Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel…

…/git/s390/linux

Pull more s390 updates from Martin Schwidefsky:
 "The second patch set for the 4.14 merge window:

   - Convert the dasd device driver to the blk-mq interface.

   - Provide three zcrypt interfaces for vfio_ap. These will be required
     for KVM guest access to the crypto cards attached via the AP bus.

   - A couple of memory management bug fixes."

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/dasd: blk-mq conversion
  s390/mm: use a single lock for the fields in mm_context_t
  s390/mm: fix race on mm->context.flush_mm
  s390/mm: fix local TLB flushing vs. detach of an mm address space
  s390/zcrypt: externalize AP queue interrupt control
  s390/zcrypt: externalize AP config info query
  s390/zcrypt: externalize test AP queue
  s390/mm: use VM_BUG_ON in crst_table_[upgrade|downgrade]

arch/s390/include/asm/ap.h

@@ -0,0 +1,126 @@
+/*
+ * Adjunct processor (AP) interfaces
+ *
+ * Copyright IBM Corp. 2017
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License (version 2 only)
+ * as published by the Free Software Foundation.
+ *
+ * Author(s): Tony Krowiak <akrowia@linux.vnet.ibm.com>
+ *	      Martin Schwidefsky <schwidefsky@de.ibm.com>
+ *	      Harald Freudenberger <freude@de.ibm.com>
+ */
+
+#ifndef _ASM_S390_AP_H_
+#define _ASM_S390_AP_H_
+
+/**
+ * The ap_qid_t identifier of an ap queue.
+ * If the AP facilities test (APFT) facility is available,
+ * card and queue index are 8 bit values, otherwise
+ * card index is 6 bit and queue index a 4 bit value.
+ */
+typedef unsigned int ap_qid_t;
+
+#define AP_MKQID(_card, _queue) (((_card) & 63) << 8 | ((_queue) & 255))
+#define AP_QID_CARD(_qid) (((_qid) >> 8) & 63)
+#define AP_QID_QUEUE(_qid) ((_qid) & 255)
+
+/**
+ * struct ap_queue_status - Holds the AP queue status.
+ * @queue_empty: Shows if queue is empty
+ * @replies_waiting: Waiting replies
+ * @queue_full: Is 1 if the queue is full
+ * @irq_enabled: Shows if interrupts are enabled for the AP
+ * @response_code: Holds the 8 bit response code
+ *
+ * The ap queue status word is returned by all three AP functions
+ * (PQAP, NQAP and DQAP).  There's a set of flags in the first
+ * byte, followed by a 1 byte response code.
+ */
+struct ap_queue_status {
+	unsigned int queue_empty	: 1;
+	unsigned int replies_waiting	: 1;
+	unsigned int queue_full		: 1;
+	unsigned int _pad1		: 4;
+	unsigned int irq_enabled	: 1;
+	unsigned int response_code	: 8;
+	unsigned int _pad2		: 16;
+};
+
+/**
+ * ap_test_queue(): Test adjunct processor queue.
+ * @qid: The AP queue number
+ * @tbit: Test facilities bit
+ * @info: Pointer to queue descriptor
+ *
+ * Returns AP queue status structure.
+ */
+struct ap_queue_status ap_test_queue(ap_qid_t qid,
+				     int tbit,
+				     unsigned long *info);
+
+struct ap_config_info {
+	unsigned int apsc	 : 1;	/* S bit */
+	unsigned int apxa	 : 1;	/* N bit */
+	unsigned int qact	 : 1;	/* C bit */
+	unsigned int rc8a	 : 1;	/* R bit */
+	unsigned char _reserved1 : 4;
+	unsigned char _reserved2[3];
+	unsigned char Na;		/* max # of APs - 1 */
+	unsigned char Nd;		/* max # of Domains - 1 */
+	unsigned char _reserved3[10];
+	unsigned int apm[8];		/* AP ID mask */
+	unsigned int aqm[8];		/* AP queue mask */
+	unsigned int adm[8];		/* AP domain mask */
+	unsigned char _reserved4[16];
+} __aligned(8);
+
+/*
+ * ap_query_configuration(): Fetch cryptographic config info
+ *
+ * Returns the ap configuration info fetched via PQAP(QCI).
+ * On success 0 is returned, on failure a negative errno
+ * is returned, e.g. if the PQAP(QCI) instruction is not
+ * available, the return value will be -EOPNOTSUPP.
+ */
+int ap_query_configuration(struct ap_config_info *info);
+
+/*
+ * struct ap_qirq_ctrl - convenient struct for easy invocation
+ * of the ap_queue_irq_ctrl() function. This struct is passed
+ * as GR1 parameter to the PQAP(AQIC) instruction. For details
+ * please see the AR documentation.
+ */
+struct ap_qirq_ctrl {
+	unsigned int _res1 : 8;
+	unsigned int zone  : 8;  /* zone info */
+	unsigned int ir    : 1;  /* ir flag: enable (1) or disable (0) irq */
+	unsigned int _res2 : 4;
+	unsigned int gisc  : 3;  /* guest isc field */
+	unsigned int _res3 : 6;
+	unsigned int gf    : 2;  /* gisa format */
+	unsigned int _res4 : 1;
+	unsigned int gisa  : 27; /* gisa origin */
+	unsigned int _res5 : 1;
+	unsigned int isc   : 3;  /* irq sub class */
+};
+
+/**
+ * ap_queue_irq_ctrl(): Control interruption on a AP queue.
+ * @qid: The AP queue number
+ * @qirqctrl: struct ap_qirq_ctrl, see above
+ * @ind: The notification indicator byte
+ *
+ * Returns AP queue status.
+ *
+ * Control interruption on the given AP queue.
+ * Just a simple wrapper function for the low level PQAP(AQIC)
+ * instruction available for other kernel modules.
+ */
+struct ap_queue_status ap_queue_irq_ctrl(ap_qid_t qid,
+					 struct ap_qirq_ctrl qirqctrl,
+					 void *ind);
+
+#endif /* _ASM_S390_AP_H_ */

arch/s390/include/asm/mmu.h

@@ -5,12 +5,11 @@
 #include <linux/errno.h>
 
 typedef struct {
+	spinlock_t lock;
 	cpumask_t cpu_attach_mask;
 	atomic_t flush_count;
 	unsigned int flush_mm;
-	spinlock_t pgtable_lock;
 	struct list_head pgtable_list;
-	spinlock_t gmap_lock;
 	struct list_head gmap_list;
 	unsigned long gmap_asce;
 	unsigned long asce;
@@ -27,10 +26,8 @@ typedef struct {
 } mm_context_t;
 
 #define INIT_MM_CONTEXT(name)						   \
-	.context.pgtable_lock =						   \
-			__SPIN_LOCK_UNLOCKED(name.context.pgtable_lock),   \
+	.context.lock =	__SPIN_LOCK_UNLOCKED(name.context.lock),	   \
 	.context.pgtable_list = LIST_HEAD_INIT(name.context.pgtable_list), \
-	.context.gmap_lock = __SPIN_LOCK_UNLOCKED(name.context.gmap_lock), \
 	.context.gmap_list = LIST_HEAD_INIT(name.context.gmap_list),
 
 static inline int tprot(unsigned long addr)

arch/s390/include/asm/mmu_context.h

@@ -17,9 +17,8 @@
 static inline int init_new_context(struct task_struct *tsk,
 				   struct mm_struct *mm)
 {
-	spin_lock_init(&mm->context.pgtable_lock);
+	spin_lock_init(&mm->context.lock);
 	INIT_LIST_HEAD(&mm->context.pgtable_list);
-	spin_lock_init(&mm->context.gmap_lock);
 	INIT_LIST_HEAD(&mm->context.gmap_list);
 	cpumask_clear(&mm->context.cpu_attach_mask);
 	atomic_set(&mm->context.flush_count, 0);
@@ -103,7 +102,6 @@ static inline void switch_mm(struct mm_struct *prev, struct mm_struct *next,
 	if (prev == next)
 		return;
 	cpumask_set_cpu(cpu, &next->context.cpu_attach_mask);
-	cpumask_set_cpu(cpu, mm_cpumask(next));
 	/* Clear old ASCE by loading the kernel ASCE. */
 	__ctl_load(S390_lowcore.kernel_asce, 1, 1);
 	__ctl_load(S390_lowcore.kernel_asce, 7, 7);
@@ -121,9 +119,8 @@ static inline void finish_arch_post_lock_switch(void)
 		preempt_disable();
 		while (atomic_read(&mm->context.flush_count))
 			cpu_relax();
-
-		if (mm->context.flush_mm)
-			__tlb_flush_mm(mm);
+		cpumask_set_cpu(smp_processor_id(), mm_cpumask(mm));
+		__tlb_flush_mm_lazy(mm);
 		preempt_enable();
 	}
 	set_fs(current->thread.mm_segment);
@@ -136,6 +133,7 @@ static inline void activate_mm(struct mm_struct *prev,
                                struct mm_struct *next)
 {
 	switch_mm(prev, next, current);
+	cpumask_set_cpu(smp_processor_id(), mm_cpumask(next));
 	set_user_asce(next);
 }
 

arch/s390/include/asm/tlbflush.h

@@ -48,23 +48,6 @@ static inline void __tlb_flush_global(void)
  * Flush TLB entries for a specific mm on all CPUs (in case gmap is used
  * this implicates multiple ASCEs!).
  */
-static inline void __tlb_flush_full(struct mm_struct *mm)
-{
-	preempt_disable();
-	atomic_inc(&mm->context.flush_count);
-	if (cpumask_equal(mm_cpumask(mm), cpumask_of(smp_processor_id()))) {
-		/* Local TLB flush */
-		__tlb_flush_local();
-	} else {
-		/* Global TLB flush */
-		__tlb_flush_global();
-		/* Reset TLB flush mask */
-		cpumask_copy(mm_cpumask(mm), &mm->context.cpu_attach_mask);
-	}
-	atomic_dec(&mm->context.flush_count);
-	preempt_enable();
-}
-
 static inline void __tlb_flush_mm(struct mm_struct *mm)
 {
 	unsigned long gmap_asce;
@@ -76,16 +59,18 @@ static inline void __tlb_flush_mm(struct mm_struct *mm)
 	 */
 	preempt_disable();
 	atomic_inc(&mm->context.flush_count);
+	/* Reset TLB flush mask */
+	cpumask_copy(mm_cpumask(mm), &mm->context.cpu_attach_mask);
+	barrier();
 	gmap_asce = READ_ONCE(mm->context.gmap_asce);
 	if (MACHINE_HAS_IDTE && gmap_asce != -1UL) {
 		if (gmap_asce)
 			__tlb_flush_idte(gmap_asce);
 		__tlb_flush_idte(mm->context.asce);
 	} else {
-		__tlb_flush_full(mm);
+		/* Global TLB flush */
+		__tlb_flush_global();
 	}
-	/* Reset TLB flush mask */
-	cpumask_copy(mm_cpumask(mm), &mm->context.cpu_attach_mask);
 	atomic_dec(&mm->context.flush_count);
 	preempt_enable();
 }
@@ -99,7 +84,6 @@ static inline void __tlb_flush_kernel(void)
 }
 #else
 #define __tlb_flush_global()	__tlb_flush_local()
-#define __tlb_flush_full(mm)	__tlb_flush_local()
 
 /*
  * Flush TLB entries for a specific ASCE on all CPUs.
@@ -117,10 +101,12 @@ static inline void __tlb_flush_kernel(void)
 
 static inline void __tlb_flush_mm_lazy(struct mm_struct * mm)
 {
+	spin_lock(&mm->context.lock);
 	if (mm->context.flush_mm) {
-		__tlb_flush_mm(mm);
 		mm->context.flush_mm = 0;
+		__tlb_flush_mm(mm);
 	}
+	spin_unlock(&mm->context.lock);
 }
 
 /*

arch/s390/mm/gmap.c

@@ -100,14 +100,14 @@ struct gmap *gmap_create(struct mm_struct *mm, unsigned long limit)
 	if (!gmap)
 		return NULL;
 	gmap->mm = mm;
-	spin_lock(&mm->context.gmap_lock);
+	spin_lock(&mm->context.lock);
 	list_add_rcu(&gmap->list, &mm->context.gmap_list);
 	if (list_is_singular(&mm->context.gmap_list))
 		gmap_asce = gmap->asce;
 	else
 		gmap_asce = -1UL;
 	WRITE_ONCE(mm->context.gmap_asce, gmap_asce);
-	spin_unlock(&mm->context.gmap_lock);
+	spin_unlock(&mm->context.lock);
 	return gmap;
 }
 EXPORT_SYMBOL_GPL(gmap_create);
@@ -248,7 +248,7 @@ void gmap_remove(struct gmap *gmap)
 		spin_unlock(&gmap->shadow_lock);
 	}
 	/* Remove gmap from the pre-mm list */
-	spin_lock(&gmap->mm->context.gmap_lock);
+	spin_lock(&gmap->mm->context.lock);
 	list_del_rcu(&gmap->list);
 	if (list_empty(&gmap->mm->context.gmap_list))
 		gmap_asce = 0;
@@ -258,7 +258,7 @@ void gmap_remove(struct gmap *gmap)
 	else
 		gmap_asce = -1UL;
 	WRITE_ONCE(gmap->mm->context.gmap_asce, gmap_asce);
-	spin_unlock(&gmap->mm->context.gmap_lock);
+	spin_unlock(&gmap->mm->context.lock);
 	synchronize_rcu();
 	/* Put reference */
 	gmap_put(gmap);

arch/s390/mm/pgalloc.c

@@ -83,7 +83,7 @@ int crst_table_upgrade(struct mm_struct *mm, unsigned long end)
 	int rc, notify;
 
 	/* upgrade should only happen from 3 to 4, 3 to 5, or 4 to 5 levels */
-	BUG_ON(mm->context.asce_limit < _REGION2_SIZE);
+	VM_BUG_ON(mm->context.asce_limit < _REGION2_SIZE);
 	if (end >= TASK_SIZE_MAX)
 		return -ENOMEM;
 	rc = 0;
@@ -124,7 +124,7 @@ void crst_table_downgrade(struct mm_struct *mm)
 	pgd_t *pgd;
 
 	/* downgrade should only happen from 3 to 2 levels (compat only) */
-	BUG_ON(mm->context.asce_limit != _REGION2_SIZE);
+	VM_BUG_ON(mm->context.asce_limit != _REGION2_SIZE);
 
 	if (current->active_mm == mm) {
 		clear_user_asce();
@@ -188,7 +188,7 @@ unsigned long *page_table_alloc(struct mm_struct *mm)
 	/* Try to get a fragment of a 4K page as a 2K page table */
 	if (!mm_alloc_pgste(mm)) {
 		table = NULL;
-		spin_lock_bh(&mm->context.pgtable_lock);
+		spin_lock_bh(&mm->context.lock);
 		if (!list_empty(&mm->context.pgtable_list)) {
 			page = list_first_entry(&mm->context.pgtable_list,
 						struct page, lru);
@@ -203,7 +203,7 @@ unsigned long *page_table_alloc(struct mm_struct *mm)
 				list_del(&page->lru);
 			}
 		}
-		spin_unlock_bh(&mm->context.pgtable_lock);
+		spin_unlock_bh(&mm->context.lock);
 		if (table)
 			return table;
 	}
@@ -227,9 +227,9 @@ unsigned long *page_table_alloc(struct mm_struct *mm)
 		/* Return the first 2K fragment of the page */
 		atomic_set(&page->_mapcount, 1);
 		clear_table(table, _PAGE_INVALID, PAGE_SIZE);
-		spin_lock_bh(&mm->context.pgtable_lock);
+		spin_lock_bh(&mm->context.lock);
 		list_add(&page->lru, &mm->context.pgtable_list);
-		spin_unlock_bh(&mm->context.pgtable_lock);
+		spin_unlock_bh(&mm->context.lock);
 	}
 	return table;
 }
@@ -243,13 +243,13 @@ void page_table_free(struct mm_struct *mm, unsigned long *table)
 	if (!mm_alloc_pgste(mm)) {
 		/* Free 2K page table fragment of a 4K page */
 		bit = (__pa(table) & ~PAGE_MASK)/(PTRS_PER_PTE*sizeof(pte_t));
-		spin_lock_bh(&mm->context.pgtable_lock);
+		spin_lock_bh(&mm->context.lock);
 		mask = atomic_xor_bits(&page->_mapcount, 1U << bit);
 		if (mask & 3)
 			list_add(&page->lru, &mm->context.pgtable_list);
 		else
 			list_del(&page->lru);
-		spin_unlock_bh(&mm->context.pgtable_lock);
+		spin_unlock_bh(&mm->context.lock);
 		if (mask != 0)
 			return;
 	}
@@ -275,13 +275,13 @@ void page_table_free_rcu(struct mmu_gather *tlb, unsigned long *table,
 		return;
 	}
 	bit = (__pa(table) & ~PAGE_MASK) / (PTRS_PER_PTE*sizeof(pte_t));
-	spin_lock_bh(&mm->context.pgtable_lock);
+	spin_lock_bh(&mm->context.lock);
 	mask = atomic_xor_bits(&page->_mapcount, 0x11U << bit);
 	if (mask & 3)
 		list_add_tail(&page->lru, &mm->context.pgtable_list);
 	else
 		list_del(&page->lru);
-	spin_unlock_bh(&mm->context.pgtable_lock);
+	spin_unlock_bh(&mm->context.lock);
 	table = (unsigned long *) (__pa(table) | (1U << bit));
 	tlb_remove_table(tlb, table);
 }