* [PATCH 6.1.y] octeontx2-pf: tc: fix egress ratelimiting
@ 2026-11-19 8:33 Rand Deeb
0 siblings, 0 replies; only message in thread
From: Rand Deeb @ 2026-11-19 8:33 UTC (permalink / raw)
To: stable
Cc: Hariprasad Kelam, Nitin Shetty J, deeb.rand, lvc-project, shda,
Jakub Kicinski, Rand Deeb
From: Hariprasad Kelam <hkelam@marvell.com>
[ Upstream commit bb0d96ebe5f4d1acccf4dc36ca7f01f9a8fa1ba1 ]
The egress rate calculation computes an incorrect mantissa and exponent,
causing up to ~50% deviation from the configured rate at lower speeds.
Rework the computation to follow the hardware rate formula:
rate = 2 * (1 + mantissa/256) * 2^exp / (1 << div_exp)
Keep div_exp = 0 and derive exp and mantissa from half of the requested
rate. Rates below 2 Mbps are floored to the smallest encodable step
(exp = 0, mantissa = 0).
Fixes: e638a83f167e ("octeontx2-pf: TC_MATCHALL egress ratelimiting offload")
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Nitin Shetty J <nshettyj@marvell.com>
Link: https://patch.msgid.link/20260717084349.2227796-1-nshettyj@marvell.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
[ Context adjusted: RATE_MANTISSA_BITS is placed with the other rate
defines, as MCAST_INVALID_GRP does not exist in this tree. ]
Signed-off-by: Rand Deeb <rand.sec96@gmail.com>
---
.../ethernet/marvell/octeontx2/nic/otx2_tc.c | 29 ++++++++++---------
1 file changed, 16 insertions(+), 13 deletions(-)
diff --git a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_tc.c b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_tc.c
index b23ac8f38..db67fb6b8 100644
--- a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_tc.c
+++ b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_tc.c
@@ -27,6 +27,7 @@
#define MAX_RATE_DIVIDER_EXPONENT 12ULL
#define MAX_RATE_EXPONENT 0x0FULL
#define MAX_RATE_MANTISSA 0xFFULL
+#define RATE_MANTISSA_BITS 8
#define CN10K_MAX_BURST_MANTISSA 0x7FFFULL
#define CN10K_MAX_BURST_SIZE 8453888ULL
@@ -101,28 +102,30 @@ static void otx2_get_egress_burst_cfg(struct otx2_nic *nic, u32 burst,
static void otx2_get_egress_rate_cfg(u64 maxrate, u32 *exp,
u32 *mantissa, u32 *div_exp)
{
- u64 tmp;
-
/* Rate calculation by hardware
*
* PIR_ADD = ((256 + mantissa) << exp) / 256
* rate = (2 * PIR_ADD) / ( 1 << div_exp)
* The resultant rate is in Mbps.
+ *
+ * Use div_exp = 0 and compute exp/mantissa for maxrate / 2; the
+ * leading factor of two yields the full rate. Rates below 2 Mbps
+ * are floored to the smallest step (exp = 0, mantissa = 0).
*/
- /* 2Mbps to 100Gbps can be expressed with div_exp = 0.
- * Setting this to '0' will ease the calculation of
- * exponent and mantissa.
- */
*div_exp = 0;
-
if (maxrate) {
- *exp = ilog2(maxrate) ? ilog2(maxrate) - 1 : 0;
- tmp = maxrate - rounddown_pow_of_two(maxrate);
- if (maxrate < MAX_RATE_MANTISSA)
- *mantissa = tmp * 2;
- else
- *mantissa = tmp / (1ULL << (*exp - 7));
+ maxrate = maxrate / 2;
+ if (!maxrate) {
+ /* Rates below 2 Mbps map to the smallest step */
+ *exp = 0;
+ *mantissa = 0;
+ } else {
+ *exp = ilog2(maxrate);
+ /* Clear MSB and derive fractional bits */
+ maxrate &= ~BIT(*exp);
+ *mantissa = (maxrate << RATE_MANTISSA_BITS) >> *exp;
+ }
} else {
/* Instead of disabling rate limiting, set all values to max */
*exp = MAX_RATE_EXPONENT;
base-commit: cf82dcca96346600c7068cf3f841335f9fa08f54
--
2.39.5
^ permalink raw reply related [flat|nested] only message in thread
only message in thread, other threads:[~2026-11-19 8:33 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-19 8:33 [PATCH 6.1.y] octeontx2-pf: tc: fix egress ratelimiting Rand Deeb
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox