cxl: Poll for outstanding IRQs when detaching a context

When detaching contexts, we may still have interrupts in the system
which are yet to be delivered to any CPU and be acked in the PSL.
This can result in a subsequent unrelated process getting an spurious
IRQ or an interrupt for a non-existent context.

This polls the PSL to ensure that the PSL is clear of IRQs for the
detached context, before removing the context from the idr.

Signed-off-by: Michael Neuling <mikey@neuling.org>
Tested-by: Andrew Donnellan <andrew.donnellan@au1.ibm.com>
Acked-by: Ian Munsie <imunsie@au1.ibm.com>
Tested-by: Vaibhav Jain <vaibhav@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>

drivers/misc/cxl/context.c

@@ -223,6 +223,13 @@ int __detach_context(struct cxl_context *ctx)
 		cxl_ops->link_ok(ctx->afu->adapter, ctx->afu));
 	flush_work(&ctx->fault_work); /* Only needed for dedicated process */
 
+	/*
+	 * Wait until no further interrupts are presented by the PSL
+	 * for this context.
+	 */
+	if (cxl_ops->irq_wait)
+		cxl_ops->irq_wait(ctx);
+
 	/* release the reference to the group leader and mm handling pid */
 	put_pid(ctx->pid);
 	put_pid(ctx->glpid);

drivers/misc/cxl/cxl.h

@@ -274,6 +274,7 @@ static const cxl_p2n_reg_t CXL_PSL_WED_An     = {0x0A0};
 #define CXL_PSL_DSISR_An_PE (1ull << (63-4))  /* PSL Error (implementation specific) */
 #define CXL_PSL_DSISR_An_AE (1ull << (63-5))  /* AFU Error */
 #define CXL_PSL_DSISR_An_OC (1ull << (63-6))  /* OS Context Warning */
+#define CXL_PSL_DSISR_PENDING (CXL_PSL_DSISR_TRANS | CXL_PSL_DSISR_An_PE | CXL_PSL_DSISR_An_AE | CXL_PSL_DSISR_An_OC)
 /* NOTE: Bits 32:63 are undefined if DSISR[DS] = 1 */
 #define CXL_PSL_DSISR_An_M  DSISR_NOHPTE      /* PTE not found */
 #define CXL_PSL_DSISR_An_P  DSISR_PROTFAULT   /* Storage protection violation */
@@ -855,6 +856,7 @@ struct cxl_backend_ops {
 					u64 dsisr, u64 errstat);
 	irqreturn_t (*psl_interrupt)(int irq, void *data);
 	int (*ack_irq)(struct cxl_context *ctx, u64 tfc, u64 psl_reset_mask);
+	void (*irq_wait)(struct cxl_context *ctx);
 	int (*attach_process)(struct cxl_context *ctx, bool kernel,
 			u64 wed, u64 amr);
 	int (*detach_process)(struct cxl_context *ctx);

drivers/misc/cxl/native.c

@@ -14,6 +14,7 @@
 #include <linux/mutex.h>
 #include <linux/mm.h>
 #include <linux/uaccess.h>
+#include <linux/delay.h>
 #include <asm/synch.h>
 #include <misc/cxl-base.h>
 
@@ -797,6 +798,35 @@ static irqreturn_t native_irq_multiplexed(int irq, void *data)
 	return fail_psl_irq(afu, &irq_info);
 }
 
+void native_irq_wait(struct cxl_context *ctx)
+{
+	u64 dsisr;
+	int timeout = 1000;
+	int ph;
+
+	/*
+	 * Wait until no further interrupts are presented by the PSL
+	 * for this context.
+	 */
+	while (timeout--) {
+		ph = cxl_p2n_read(ctx->afu, CXL_PSL_PEHandle_An) & 0xffff;
+		if (ph != ctx->pe)
+			return;
+		dsisr = cxl_p2n_read(ctx->afu, CXL_PSL_DSISR_An);
+		if ((dsisr & CXL_PSL_DSISR_PENDING) == 0)
+			return;
+		/*
+		 * We are waiting for the workqueue to process our
+		 * irq, so need to let that run here.
+		 */
+		msleep(1);
+	}
+
+	dev_warn(&ctx->afu->dev, "WARNING: waiting on DSI for PE %i"
+		 " DSISR %016llx!\n", ph, dsisr);
+	return;
+}
+
 static irqreturn_t native_slice_irq_err(int irq, void *data)
 {
 	struct cxl_afu *afu = data;
@@ -1076,6 +1106,7 @@ const struct cxl_backend_ops cxl_native_ops = {
 	.handle_psl_slice_error = native_handle_psl_slice_error,
 	.psl_interrupt = NULL,
 	.ack_irq = native_ack_irq,
+	.irq_wait = native_irq_wait,
 	.attach_process = native_attach_process,
 	.detach_process = native_detach_process,
 	.support_attributes = native_support_attributes,