[deliverable/linux.git] / arch / powerpc / mm / hugepage-hash64.c

/*
 * Copyright IBM Corporation, 2013
 * Author Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
 *
 * This program is free software; you can redistribute it and/or modify it
 * under the terms of version 2.1 of the GNU Lesser General Public License
 * as published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it would be useful, but
 * WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
 *
 */

/*
 * PPC64 THP Support for hash based MMUs
 */
#include <linux/mm.h>
#include <asm/machdep.h>

static void invalidate_old_hpte(unsigned long vsid, unsigned long addr,
				pmd_t *pmdp, unsigned int psize, int ssize)
{
	int i, max_hpte_count, valid;
	unsigned long s_addr;
	unsigned char *hpte_slot_array;
	unsigned long hidx, shift, vpn, hash, slot;

	s_addr = addr & HPAGE_PMD_MASK;
	hpte_slot_array = get_hpte_slot_array(pmdp);
	/*
	 * IF we try to do a HUGE PTE update after a withdraw is done.
	 * we will find the below NULL. This happens when we do
	 * split_huge_page_pmd
	 */
	if (!hpte_slot_array)
		return;

	if (ppc_md.hugepage_invalidate)
		return ppc_md.hugepage_invalidate(vsid, s_addr, hpte_slot_array,
						  psize, ssize);
	/*
	 * No bluk hpte removal support, invalidate each entry
	 */
	shift = mmu_psize_defs[psize].shift;
	max_hpte_count = HPAGE_PMD_SIZE >> shift;
	for (i = 0; i < max_hpte_count; i++) {
		/*
		 * 8 bits per each hpte entries
		 * 000| [ secondary group (one bit) | hidx (3 bits) | valid bit]
		 */
		valid = hpte_valid(hpte_slot_array, i);
		if (!valid)
			continue;
		hidx =  hpte_hash_index(hpte_slot_array, i);

		/* get the vpn */
		addr = s_addr + (i * (1ul << shift));
		vpn = hpt_vpn(addr, vsid, ssize);
		hash = hpt_hash(vpn, shift, ssize);
		if (hidx & _PTEIDX_SECONDARY)
			hash = ~hash;

		slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
		slot += hidx & _PTEIDX_GROUP_IX;
		ppc_md.hpte_invalidate(slot, vpn, psize,
				       MMU_PAGE_16M, ssize, 0);
	}
}


int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid,
		    pmd_t *pmdp, unsigned long trap, int local, int ssize,
		    unsigned int psize)
{
	unsigned int index, valid;
	unsigned char *hpte_slot_array;
	unsigned long rflags, pa, hidx;
	unsigned long old_pmd, new_pmd;
	int ret, lpsize = MMU_PAGE_16M;
	unsigned long vpn, hash, shift, slot;

	/*
	 * atomically mark the linux large page PMD busy and dirty
	 */
	do {
		pmd_t pmd = ACCESS_ONCE(*pmdp);

		old_pmd = pmd_val(pmd);
		/* If PMD busy, retry the access */
		if (unlikely(old_pmd & _PAGE_BUSY))
			return 0;
		/* If PMD is trans splitting retry the access */
		if (unlikely(old_pmd & _PAGE_SPLITTING))
			return 0;
		/* If PMD permissions don't match, take page fault */
		if (unlikely(access & ~old_pmd))
			return 1;
		/*
		 * Try to lock the PTE, add ACCESSED and DIRTY if it was
		 * a write access
		 */
		new_pmd = old_pmd | _PAGE_BUSY | _PAGE_ACCESSED;
		if (access & _PAGE_RW)
			new_pmd |= _PAGE_DIRTY;
	} while (old_pmd != __cmpxchg_u64((unsigned long *)pmdp,
					  old_pmd, new_pmd));
	/*
	 * PP bits. _PAGE_USER is already PP bit 0x2, so we only
	 * need to add in 0x1 if it's a read-only user page
	 */
	rflags = new_pmd & _PAGE_USER;
	if ((new_pmd & _PAGE_USER) && !((new_pmd & _PAGE_RW) &&
					   (new_pmd & _PAGE_DIRTY)))
		rflags |= 0x1;
	/*
	 * _PAGE_EXEC -> HW_NO_EXEC since it's inverted
	 */
	rflags |= ((new_pmd & _PAGE_EXEC) ? 0 : HPTE_R_N);

#if 0
	if (!cpu_has_feature(CPU_FTR_COHERENT_ICACHE)) {

		/*
		 * No CPU has hugepages but lacks no execute, so we
		 * don't need to worry about that case
		 */
		rflags = hash_page_do_lazy_icache(rflags, __pte(old_pte), trap);
	}
#endif
	/*
	 * Find the slot index details for this ea, using base page size.
	 */
	shift = mmu_psize_defs[psize].shift;
	index = (ea & ~HPAGE_PMD_MASK) >> shift;
	BUG_ON(index >= 4096);

	vpn = hpt_vpn(ea, vsid, ssize);
	hash = hpt_hash(vpn, shift, ssize);
	hpte_slot_array = get_hpte_slot_array(pmdp);
	if (psize == MMU_PAGE_4K) {
		/*
		 * invalidate the old hpte entry if we have that mapped via 64K
		 * base page size. This is because demote_segment won't flush
		 * hash page table entries.
		 */
		if ((old_pmd & _PAGE_HASHPTE) && !(old_pmd & _PAGE_COMBO))
			invalidate_old_hpte(vsid, ea, pmdp, MMU_PAGE_64K, ssize);
	}

	valid = hpte_valid(hpte_slot_array, index);
	if (valid) {
		/* update the hpte bits */
		hidx =  hpte_hash_index(hpte_slot_array, index);
		if (hidx & _PTEIDX_SECONDARY)
			hash = ~hash;
		slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
		slot += hidx & _PTEIDX_GROUP_IX;

		ret = ppc_md.hpte_updatepp(slot, rflags, vpn,
					   psize, lpsize, ssize, local);
		/*
		 * We failed to update, try to insert a new entry.
		 */
		if (ret == -1) {
			/*
			 * large pte is marked busy, so we can be sure
			 * nobody is looking at hpte_slot_array. hence we can
			 * safely update this here.
			 */
			valid = 0;
			hpte_slot_array[index] = 0;
		}
	}

	if (!valid) {
		unsigned long hpte_group;

		/* insert new entry */
		pa = pmd_pfn(__pmd(old_pmd)) << PAGE_SHIFT;
		new_pmd |= _PAGE_HASHPTE;

		/* Add in WIMG bits */
		rflags |= (new_pmd & (_PAGE_WRITETHRU | _PAGE_NO_CACHE |
				      _PAGE_GUARDED));
		/*
		 * enable the memory coherence always
		 */
		rflags |= HPTE_R_M;
repeat:
		hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL;

		/* Insert into the hash table, primary slot */
		slot = ppc_md.hpte_insert(hpte_group, vpn, pa, rflags, 0,
					  psize, lpsize, ssize);
		/*
		 * Primary is full, try the secondary
		 */
		if (unlikely(slot == -1)) {
			hpte_group = ((~hash & htab_hash_mask) *
				      HPTES_PER_GROUP) & ~0x7UL;
			slot = ppc_md.hpte_insert(hpte_group, vpn, pa,
						  rflags, HPTE_V_SECONDARY,
						  psize, lpsize, ssize);
			if (slot == -1) {
				if (mftb() & 0x1)
					hpte_group = ((hash & htab_hash_mask) *
						      HPTES_PER_GROUP) & ~0x7UL;

				ppc_md.hpte_remove(hpte_group);
				goto repeat;
			}
		}
		/*
		 * Hypervisor failure. Restore old pmd and return -1
		 * similar to __hash_page_*
		 */
		if (unlikely(slot == -2)) {
			*pmdp = __pmd(old_pmd);
			hash_failure_debug(ea, access, vsid, trap, ssize,
					   psize, lpsize, old_pmd);
			return -1;
		}
		/*
		 * large pte is marked busy, so we can be sure
		 * nobody is looking at hpte_slot_array. hence we can
		 * safely update this here.
		 */
		mark_hpte_slot_valid(hpte_slot_array, index, slot);
	}
	/*
	 * Mark the pte with _PAGE_COMBO, if we are trying to hash it with
	 * base page size 4k.
	 */
	if (psize == MMU_PAGE_4K)
		new_pmd |= _PAGE_COMBO;
	/*
	 * The hpte valid is stored in the pgtable whose address is in the
	 * second half of the PMD. Order this against clearing of the busy bit in
	 * huge pmd.
	 */
	smp_wmb();
	*pmdp = __pmd(new_pmd & ~_PAGE_BUSY);
	return 0;
}
Commit	Line	Data
6d492ecc AK	1	/*
	2	* Copyright IBM Corporation, 2013
	3	* Author Aneesh Kumar K.V <aneesh.kumar@linux.vnet.ibm.com>
	4	*
	5	* This program is free software; you can redistribute it and/or modify it
	6	* under the terms of version 2.1 of the GNU Lesser General Public License
	7	* as published by the Free Software Foundation.
	8	*
	9	* This program is distributed in the hope that it would be useful, but
	10	* WITHOUT ANY WARRANTY; without even the implied warranty of
	11	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
	12	*
	13	*/
	14
	15	/*
	16	* PPC64 THP Support for hash based MMUs
	17	*/
	18	#include <linux/mm.h>
	19	#include <asm/machdep.h>
	20
629149fa AK	21	static void invalidate_old_hpte(unsigned long vsid, unsigned long addr,
	22	pmd_t *pmdp, unsigned int psize, int ssize)
	23	{
	24	int i, max_hpte_count, valid;
	25	unsigned long s_addr;
	26	unsigned char *hpte_slot_array;
	27	unsigned long hidx, shift, vpn, hash, slot;
	28
	29	s_addr = addr & HPAGE_PMD_MASK;
	30	hpte_slot_array = get_hpte_slot_array(pmdp);
	31	/*
	32	* IF we try to do a HUGE PTE update after a withdraw is done.
	33	* we will find the below NULL. This happens when we do
	34	* split_huge_page_pmd
	35	*/
	36	if (!hpte_slot_array)
	37	return;
	38
	39	if (ppc_md.hugepage_invalidate)
	40	return ppc_md.hugepage_invalidate(vsid, s_addr, hpte_slot_array,
	41	psize, ssize);
	42	/*
	43	* No bluk hpte removal support, invalidate each entry
	44	*/
	45	shift = mmu_psize_defs[psize].shift;
	46	max_hpte_count = HPAGE_PMD_SIZE >> shift;
	47	for (i = 0; i < max_hpte_count; i++) {
	48	/*
	49	* 8 bits per each hpte entries
	50	* 000\| [ secondary group (one bit) \| hidx (3 bits) \| valid bit]
	51	*/
	52	valid = hpte_valid(hpte_slot_array, i);
	53	if (!valid)
	54	continue;
	55	hidx = hpte_hash_index(hpte_slot_array, i);
	56
	57	/* get the vpn */
	58	addr = s_addr + (i * (1ul << shift));
	59	vpn = hpt_vpn(addr, vsid, ssize);
	60	hash = hpt_hash(vpn, shift, ssize);
	61	if (hidx & _PTEIDX_SECONDARY)
	62	hash = ~hash;
	63
	64	slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
	65	slot += hidx & _PTEIDX_GROUP_IX;
	66	ppc_md.hpte_invalidate(slot, vpn, psize,
	67	MMU_PAGE_16M, ssize, 0);
	68	}
	69	}
	70
	71
6d492ecc AK	72	int __hash_page_thp(unsigned long ea, unsigned long access, unsigned long vsid,
	73	pmd_t *pmdp, unsigned long trap, int local, int ssize,
	74	unsigned int psize)
	75	{
	76	unsigned int index, valid;
	77	unsigned char *hpte_slot_array;
	78	unsigned long rflags, pa, hidx;
	79	unsigned long old_pmd, new_pmd;
	80	int ret, lpsize = MMU_PAGE_16M;
	81	unsigned long vpn, hash, shift, slot;
	82
	83	/*
	84	* atomically mark the linux large page PMD busy and dirty
	85	*/
	86	do {
7e467245 AK	87	pmd_t pmd = ACCESS_ONCE(*pmdp);
	88
	89	old_pmd = pmd_val(pmd);
6d492ecc AK	90	/* If PMD busy, retry the access */
	91	if (unlikely(old_pmd & _PAGE_BUSY))
	92	return 0;
0ac52dd7 AK	93	/* If PMD is trans splitting retry the access */
	94	if (unlikely(old_pmd & _PAGE_SPLITTING))
	95	return 0;
6d492ecc AK	96	/* If PMD permissions don't match, take page fault */
	97	if (unlikely(access & ~old_pmd))
	98	return 1;
	99	/*
	100	* Try to lock the PTE, add ACCESSED and DIRTY if it was
	101	* a write access
	102	*/
	103	new_pmd = old_pmd \| _PAGE_BUSY \| _PAGE_ACCESSED;
	104	if (access & _PAGE_RW)
	105	new_pmd \|= _PAGE_DIRTY;
	106	} while (old_pmd != __cmpxchg_u64((unsigned long *)pmdp,
	107	old_pmd, new_pmd));
	108	/*
	109	* PP bits. _PAGE_USER is already PP bit 0x2, so we only
	110	* need to add in 0x1 if it's a read-only user page
	111	*/
	112	rflags = new_pmd & _PAGE_USER;
	113	if ((new_pmd & _PAGE_USER) && !((new_pmd & _PAGE_RW) &&
	114	(new_pmd & _PAGE_DIRTY)))
	115	rflags \|= 0x1;
	116	/*
	117	* _PAGE_EXEC -> HW_NO_EXEC since it's inverted
	118	*/
	119	rflags \|= ((new_pmd & _PAGE_EXEC) ? 0 : HPTE_R_N);
	120
	121	#if 0
	122	if (!cpu_has_feature(CPU_FTR_COHERENT_ICACHE)) {
	123
	124	/*
	125	* No CPU has hugepages but lacks no execute, so we
	126	* don't need to worry about that case
	127	*/
	128	rflags = hash_page_do_lazy_icache(rflags, __pte(old_pte), trap);
	129	}
	130	#endif
	131	/*
	132	* Find the slot index details for this ea, using base page size.
	133	*/
	134	shift = mmu_psize_defs[psize].shift;
	135	index = (ea & ~HPAGE_PMD_MASK) >> shift;
	136	BUG_ON(index >= 4096);
	137
	138	vpn = hpt_vpn(ea, vsid, ssize);
	139	hash = hpt_hash(vpn, shift, ssize);
	140	hpte_slot_array = get_hpte_slot_array(pmdp);
629149fa AK	141	if (psize == MMU_PAGE_4K) {
	142	/*
	143	* invalidate the old hpte entry if we have that mapped via 64K
	144	* base page size. This is because demote_segment won't flush
	145	* hash page table entries.
	146	*/
	147	if ((old_pmd & _PAGE_HASHPTE) && !(old_pmd & _PAGE_COMBO))
	148	invalidate_old_hpte(vsid, ea, pmdp, MMU_PAGE_64K, ssize);
	149	}
6d492ecc AK	150
	151	valid = hpte_valid(hpte_slot_array, index);
	152	if (valid) {
	153	/* update the hpte bits */
	154	hidx = hpte_hash_index(hpte_slot_array, index);
	155	if (hidx & _PTEIDX_SECONDARY)
	156	hash = ~hash;
	157	slot = (hash & htab_hash_mask) * HPTES_PER_GROUP;
	158	slot += hidx & _PTEIDX_GROUP_IX;
	159
	160	ret = ppc_md.hpte_updatepp(slot, rflags, vpn,
	161	psize, lpsize, ssize, local);
	162	/*
	163	* We failed to update, try to insert a new entry.
	164	*/
	165	if (ret == -1) {
	166	/*
	167	* large pte is marked busy, so we can be sure
	168	* nobody is looking at hpte_slot_array. hence we can
	169	* safely update this here.
	170	*/
	171	valid = 0;
6d492ecc	172	hpte_slot_array[index] = 0;
629149fa	173	}
6d492ecc AK	174	}
	175
	176	if (!valid) {
	177	unsigned long hpte_group;
	178
	179	/* insert new entry */
	180	pa = pmd_pfn(__pmd(old_pmd)) << PAGE_SHIFT;
629149fa	181	new_pmd \|= _PAGE_HASHPTE;
6d492ecc AK	182
	183	/* Add in WIMG bits */
	184	rflags \|= (new_pmd & (_PAGE_WRITETHRU \| _PAGE_NO_CACHE \|
c8c06f5a AK	185	_PAGE_GUARDED));
	186	/*
	187	* enable the memory coherence always
	188	*/
	189	rflags \|= HPTE_R_M;
629149fa AK	190	repeat:
629149fa AK	191	hpte_group = ((hash & htab_hash_mask) * HPTES_PER_GROUP) & ~0x7UL;
6d492ecc AK	192
	193	/* Insert into the hash table, primary slot */
	194	slot = ppc_md.hpte_insert(hpte_group, vpn, pa, rflags, 0,
	195	psize, lpsize, ssize);
	196	/*
	197	* Primary is full, try the secondary
	198	*/
	199	if (unlikely(slot == -1)) {
	200	hpte_group = ((~hash & htab_hash_mask) *
	201	HPTES_PER_GROUP) & ~0x7UL;
	202	slot = ppc_md.hpte_insert(hpte_group, vpn, pa,
	203	rflags, HPTE_V_SECONDARY,
	204	psize, lpsize, ssize);
	205	if (slot == -1) {
	206	if (mftb() & 0x1)
	207	hpte_group = ((hash & htab_hash_mask) *
	208	HPTES_PER_GROUP) & ~0x7UL;
	209
	210	ppc_md.hpte_remove(hpte_group);
	211	goto repeat;
	212	}
	213	}
	214	/*
	215	* Hypervisor failure. Restore old pmd and return -1
	216	* similar to __hash_page_*
	217	*/
	218	if (unlikely(slot == -2)) {
	219	*pmdp = __pmd(old_pmd);
	220	hash_failure_debug(ea, access, vsid, trap, ssize,
	221	psize, lpsize, old_pmd);
	222	return -1;
	223	}
	224	/*
	225	* large pte is marked busy, so we can be sure
	226	* nobody is looking at hpte_slot_array. hence we can
	227	* safely update this here.
	228	*/
	229	mark_hpte_slot_valid(hpte_slot_array, index, slot);
	230	}
629149fa AK	231	/*
	232	* Mark the pte with _PAGE_COMBO, if we are trying to hash it with
	233	* base page size 4k.
	234	*/
	235	if (psize == MMU_PAGE_4K)
	236	new_pmd \|= _PAGE_COMBO;
6d492ecc	237	/*
b0aa44a3 AK	238	* The hpte valid is stored in the pgtable whose address is in the
	239	* second half of the PMD. Order this against clearing of the busy bit in
	240	* huge pmd.
6d492ecc	241	*/
b0aa44a3	242	smp_wmb();
6d492ecc AK	243	*pmdp = __pmd(new_pmd & ~_PAGE_BUSY);
	244	return 0;
	245	}