diff options
Diffstat (limited to 'kernel/gsp_phys_gpu.c')
| -rw-r--r-- | kernel/gsp_phys_gpu.c | 35 |
1 files changed, 35 insertions, 0 deletions
diff --git a/kernel/gsp_phys_gpu.c b/kernel/gsp_phys_gpu.c index 9568d22..e845493 100644 --- a/kernel/gsp_phys_gpu.c +++ b/kernel/gsp_phys_gpu.c @@ -1069,6 +1069,41 @@ module_param_named(gsp_phys_gr_nowrite_retries, MODULE_PARM_DESC(gsp_phys_gr_nowrite_retries, "Dedicated re-execution budget for a data-bearing GR launch that wrote nothing tracked, clamped [0,4] (default 3)"); +/* + * STALLFREE: absolute per-launch wall-clock budget (gsp_phys_gr_wait_max_ms; + * default 60000, clamped [10000, 600000]; 0 disables the extra cap and the + * built-in ladder bounds stand alone). The GR submit ladder and its retirement + * (no-write) retries are individually bounded, but their SUM grows with every + * re-arm/re-publish: a launch whose engine ack never arrives could hold the + * single-inflight work item for tens of seconds while the guest client waits + * on its QMD completion word. When this budget is crossed the submit stops its + * ladder immediately (no further re-arm/publish/retry), logs + * 'vgsp-gr: GR-STALL-wait-max ... action=fail' and returns -ETIMEDOUT; the + * sema service's failure edge then authors the QMD completion release in + * software, so the client always completes. A green launch (~0.2-11 s end to + * end, including the fast-re-arm recovery) never approaches the default, so + * the default does not change green behavior. + */ +static unsigned int gsp_phys_gr_wait_max_ms_param = 60000u; +module_param_named(gsp_phys_gr_wait_max_ms, + gsp_phys_gr_wait_max_ms_param, uint, 0644); +MODULE_PARM_DESC(gsp_phys_gr_wait_max_ms, + "Absolute per-launch GR submit budget in ms, clamped [10000,600000]; 0 = only the built-in per-ladder bounds (default 60000)"); + +unsigned int gsp_phys_gr_wait_max_ms_value(void) +{ + unsigned int v = gsp_phys_gr_wait_max_ms_param; + + if (!v) + return 0; + if (v < 10000u) + v = 10000u; + if (v > 600000u) + v = 600000u; + return v; +} +EXPORT_SYMBOL_GPL(gsp_phys_gr_wait_max_ms_value); + int gsp_phys_gr_nowrite_retries_value(void) { int v = gsp_phys_gr_nowrite_retries_param; |