KVM: Don't pre-reserve xarray entries when storing empty/NULL attributesgrafted

Skip the xarray reservation loop when clearing all memory attributes, as
storing NULL only erases the entry and never needs to allocate, so no
reservation (and no cleanup of a failed one) is required in that case.

Suggested-by: Sean Christopherson <seanjc@google.com>
Cc: David Ballesteros <davimaba.v@proton.me>
Signed-off-by: Zeng Chi <zengchi@kylinos.cn>
Link: https://patch.msgid.link/20260921102442.1232375-1-zeng_chi911@163.com
[sean: split to separate patch]
Signed-off-by: Sean Christopherson <seanjc@google.com>

1 files changed, 116 insertions, 0 deletions

diff --git a/arch/riscv/lib/delay.c b/arch/riscv/lib/delay.c
new file mode 100644
index 000000000..bd9b38256
--- /dev/null
+++ b/arch/riscv/lib/delay.c
@@ -0,0 +1,116 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Copyright (C) 2012 Regents of the University of California
+ */
+
+#include <linux/delay.h>
+#include <linux/math.h>
+#include <linux/param.h>
+#include <linux/types.h>
+#include <linux/export.h>
+
+#include <asm/processor.h>
+
+/*
+ * This is copies from arch/arm/include/asm/delay.h
+ *
+ * Loop (or tick) based delay:
+ *
+ * loops = loops_per_jiffy * jiffies_per_sec * delay_us / us_per_sec
+ *
+ * where:
+ *
+ * jiffies_per_sec = HZ
+ * us_per_sec = 1000000
+ *
+ * Therefore the constant part is HZ / 1000000 which is a small
+ * fractional number. To make this usable with integer math, we
+ * scale up this constant by 2^31, perform the actual multiplication,
+ * and scale the result back down by 2^31 with a simple shift:
+ *
+ * loops = (loops_per_jiffy * delay_us * UDELAY_MULT) >> 31
+ *
+ * where:
+ *
+ * UDELAY_MULT = 2^31 * HZ / 1000000
+ *             = (2^31 / 1000000) * HZ
+ *             = 2147.483648 * HZ
+ *             = 2147 * HZ + 483648 * HZ / 1000000
+ *
+ * 31 is the biggest scale shift value that won't overflow 32 bits for
+ * delay_us * UDELAY_MULT assuming HZ <= 1000 and delay_us <= 2000.
+ */
+#define MAX_UDELAY_US	2000
+#define MAX_UDELAY_HZ	1000
+#define UDELAY_MULT	(2147UL * HZ + 483648UL * HZ / 1000000UL)
+#define UDELAY_SHIFT	31
+
+#if HZ > MAX_UDELAY_HZ
+#error "HZ > MAX_UDELAY_HZ"
+#endif
+
+/*
+ * RISC-V supports both UDELAY and NDELAY.  This is largely the same as above,
+ * but with different constants.  I added 10 bits to the shift to get this, but
+ * the result is that I need a 64-bit multiply, which is slow on 32-bit
+ * platforms.
+ *
+ * NDELAY_MULT = 2^41 * HZ / 1000000000
+ *             = (2^41 / 1000000000) * HZ
+ *             = 2199.02325555 * HZ
+ *             = 2199 * HZ + 23255550 * HZ / 1000000000
+ *
+ * The maximum here is to avoid 64-bit overflow, but it isn't checked as it
+ * won't happen.
+ */
+#define MAX_NDELAY_NS   (1ULL << 42)
+#define MAX_NDELAY_HZ	MAX_UDELAY_HZ
+#define NDELAY_MULT	((unsigned long long)(2199ULL * HZ + 23255550ULL * HZ / 1000000000ULL))
+#define NDELAY_SHIFT	41
+
+#if HZ > MAX_NDELAY_HZ
+#error "HZ > MAX_NDELAY_HZ"
+#endif
+
+void __delay(unsigned long cycles)
+{
+	u64 t0 = get_cycles();
+
+	while ((unsigned long)(get_cycles() - t0) < cycles)
+		cpu_relax();
+}
+EXPORT_SYMBOL(__delay);
+
+void udelay(unsigned long usecs)
+{
+	u64 ucycles = (u64)usecs * lpj_fine * UDELAY_MULT;
+	u64 n;
+
+	if (unlikely(usecs > MAX_UDELAY_US)) {
+		n = (u64)usecs * riscv_timebase;
+		do_div(n, 1000000);
+
+		__delay(n);
+		return;
+	}
+
+	__delay(ucycles >> UDELAY_SHIFT);
+}
+EXPORT_SYMBOL(udelay);
+
+void ndelay(unsigned long nsecs)
+{
+	/*
+	 * This doesn't bother checking for overflow, as it won't happen (it's
+	 * an hour) of delay.
+	 */
+	unsigned long long ncycles = nsecs * lpj_fine * NDELAY_MULT;
+	__delay(ncycles >> NDELAY_SHIFT);
+}
+EXPORT_SYMBOL(ndelay);
+
+bool delay_read_timer(unsigned long *timer_val)
+{
+	*timer_val = get_cycles();
+	return true;
+}