1/*	$NetBSD: hpc_machdep.c,v 1.70 2003/09/16 08:18:22 agc Exp $	*/
2
3/*-
4 * Copyright (c) 1994-1998 Mark Brinicombe.
5 * Copyright (c) 1994 Brini.
6 * All rights reserved.
7 *
8 * This code is derived from software written for Brini by Mark Brinicombe
9 *
10 * Redistribution and use in source and binary forms, with or without
11 * modification, are permitted provided that the following conditions
12 * are met:
13 * 1. Redistributions of source code must retain the above copyright
14 *    notice, this list of conditions and the following disclaimer.
15 * 2. Redistributions in binary form must reproduce the above copyright
16 *    notice, this list of conditions and the following disclaimer in the
17 *    documentation and/or other materials provided with the distribution.
18 * 3. All advertising materials mentioning features or use of this software
19 *    must display the following acknowledgement:
20 *      This product includes software developed by Brini.
21 * 4. The name of the company nor the name of the author may be used to
22 *    endorse or promote products derived from this software without specific
23 *    prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY BRINI ``AS IS'' AND ANY EXPRESS OR IMPLIED
26 * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27 * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
28 * IN NO EVENT SHALL BRINI OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
29 * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
30 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
31 * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * RiscBSD kernel project
38 *
39 * machdep.c
40 *
41 * Machine dependant functions for kernel setup
42 *
43 * This file needs a lot of work.
44 *
45 * Created      : 17/09/94
46 */
47
48#include "opt_ddb.h"
49
50#include <sys/cdefs.h>
51__FBSDID("$FreeBSD$");
52
53#define _ARM32_BUS_DMA_PRIVATE
54#include <sys/param.h>
55#include <sys/systm.h>
56#include <sys/sysproto.h>
57#include <sys/signalvar.h>
58#include <sys/imgact.h>
59#include <sys/kernel.h>
60#include <sys/ktr.h>
61#include <sys/linker.h>
62#include <sys/lock.h>
63#include <sys/malloc.h>
64#include <sys/mutex.h>
65#include <sys/pcpu.h>
66#include <sys/proc.h>
67#include <sys/ptrace.h>
68#include <sys/cons.h>
69#include <sys/bio.h>
70#include <sys/bus.h>
71#include <sys/buf.h>
72#include <sys/exec.h>
73#include <sys/kdb.h>
74#include <sys/msgbuf.h>
75#include <machine/reg.h>
76#include <machine/cpu.h>
77
78#include <vm/vm.h>
79#include <vm/pmap.h>
80#include <vm/vm_object.h>
81#include <vm/vm_page.h>
82#include <vm/vm_map.h>
83#include <machine/devmap.h>
84#include <machine/vmparam.h>
85#include <machine/pcb.h>
86#include <machine/undefined.h>
87#include <machine/machdep.h>
88#include <machine/metadata.h>
89#include <machine/armreg.h>
90#include <machine/bus.h>
91#include <machine/physmem.h>
92#include <sys/reboot.h>
93
94#include <arm/xscale/pxa/pxareg.h>
95#include <arm/xscale/pxa/pxavar.h>
96
97#define KERNEL_PT_SYS		0	/* Page table for mapping proc0 zero page */
98#define	KERNEL_PT_IOPXS		1
99#define KERNEL_PT_BEFOREKERN	2
100#define KERNEL_PT_AFKERNEL	3	/* L2 table for mapping after kernel */
101#define	KERNEL_PT_AFKERNEL_NUM	9
102
103/* this should be evenly divisable by PAGE_SIZE / L2_TABLE_SIZE_REAL (or 4) */
104#define NUM_KERNEL_PTS		(KERNEL_PT_AFKERNEL + KERNEL_PT_AFKERNEL_NUM)
105
106struct pv_addr kernel_pt_table[NUM_KERNEL_PTS];
107
108/* Physical and virtual addresses for some global pages */
109
110struct pv_addr systempage;
111struct pv_addr msgbufpv;
112struct pv_addr irqstack;
113struct pv_addr undstack;
114struct pv_addr abtstack;
115struct pv_addr kernelstack;
116struct pv_addr minidataclean;
117
118static void	pxa_probe_sdram(bus_space_tag_t, bus_space_handle_t,
119		    uint32_t *, uint32_t *);
120
121/* Static device mappings. */
122static const struct arm_devmap_entry pxa_devmap[] = {
123	/*
124	 * Map the on-board devices up into the KVA region so we don't muck
125	 * up user-space.
126	 */
127	{
128		PXA2X0_PERIPH_START + PXA2X0_PERIPH_OFFSET,
129		PXA2X0_PERIPH_START,
130		PXA250_PERIPH_END - PXA2X0_PERIPH_START,
131		VM_PROT_READ|VM_PROT_WRITE,
132		PTE_DEVICE,
133	},
134	{ 0, 0, 0, 0, 0, }
135};
136
137#define SDRAM_START 0xa0000000
138
139extern vm_offset_t xscale_cache_clean_addr;
140
141void *
142initarm(struct arm_boot_params *abp)
143{
144	struct pv_addr  kernel_l1pt;
145	struct pv_addr  dpcpu;
146	int loop;
147	u_int l1pagetable;
148	vm_offset_t freemempos;
149	vm_offset_t freemem_pt;
150	vm_offset_t afterkern;
151	vm_offset_t freemem_after;
152	vm_offset_t lastaddr;
153	int i, j;
154	uint32_t memsize[PXA2X0_SDRAM_BANKS], memstart[PXA2X0_SDRAM_BANKS];
155
156	lastaddr = parse_boot_param(abp);
157	arm_physmem_kernaddr = abp->abp_physaddr;
158	set_cpufuncs();
159	pcpu_init(pcpup, 0, sizeof(struct pcpu));
160	PCPU_SET(curthread, &thread0);
161
162	/* Do basic tuning, hz etc */
163	init_param1();
164
165	freemempos = 0xa0200000;
166	/* Define a macro to simplify memory allocation */
167#define	valloc_pages(var, np)			\
168	alloc_pages((var).pv_pa, (np));		\
169	(var).pv_va = (var).pv_pa + 0x20000000;
170
171#define alloc_pages(var, np)			\
172	freemempos -= (np * PAGE_SIZE);		\
173	(var) = freemempos;		\
174	memset((char *)(var), 0, ((np) * PAGE_SIZE));
175
176	while (((freemempos - L1_TABLE_SIZE) & (L1_TABLE_SIZE - 1)) != 0)
177		freemempos -= PAGE_SIZE;
178	valloc_pages(kernel_l1pt, L1_TABLE_SIZE / PAGE_SIZE);
179	for (loop = 0; loop < NUM_KERNEL_PTS; ++loop) {
180		if (!(loop % (PAGE_SIZE / L2_TABLE_SIZE_REAL))) {
181			valloc_pages(kernel_pt_table[loop],
182			    L2_TABLE_SIZE / PAGE_SIZE);
183		} else {
184			kernel_pt_table[loop].pv_pa = freemempos +
185			    (loop % (PAGE_SIZE / L2_TABLE_SIZE_REAL)) *
186			    L2_TABLE_SIZE_REAL;
187			kernel_pt_table[loop].pv_va =
188			    kernel_pt_table[loop].pv_pa + 0x20000000;
189		}
190	}
191	freemem_pt = freemempos;
192	freemempos = 0xa0100000;
193	/*
194	 * Allocate a page for the system page mapped to V0x00000000
195	 * This page will just contain the system vectors and can be
196	 * shared by all processes.
197	 */
198	valloc_pages(systempage, 1);
199
200	/* Allocate dynamic per-cpu area. */
201	valloc_pages(dpcpu, DPCPU_SIZE / PAGE_SIZE);
202	dpcpu_init((void *)dpcpu.pv_va, 0);
203
204	/* Allocate stacks for all modes */
205	valloc_pages(irqstack, IRQ_STACK_SIZE);
206	valloc_pages(abtstack, ABT_STACK_SIZE);
207	valloc_pages(undstack, UND_STACK_SIZE);
208	valloc_pages(kernelstack, KSTACK_PAGES);
209	alloc_pages(minidataclean.pv_pa, 1);
210	valloc_pages(msgbufpv, round_page(msgbufsize) / PAGE_SIZE);
211	/*
212	 * Allocate memory for the l1 and l2 page tables. The scheme to avoid
213	 * wasting memory by allocating the l1pt on the first 16k memory was
214	 * taken from NetBSD rpc_machdep.c. NKPT should be greater than 12 for
215	 * this to work (which is supposed to be the case).
216	 */
217
218	/*
219	 * Now we start construction of the L1 page table
220	 * We start by mapping the L2 page tables into the L1.
221	 * This means that we can replace L1 mappings later on if necessary
222	 */
223	l1pagetable = kernel_l1pt.pv_va;
224
225	/* Map the L2 pages tables in the L1 page table */
226	pmap_link_l2pt(l1pagetable, ARM_VECTORS_HIGH & ~(0x00100000 - 1),
227	    &kernel_pt_table[KERNEL_PT_SYS]);
228#if 0 /* XXXBJR: What is this?  Don't know if there's an analogue. */
229	pmap_link_l2pt(l1pagetable, IQ80321_IOPXS_VBASE,
230	                &kernel_pt_table[KERNEL_PT_IOPXS]);
231#endif
232	pmap_link_l2pt(l1pagetable, KERNBASE,
233	    &kernel_pt_table[KERNEL_PT_BEFOREKERN]);
234	pmap_map_chunk(l1pagetable, KERNBASE, SDRAM_START, 0x100000,
235	    VM_PROT_READ|VM_PROT_WRITE, PTE_CACHE);
236	pmap_map_chunk(l1pagetable, KERNBASE + 0x100000, SDRAM_START + 0x100000,
237	    0x100000, VM_PROT_READ|VM_PROT_WRITE, PTE_PAGETABLE);
238	pmap_map_chunk(l1pagetable, KERNBASE + 0x200000, SDRAM_START + 0x200000,
239	   (((uint32_t)(lastaddr) - KERNBASE - 0x200000) + L1_S_SIZE) & ~(L1_S_SIZE - 1),
240	    VM_PROT_READ|VM_PROT_WRITE, PTE_CACHE);
241	freemem_after = ((int)lastaddr + PAGE_SIZE) & ~(PAGE_SIZE - 1);
242	afterkern = round_page(((vm_offset_t)lastaddr + L1_S_SIZE) &
243	    ~(L1_S_SIZE - 1));
244	for (i = 0; i < KERNEL_PT_AFKERNEL_NUM; i++) {
245		pmap_link_l2pt(l1pagetable, afterkern + i * 0x00100000,
246		    &kernel_pt_table[KERNEL_PT_AFKERNEL + i]);
247	}
248	pmap_map_entry(l1pagetable, afterkern, minidataclean.pv_pa,
249	    VM_PROT_READ|VM_PROT_WRITE, PTE_CACHE);
250
251
252	/* Map the Mini-Data cache clean area. */
253	xscale_setup_minidata(l1pagetable, afterkern,
254	    minidataclean.pv_pa);
255
256	/* Map the vector page. */
257	pmap_map_entry(l1pagetable, ARM_VECTORS_HIGH, systempage.pv_pa,
258	    VM_PROT_READ|VM_PROT_WRITE, PTE_CACHE);
259	arm_devmap_bootstrap(l1pagetable, pxa_devmap);
260
261	/*
262	 * Give the XScale global cache clean code an appropriately
263	 * sized chunk of unmapped VA space starting at 0xff000000
264	 * (our device mappings end before this address).
265	 */
266	xscale_cache_clean_addr = 0xff000000U;
267
268	cpu_domains((DOMAIN_CLIENT << (PMAP_DOMAIN_KERNEL*2)) | DOMAIN_CLIENT);
269	setttb(kernel_l1pt.pv_pa);
270	cpu_tlb_flushID();
271	cpu_domains(DOMAIN_CLIENT << (PMAP_DOMAIN_KERNEL*2));
272
273	/*
274	 * Pages were allocated during the secondary bootstrap for the
275	 * stacks for different CPU modes.
276	 * We must now set the r13 registers in the different CPU modes to
277	 * point to these stacks.
278	 * Since the ARM stacks use STMFD etc. we must set r13 to the top end
279	 * of the stack memory.
280	 */
281	set_stackptrs(0);
282
283	/*
284	 * We must now clean the cache again....
285	 * Cleaning may be done by reading new data to displace any
286	 * dirty data in the cache. This will have happened in setttb()
287	 * but since we are boot strapping the addresses used for the read
288	 * may have just been remapped and thus the cache could be out
289	 * of sync. A re-clean after the switch will cure this.
290	 * After booting there are no gross relocations of the kernel thus
291	 * this problem will not occur after initarm().
292	 */
293	cpu_idcache_wbinv_all();
294	cpu_setup("");
295
296	/*
297	 * Sort out bus_space for on-board devices.
298	 */
299	pxa_obio_tag_init();
300
301	/*
302	 * Fetch the SDRAM start/size from the PXA2X0 SDRAM configration
303	 * registers.
304	 */
305	pxa_probe_sdram(obio_tag, PXA2X0_MEMCTL_BASE, memstart, memsize);
306
307	/* Fire up consoles. */
308	cninit();
309
310	undefined_init();
311
312	init_proc0(kernelstack.pv_va);
313
314	/* Enable MMU, I-cache, D-cache, write buffer. */
315	arm_vector_init(ARM_VECTORS_HIGH, ARM_VEC_ALL);
316
317	pmap_curmaxkvaddr = afterkern + PAGE_SIZE;
318	vm_max_kernel_address = 0xd0000000;
319	pmap_bootstrap(pmap_curmaxkvaddr, &kernel_l1pt);
320	msgbufp = (void*)msgbufpv.pv_va;
321	msgbufinit(msgbufp, msgbufsize);
322	mutex_init();
323
324	/*
325	 * Add the physical ram we have available.
326	 *
327	 * Exclude the kernel (and all the things we allocated which immediately
328	 * follow the kernel) from the VM allocation pool but not from crash
329	 * dumps.  virtual_avail is a global variable which tracks the kva we've
330	 * "allocated" while setting up pmaps.
331	 *
332	 * Prepare the list of physical memory available to the vm subsystem.
333	 */
334	for (j = 0; j < PXA2X0_SDRAM_BANKS; j++) {
335		if (memsize[j] > 0)
336			arm_physmem_hardware_region(memstart[j], memsize[j]);
337	}
338	arm_physmem_exclude_region(abp->abp_physaddr,
339	    virtual_avail - KERNVIRTADDR, EXFLAG_NOALLOC);
340	arm_physmem_init_kernel_globals();
341
342	init_param2(physmem);
343	kdb_init();
344	return ((void *)(kernelstack.pv_va + USPACE_SVC_STACK_TOP -
345	    sizeof(struct pcb)));
346}
347
348static void
349pxa_probe_sdram(bus_space_tag_t bst, bus_space_handle_t bsh,
350    uint32_t *memstart, uint32_t *memsize)
351{
352	uint32_t	mdcnfg, dwid, dcac, drac, dnb;
353	int		i;
354
355	mdcnfg = bus_space_read_4(bst, bsh, MEMCTL_MDCNFG);
356
357	/*
358	 * Scan all 4 SDRAM banks
359	 */
360	for (i = 0; i < PXA2X0_SDRAM_BANKS; i++) {
361		memstart[i] = 0;
362		memsize[i] = 0;
363
364		switch (i) {
365		case 0:
366		case 1:
367			if ((i == 0 && (mdcnfg & MDCNFG_DE0) == 0) ||
368			    (i == 1 && (mdcnfg & MDCNFG_DE1) == 0))
369				continue;
370			dwid = mdcnfg >> MDCNFD_DWID01_SHIFT;
371			dcac = mdcnfg >> MDCNFD_DCAC01_SHIFT;
372			drac = mdcnfg >> MDCNFD_DRAC01_SHIFT;
373			dnb = mdcnfg >> MDCNFD_DNB01_SHIFT;
374			break;
375
376		case 2:
377		case 3:
378			if ((i == 2 && (mdcnfg & MDCNFG_DE2) == 0) ||
379			    (i == 3 && (mdcnfg & MDCNFG_DE3) == 0))
380				continue;
381			dwid = mdcnfg >> MDCNFD_DWID23_SHIFT;
382			dcac = mdcnfg >> MDCNFD_DCAC23_SHIFT;
383			drac = mdcnfg >> MDCNFD_DRAC23_SHIFT;
384			dnb = mdcnfg >> MDCNFD_DNB23_SHIFT;
385			break;
386		default:
387			panic("pxa_probe_sdram: impossible");
388		}
389
390		dwid = 2 << (1 - (dwid & MDCNFD_DWID_MASK));  /* 16/32 width */
391		dcac = 1 << ((dcac & MDCNFD_DCAC_MASK) + 8);  /* 8-11 columns */
392		drac = 1 << ((drac & MDCNFD_DRAC_MASK) + 11); /* 11-13 rows */
393		dnb = 2 << (dnb & MDCNFD_DNB_MASK);	      /* # of banks */
394
395		memsize[i] = dwid * dcac * drac * dnb;
396		memstart[i] = PXA2X0_SDRAM0_START +
397		    (i * PXA2X0_SDRAM_BANK_SIZE);
398	}
399}
400
401#define	TIMER_FREQUENCY	3686400
402#define	UNIMPLEMENTED	panic("%s: unimplemented", __func__)
403
404/* XXXBJR: Belongs with DELAY in a timer.c of some sort. */
405void
406cpu_startprofclock(void)
407{
408	UNIMPLEMENTED;
409}
410
411void
412cpu_stopprofclock(void)
413{
414	UNIMPLEMENTED;
415}
416
417static struct arm32_dma_range pxa_range = {
418	.dr_sysbase = 0,
419	.dr_busbase = 0,
420	.dr_len = ~0u,
421};
422
423struct arm32_dma_range *
424bus_dma_get_range(void)
425{
426
427	return (&pxa_range);
428}
429
430int
431bus_dma_get_range_nb(void)
432{
433
434	return (1);
435}
436