1/*
2 * TX ucode for the Intel IXP2400 in POS-PHY mode.
3 * Copyright (C) 2004, 2005 Lennert Buytenhek
4 * Dedicated to Marija Kulikova.
5 *
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (at your option) any later version.
10 *
11 * Assumptions made in this code:
12 * - The IXP2400 MSF is configured for POS-PHY mode, in a mode where
13 *   only one TBUF partition is used.  This includes, for example,
14 *   1x32 SPHY and 1x32 MPHY32, but not 4x8 SPHY or 1x32 MPHY4. (This
15 *   is not an exhaustive list.)
16 * - The TBUF uses 64-byte mpackets.
17 * - TX descriptors reside in SRAM, and have the following format:
18 *	struct tx_desc
19 *	{
20 *	// to uengine
21 *		u32	buf_phys_addr;
22 *		u32	pkt_length;
23 *		u32	channel;
24 *	};
25 * - Packet data resides in DRAM.
26 * - Packet buffer addresses are 8-byte aligned.
27 * - Scratch ring 2 is tx_pending.
28 * - Scratch ring 3 is tx_done, and has status condition 'full'.
29 * - This code is run on all eight threads of the microengine it runs on.
30 */
31
32#define TX_SEQUENCE_0		0x0060
33#define TBUF_CTRL		0x1800
34
35#define PARTITION_SIZE		128
36#define PARTITION_THRESH	96
37
38
39	.sig volatile sig1
40	.sig volatile sig2
41	.sig volatile sig3
42
43	.reg @old_tx_seq_0
44	.reg @mpkts_in_flight
45	.reg @next_tbuf_mpacket
46
47	.reg @buffer_handle
48	.reg @buffer_start
49	.reg @packet_length
50	.reg @channel
51	.reg @packet_offset
52
53	.reg zero
54
55	immed[zero, 0]
56
57	/*
58	 * Skip context 0 initialisation?
59	 */
60	.begin
61		br!=ctx[0, mpacket_tx_loop#]
62	.end
63
64	/*
65	 * Wait until all pending TBUF elements have been transmitted.
66	 */
67	.begin
68		.reg read $tx
69		.sig zzz
70
71	loop_empty#:
72		msf[read, $tx, zero, TX_SEQUENCE_0, 1], ctx_swap[zzz]
73		alu_shf[--, --, b, $tx, >>31]
74		beq[loop_empty#]
75
76		alu[@old_tx_seq_0, --, b, $tx]
77	.end
78
79	immed[@mpkts_in_flight, 0]
80	alu[@next_tbuf_mpacket, @old_tx_seq_0, and, (PARTITION_SIZE - 1)]
81
82	immed[@buffer_handle, 0]
83
84	/*
85	 * Initialise signal pipeline.
86	 */
87	.begin
88		local_csr_wr[SAME_ME_SIGNAL, (&sig1 << 3)]
89		.set_sig sig1
90
91		local_csr_wr[SAME_ME_SIGNAL, (&sig2 << 3)]
92		.set_sig sig2
93
94		local_csr_wr[SAME_ME_SIGNAL, (&sig3 << 3)]
95		.set_sig sig3
96	.end
97
98mpacket_tx_loop#:
99	.begin
100		.reg tbuf_element_index
101		.reg buffer_handle
102		.reg sop_eop
103		.reg packet_data
104		.reg channel
105		.reg mpacket_size
106
107		/*
108		 * If there is no packet currently being transmitted,
109		 * dequeue the next TX descriptor, and fetch the buffer
110		 * address, packet length and destination channel number.
111		 */
112		.begin
113			.reg read $stemp $stemp2 $stemp3
114			.xfer_order $stemp $stemp2 $stemp3
115			.sig zzz
116
117			ctx_arb[sig1]
118
119			alu[--, --, b, @buffer_handle]
120			bne[already_got_packet#]
121
122		tx_nobufs#:
123			scratch[get, $stemp, zero, 8, 1], ctx_swap[zzz]
124			alu[@buffer_handle, --, b, $stemp]
125			beq[tx_nobufs#]
126
127			sram[read, $stemp, $stemp, 0, 3], ctx_swap[zzz]
128			alu[@buffer_start, --, b, $stemp]
129			alu[@packet_length, --, b, $stemp2]
130			beq[zero_byte_packet#]
131			alu[@channel, --, b, $stemp3]
132			immed[@packet_offset, 0]
133
134		already_got_packet#:
135			local_csr_wr[SAME_ME_SIGNAL, (0x80 | (&sig1 << 3))]
136		.end
137
138		/*
139		 * Determine tbuf element index, SOP/EOP flags, mpacket
140		 * offset and mpacket size and cache buffer_handle and
141		 * channel number.
142		 */
143		.begin
144			alu[tbuf_element_index, --, b, @next_tbuf_mpacket]
145			alu[@next_tbuf_mpacket, @next_tbuf_mpacket, +, 1]
146			alu[@next_tbuf_mpacket, @next_tbuf_mpacket, and,
147							(PARTITION_SIZE - 1)]
148
149			alu[buffer_handle, --, b, @buffer_handle]
150			immed[@buffer_handle, 0]
151
152			immed[sop_eop, 1]
153
154			alu[packet_data, --, b, @packet_offset]
155			bne[no_sop#]
156			alu[sop_eop, sop_eop, or, 2]
157		no_sop#:
158			alu[packet_data, packet_data, +, @buffer_start]
159
160			alu[channel, --, b, @channel]
161
162			alu[mpacket_size, @packet_length, -, @packet_offset]
163			alu[--, 64, -, mpacket_size]
164			bhs[eop#]
165			alu[@buffer_handle, --, b, buffer_handle]
166			immed[mpacket_size, 64]
167			alu[sop_eop, sop_eop, and, 2]
168		eop#:
169
170			alu[@packet_offset, @packet_offset, +, mpacket_size]
171		.end
172
173		/*
174		 * Wait until there's enough space in the TBUF.
175		 */
176		.begin
177			.reg read $tx
178			.reg temp
179			.sig zzz
180
181			ctx_arb[sig2]
182
183			br[test_space#]
184
185		loop_space#:
186			msf[read, $tx, zero, TX_SEQUENCE_0, 1], ctx_swap[zzz]
187
188			alu[temp, $tx, -, @old_tx_seq_0]
189			alu[temp, temp, and, 0xff]
190			alu[@mpkts_in_flight, @mpkts_in_flight, -, temp]
191
192			alu[@old_tx_seq_0, --, b, $tx]
193
194		test_space#:
195			alu[--, PARTITION_THRESH, -, @mpkts_in_flight]
196			blo[loop_space#]
197
198			alu[@mpkts_in_flight, @mpkts_in_flight, +, 1]
199
200			local_csr_wr[SAME_ME_SIGNAL, (0x80 | (&sig2 << 3))]
201		.end
202
203		/*
204		 * Copy the packet data to the TBUF.
205		 */
206		.begin
207			.reg temp
208			.sig copy_sig
209
210			alu[temp, mpacket_size, -, 1]
211			alu_shf[temp, 0x10, or, temp, >>3]
212			alu_shf[temp, 0x10, or, temp, <<21]
213			alu_shf[temp, temp, or, tbuf_element_index, <<11]
214			alu_shf[--, temp, or, 1, <<18]
215
216			dram[tbuf_wr, --, packet_data, 0, max_8],
217					indirect_ref, sig_done[copy_sig]
218			ctx_arb[copy_sig]
219		.end
220
221		/*
222		 * Mark TBUF element as ready-to-be-transmitted.
223		 */
224		.begin
225			.reg write $tsw $tsw2
226			.xfer_order $tsw $tsw2
227			.reg temp
228			.sig zzz
229
230			alu_shf[temp, channel, or, mpacket_size, <<24]
231			alu_shf[$tsw, temp, or, sop_eop, <<8]
232			immed[$tsw2, 0]
233
234			immed[temp, TBUF_CTRL]
235			alu_shf[temp, temp, or, tbuf_element_index, <<3]
236			msf[write, $tsw, temp, 0, 2], ctx_swap[zzz]
237		.end
238
239		/*
240		 * Resynchronise.
241		 */
242		.begin
243			ctx_arb[sig3]
244			local_csr_wr[SAME_ME_SIGNAL, (0x80 | (&sig3 << 3))]
245		.end
246
247		/*
248		 * If this was an EOP mpacket, recycle the TX buffer
249	 	 * and signal the host.
250		 */
251		.begin
252			.reg write $stemp
253			.sig zzz
254
255			alu[--, sop_eop, and, 1]
256			beq[mpacket_tx_loop#]
257
258		tx_done_ring_full#:
259			br_inp_state[SCR_Ring3_Status, tx_done_ring_full#]
260
261			alu[$stemp, --, b, buffer_handle]
262			scratch[put, $stemp, zero, 12, 1], ctx_swap[zzz]
263			cap[fast_wr, 0, XSCALE_INT_A]
264			br[mpacket_tx_loop#]
265		.end
266	.end
267
268
269zero_byte_packet#:
270	halt
271
272
273