provide support for submitting async dcmd frames.

async dcmds are submitted via an mpii request (like the scsi commands
are) which uses the ccb_request buffer, meaning that the dcmd itself
has to go somewhere else. this reuses the sense buffer on the ccb
for the dcmd, and provides wrappers for accessing that space and
submitting a dcmd via the passthru command.

1 files changed, 53 insertions, 1 deletions

diff --git a/sys/dev/pci/mfii.c b/sys/dev/pci/mfii.c
index c1ed2bc5dda..ca420e7d3be 100644
--- a/sys/dev/pci/mfii.c
+++ b/sys/dev/pci/mfii.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: mfii.c,v 1.35 2017/02/07 00:25:40 dlg Exp $ */
+/* $OpenBSD: mfii.c,v 1.36 2017/02/07 02:47:19 dlg Exp $ */
 
 /*
  * Copyright (c) 2012 David Gwynne <dlg@openbsd.org>
@@ -365,6 +365,40 @@ void			mfii_abort(struct mfii_softc *, struct mfii_ccb *,
 void			mfii_scsi_cmd_abort_done(struct mfii_softc *,
 			    struct mfii_ccb *);
 
+/*
+ * mfii boards support asynchronous (and non-polled) completion of
+ * dcmds by proxying them through a passthru mpii command that points
+ * at a dcmd frame. since the passthru command is submitted like
+ * the scsi commands using an SMID in the request descriptor,
+ * ccb_request memory * must contain the passthru command because
+ * that is what the SMID refers to. this means ccb_request cannot
+ * contain the dcmd. rather than allocating separate dma memory to
+ * hold the dcmd, we reuse the sense memory buffer for it.
+ */
+
+void			mfii_dcmd_start(struct mfii_softc *,
+			    struct mfii_ccb *);
+
+static inline void
+mfii_dcmd_scrub(struct mfii_ccb *ccb)
+{
+	memset(ccb->ccb_sense, 0, sizeof(*ccb->ccb_sense));
+}
+
+static inline struct mfi_dcmd_frame *
+mfii_dcmd_frame(struct mfii_ccb *ccb)
+{
+	CTASSERT(sizeof(struct mfi_dcmd_frame) <= sizeof(*ccb->ccb_sense));
+	return ((struct mfi_dcmd_frame *)ccb->ccb_sense);
+}
+
+static inline void
+mfii_dcmd_sync(struct mfii_softc *sc, struct mfii_ccb *ccb, int flags)
+{
+	bus_dmamap_sync(sc->sc_dmat, MFII_DMA_MAP(sc->sc_sense),
+	    ccb->ccb_sense_offset, sizeof(*ccb->ccb_sense), flags);
+}
+
 #define mfii_fw_state(_sc) mfii_read((_sc), MFI_OSP)
 
 const struct mfii_iop mfii_iop_thunderbolt = {
@@ -475,6 +509,7 @@ mfii_attach(struct device *parent, struct device *self, void *aux)
 	sc->sc_max_sgl = (status & MFI_STATE_MAXSGL_MASK) >> 16;
 
 	/* sense memory */
+	CTASSERT(sizeof(struct mfi_sense) == MFI_SENSE_SIZE);
 	sc->sc_sense = mfii_dmamem_alloc(sc, sc->sc_max_cmds * MFI_SENSE_SIZE);
 	if (sc->sc_sense == NULL) {
 		printf("%s: unable to allocate sense memory\n", DEVNAME(sc));
@@ -743,8 +778,25 @@ mfii_dmamem_free(struct mfii_softc *sc, struct mfii_dmamem *m)
 	free(m, M_DEVBUF, 0);
 }
 
+void
+mfii_dcmd_start(struct mfii_softc *sc, struct mfii_ccb *ccb)
+{
+	struct mpii_msg_scsi_io *io = ccb->ccb_request;
+	struct mfii_raid_context *ctx = (struct mfii_raid_context *)(io + 1);
+	struct mfii_sge *sge = (struct mfii_sge *)(ctx + 1);
+
+	io->function = MFII_FUNCTION_PASSTHRU_IO;
+	io->sgl_offset0 = (uint32_t *)sge - (uint32_t *)io;
+
+	htolem64(&sge->sg_addr, ccb->ccb_sense_dva);
+	htolem32(&sge->sg_len, sizeof(*ccb->ccb_sense));
+	sge->sg_flags = MFII_SGE_CHAIN_ELEMENT | MFII_SGE_ADDR_IOCPLBNTA;
 
+	ccb->ccb_req.flags = MFII_REQ_TYPE_SCSI;
+	ccb->ccb_req.smid = letoh16(ccb->ccb_smid);
 
+	mfii_start(sc, ccb);
+}
 
 int
 mfii_transition_firmware(struct mfii_softc *sc)