diff options
Diffstat (limited to 'drivers/staging')
| -rw-r--r-- | drivers/staging/fsl_qbman/qman_high.c | 12 |
1 files changed, 8 insertions, 4 deletions
diff --git a/drivers/staging/fsl_qbman/qman_high.c b/drivers/staging/fsl_qbman/qman_high.c index 370e260..9b8a924 100644 --- a/drivers/staging/fsl_qbman/qman_high.c +++ b/drivers/staging/fsl_qbman/qman_high.c @@ -3223,7 +3223,7 @@ static int qman_ceetm_write_statistics(u16 cid, enum qm_dc_portal dcp_idx, return 0; } -int qman_ceetm_bps2tokenrate(u32 bps, struct qm_ceetm_rate *token_rate, +int qman_ceetm_bps2tokenrate(u64 bps, struct qm_ceetm_rate *token_rate, int rounding) { u16 pres; @@ -3254,16 +3254,20 @@ int qman_ceetm_bps2tokenrate(u32 bps, struct qm_ceetm_rate *token_rate, * token-rate = (bps*2^19) / (PRES*QHZ) * N = (bps*2^32) / (PRES*QHz) * + * And to avoid 64-bit overflow if 'bps' is larger than 4Gbps + * (yet minimise rounding error if 'bps' is small), we reorganise + * the formula to use two 16-bit shifts rather than 1 32-bit shift. + * N = (((bps*2^16)/PRES)*2^16)/QHz */ - temp = ROUNDING(((u64)bps << 32), pres, rounding); - temp = ROUNDING(temp, qman_freq, rounding); + temp = ROUNDING((bps << 16), pres, rounding); + temp = ROUNDING((temp << 16), qman_freq, rounding); token_rate->whole = temp >> 13; token_rate->fraction = temp & (((u64)1 << 13) - 1); return 0; } EXPORT_SYMBOL(qman_ceetm_bps2tokenrate); -int qman_ceetm_tokenrate2bps(const struct qm_ceetm_rate *token_rate, u32 *bps, +int qman_ceetm_tokenrate2bps(const struct qm_ceetm_rate *token_rate, u64 *bps, int rounding) { u16 pres; |