modeset: fix flags overwrite in nvRmMuxPre/nvRmMuxPost

[rootkiller6788]

Summary

Fixes the latent flags bug described in issue #1311.

In nvRmMuxPre() and nvRmMuxPost() (src/nvidia-modeset/src/nvkms-rm.c), params.flags is initialized with DRF_DEF(0073_CTRL_DFP, _DISP_MUX_FLAGS, _SR_ENTER_SKIP, _NO) and then overwritten (not OR-ed) with the switch-type value when a mux state is selected:

params.flags = DRF_DEF(0073_CTRL_DFP, _DISP_MUX_FLAGS, _SR_ENTER_SKIP, _NO);
if (state == MUX_STATE_DISCRETE) {
    params.flags = NV0073_CTRL_DFP_DISP_MUX_FLAGS_SWITCH_TYPE_IGPU_TO_DGPU;
} else if (state == MUX_STATE_INTEGRATED) {
    params.flags = NV0073_CTRL_DFP_DISP_MUX_FLAGS_SWITCH_TYPE_DGPU_TO_IGPU;
}

Because the second assignment replaces the whole flags word, the initial _SR_ENTER_SKIP field is dead code and the driver can never request NV0073_CTRL_DFP_DISP_MUX_FLAGS_SR_ENTER_SKIP_YES. Per the RM documentation in ctrl0073dfp.h, RM performs the PSR-enter sequence during pre-mux operations unless that bit is set — so on platforms where PSR-enter is unsupported, the entire pre-mux operation fails with NV_ERR_NOT_SUPPORTED (issue #1311 reporter hit exactly this).

Additionally, nvRmMuxPost() initializes the skip field with the _SR_ENTER_SKIP macro, but the POST command defines the field as _SR_EXIT_SKIP (bit 1:1 in both cases, so the compiled value is unchanged).

Changes

• nvRmMuxPre(): OR the switch type into params.flags instead of overwriting it (= → |=).
• nvRmMuxPost(): same = → |= fix, and use the correct _SR_EXIT_SKIP field in the initial DRF_DEF.

Verification

• Confirmed the flag macros in ctrl0073dfp.h: SWITCH_TYPE is bit 0:0 (IGPU_TO_DGPU=0x0, DGPU_TO_IGPU=0x1), SR_ENTER_SKIP/SR_EXIT_SKIP are bit 1:1 (_NO=0x0, _YES=0x1).
• DRF_DEF(...) expands to (value << DRF_SHIFT(field)); with _NO it yields 0, so with |= the resulting params.flags values are byte-identical to before (0x0 for iGPU→dGPU, 0x1 for dGPU→iGPU). No functional change for existing callers — the fix makes the intended flag composition actually reach the RM command and uses the correct field name for the POST command.

No open PR covers this change.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}