process_rx(efx, skb);
--
Ben Hutchings, Staff Engineer, Solarflare
Not speaking for my employer; that's the marketing department's job.
They asked us to note that Solarflare product names are trademarked.
^ permalink raw reply related [flat|nested] 18+ messages in thread* Re: Pull request: sfc-next 2013-12-14
2013-12-14 22:57 Pull request: sfc-next 2013-12-14 Ben Hutchings
` (14 preceding siblings ...)
2013-12-14 23:07 ` [PATCH net-next 15/15] sfc: Remove unnecessary condition for processing the TX timestamp queue Ben Hutchings
@ 2013-12-15 3:34 ` David Miller
15 siblings, 0 replies; 18+ messages in thread
From: David Miller @ 2013-12-15 3:34 UTC (permalink / raw)
To: bhutchings; +Cc: netdev, linux-net-drivers, richardcochran
From: Ben Hutchings <bhutchings@solarflare.com>
Date: Sat, 14 Dec 2013 22:57:32 +0000
> git://git.kernel.org/pub/scm/linux/kernel/git/bwh/sfc-next.git for-davem
...
> 1. Change PTP clock name to 'sfc'.
> 2. Complete support for hardware timestamping and PTP clock on the
> SFC9100 family.
> 3. Various cleanups for the PTP code.
Pulled, thanks Ben.
^ permalink raw reply [flat|nested] 18+ messages in thread