HMA/UMA bounce of int 25h doesn't actually move the data in blockio.c

[ecm-pushbx]

This is the affected part:

kernel/kernel/blockio.c

Lines 435 to 444 in 5de2eb1

	if (FP_SEG(buf) >= 0xa000 && numblocks == 1 && bufloc != LOC_CONV)
	{
	IoReqHdr.r_trans = deblock_buf;
	if (mode == DSKWRITE)
	fmemcpy(deblock_buf, buf, dpbp->dpb_secsize);
	execrh((request FAR *) & IoReqHdr, dpbp->dpb_device);
	if (mode == DSKREAD)
	fmemcpy(buf, deblock_buf, dpbp->dpb_secsize);
	}
	else

If I call int 25h using the lDebug l F001:0 A: 0 1 command (there's a free UMB in my dosemu2 setup at this address) then this code is tripped. The problem is that according to

kernel/kernel/blockio.c

Lines 412 to 414 in 5de2eb1

	case DSKREADINT25:
	case DSKREAD:
	IoReqHdr.r_command = C_INPUT;

mode may be DSKREADINT25 but this isn't recognised by the bounce code, which expects only DSKREAD for a read.

This can also be demonstrated by running the lDebug command l F001:0 A: 0 2 which doesn't trip the bounce (despite it likely should) and actually reads the data into the UMB.

The write side for int 26h is likely affected in the same way.