From: Richard Henderson <richard.henderson@linaro.org>
To: qemu-devel@nongnu.org
Subject: [PATCH v3 06/14] tcg/arm: Support unaligned access for softmmu
Date: Wed, 18 Aug 2021 11:29:04 -1000 [thread overview]
Message-ID: <20210818212912.396794-7-richard.henderson@linaro.org> (raw)
In-Reply-To: <20210818212912.396794-1-richard.henderson@linaro.org>
From armv6, the architecture supports unaligned accesses.
All we need to do is perform the correct alignment check
in tcg_out_tlb_read and not use LDRD/STRD when the access
is not aligned.
Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
---
tcg/arm/tcg-target.c.inc | 69 ++++++++++++++++++++++------------------
1 file changed, 38 insertions(+), 31 deletions(-)
diff --git a/tcg/arm/tcg-target.c.inc b/tcg/arm/tcg-target.c.inc
index 0c7e4f8411..c55167cc84 100644
--- a/tcg/arm/tcg-target.c.inc
+++ b/tcg/arm/tcg-target.c.inc
@@ -34,13 +34,6 @@ bool use_idiv_instructions;
bool use_neon_instructions;
#endif
-/* ??? Ought to think about changing CONFIG_SOFTMMU to always defined. */
-#ifdef CONFIG_SOFTMMU
-# define USING_SOFTMMU 1
-#else
-# define USING_SOFTMMU 0
-#endif
-
#ifdef CONFIG_DEBUG_TCG
static const char * const tcg_target_reg_names[TCG_TARGET_NB_REGS] = {
"%r0", "%r1", "%r2", "%r3", "%r4", "%r5", "%r6", "%r7",
@@ -1526,15 +1519,20 @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGReg addrlo, TCGReg addrhi,
int fast_off = TLB_MASK_TABLE_OFS(mem_index);
int mask_off = fast_off + offsetof(CPUTLBDescFast, mask);
int table_off = fast_off + offsetof(CPUTLBDescFast, table);
- unsigned s_bits = opc & MO_SIZE;
- unsigned a_bits = get_alignment_bits(opc);
+ unsigned s_mask = (1 << (opc & MO_SIZE)) - 1;
+ unsigned a_mask = (1 << get_alignment_bits(opc)) - 1;
+ TCGReg t_addr;
/*
- * We don't support inline unaligned acceses, but we can easily
- * support overalignment checks.
+ * For v7, support for unaligned accesses is mandatory.
+ * For v6, support for unaligned accesses is enabled by SCTLR.U,
+ * which is enabled by (at least) Linux and FreeBSD.
+ * For v4 and v5, unaligned accesses are... complicated, and
+ * unhelped by Linux having a global not per-process flag
+ * for unaligned handling.
*/
- if (a_bits < s_bits) {
- a_bits = s_bits;
+ if (!use_armv6_instructions && a_mask < s_mask) {
+ a_mask = s_mask;
}
/* Load env_tlb(env)->f[mmu_idx].{mask,table} into {r0,r1}. */
@@ -1578,27 +1576,32 @@ static TCGReg tcg_out_tlb_read(TCGContext *s, TCGReg addrlo, TCGReg addrhi,
/*
* Check alignment, check comparators.
- * Do this in no more than 3 insns. Use MOVW for v7, if possible,
+ * Do this in 2-4 insns. Use MOVW for v7, if possible,
* to reduce the number of sequential conditional instructions.
* Almost all guests have at least 4k pages, which means that we need
* to clear at least 9 bits even for an 8-byte memory, which means it
* isn't worth checking for an immediate operand for BIC.
*/
+ /* For unaligned accesses, test the page of the last byte. */
+ t_addr = addrlo;
+ if (a_mask < s_mask) {
+ t_addr = TCG_REG_R0;
+ tcg_out_dat_imm(s, COND_AL, ARITH_ADD, t_addr,
+ addrlo, s_mask - a_mask);
+ }
if (use_armv7_instructions && TARGET_PAGE_BITS <= 16) {
- tcg_target_ulong mask = ~(TARGET_PAGE_MASK | ((1 << a_bits) - 1));
-
- tcg_out_movi32(s, COND_AL, TCG_REG_TMP, mask);
+ tcg_out_movi32(s, COND_AL, TCG_REG_TMP, ~(TARGET_PAGE_MASK | a_mask));
tcg_out_dat_reg(s, COND_AL, ARITH_BIC, TCG_REG_TMP,
- addrlo, TCG_REG_TMP, 0);
+ t_addr, TCG_REG_TMP, 0);
tcg_out_dat_reg(s, COND_AL, ARITH_CMP, 0, TCG_REG_R2, TCG_REG_TMP, 0);
} else {
- if (a_bits) {
- tcg_out_dat_imm(s, COND_AL, ARITH_TST, 0, addrlo,
- (1 << a_bits) - 1);
+ if (a_mask) {
+ tcg_debug_assert(a_mask <= 0xff);
+ tcg_out_dat_imm(s, COND_AL, ARITH_TST, 0, addrlo, a_mask);
}
- tcg_out_dat_reg(s, COND_AL, ARITH_MOV, TCG_REG_TMP, 0, addrlo,
+ tcg_out_dat_reg(s, COND_AL, ARITH_MOV, TCG_REG_TMP, 0, t_addr,
SHIFT_IMM_LSR(TARGET_PAGE_BITS));
- tcg_out_dat_reg(s, (a_bits ? COND_EQ : COND_AL), ARITH_CMP,
+ tcg_out_dat_reg(s, (a_mask ? COND_EQ : COND_AL), ARITH_CMP,
0, TCG_REG_R2, TCG_REG_TMP,
SHIFT_IMM_LSL(TARGET_PAGE_BITS));
}
@@ -1763,8 +1766,9 @@ static inline void tcg_out_qemu_ld_index(TCGContext *s, MemOp opc,
tcg_out_ld32_r(s, COND_AL, datalo, addrlo, addend);
break;
case MO_Q:
- /* Avoid ldrd for user-only emulation, to handle unaligned. */
- if (USING_SOFTMMU && use_armv6_instructions
+ /* LDRD requires alignment; double-check that. */
+ if (use_armv6_instructions
+ && get_alignment_bits(opc) >= MO_64
&& (datalo & 1) == 0 && datahi == datalo + 1) {
tcg_out_ldrd_r(s, COND_AL, datalo, addrlo, addend);
} else if (datalo != addend) {
@@ -1806,8 +1810,9 @@ static inline void tcg_out_qemu_ld_direct(TCGContext *s, MemOp opc,
tcg_out_ld32_12(s, COND_AL, datalo, addrlo, 0);
break;
case MO_Q:
- /* Avoid ldrd for user-only emulation, to handle unaligned. */
- if (USING_SOFTMMU && use_armv6_instructions
+ /* LDRD requires alignment; double-check that. */
+ if (use_armv6_instructions
+ && get_alignment_bits(opc) >= MO_64
&& (datalo & 1) == 0 && datahi == datalo + 1) {
tcg_out_ldrd_8(s, COND_AL, datalo, addrlo, 0);
} else if (datalo == addrlo) {
@@ -1882,8 +1887,9 @@ static inline void tcg_out_qemu_st_index(TCGContext *s, int cond, MemOp opc,
tcg_out_st32_r(s, cond, datalo, addrlo, addend);
break;
case MO_64:
- /* Avoid strd for user-only emulation, to handle unaligned. */
- if (USING_SOFTMMU && use_armv6_instructions
+ /* STRD requires alignment; double-check that. */
+ if (use_armv6_instructions
+ && get_alignment_bits(opc) >= MO_64
&& (datalo & 1) == 0 && datahi == datalo + 1) {
tcg_out_strd_r(s, cond, datalo, addrlo, addend);
} else {
@@ -1914,8 +1920,9 @@ static inline void tcg_out_qemu_st_direct(TCGContext *s, MemOp opc,
tcg_out_st32_12(s, COND_AL, datalo, addrlo, 0);
break;
case MO_64:
- /* Avoid strd for user-only emulation, to handle unaligned. */
- if (USING_SOFTMMU && use_armv6_instructions
+ /* STRD requires alignment; double-check that. */
+ if (use_armv6_instructions
+ && get_alignment_bits(opc) >= MO_64
&& (datalo & 1) == 0 && datahi == datalo + 1) {
tcg_out_strd_8(s, COND_AL, datalo, addrlo, 0);
} else {
--
2.25.1
next prev parent reply other threads:[~2021-08-18 21:31 UTC|newest]
Thread overview: 33+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-08-18 21:28 [PATCH v3 00/14] tcg/arm: Unaligned access and other cleanup Richard Henderson
2021-08-18 21:28 ` [PATCH v3 01/14] tcg/arm: Remove fallback definition of __ARM_ARCH Richard Henderson
2021-08-20 10:38 ` Peter Maydell
2021-08-18 21:29 ` [PATCH v3 02/14] tcg/arm: Standardize on tcg_out_<branch>_{reg,imm} Richard Henderson
2021-08-18 21:58 ` Philippe Mathieu-Daudé
2021-08-20 10:39 ` [PATCH v3 02/14] tcg/arm: Standardize on tcg_out_<branch>_{reg, imm} Peter Maydell
2021-08-18 21:29 ` [PATCH v3 03/14] tcg/arm: Simplify use_armvt5_instructions Richard Henderson
2021-08-20 10:59 ` Peter Maydell
2021-08-18 21:29 ` [PATCH v3 04/14] tcg/arm: Support armv4t in tcg_out_goto and tcg_out_call Richard Henderson
2021-08-20 10:50 ` Peter Maydell
2021-08-18 21:29 ` [PATCH v3 05/14] tcg/arm: Examine QEMU_TCG_DEBUG environment variable Richard Henderson
2021-08-20 11:01 ` Peter Maydell
2021-08-18 21:29 ` Richard Henderson [this message]
2021-08-20 13:34 ` [PATCH v3 06/14] tcg/arm: Support unaligned access for softmmu Peter Maydell
2021-08-20 17:19 ` Richard Henderson
2021-08-18 21:29 ` [PATCH v3 07/14] tcg/arm: Split out tcg_out_ldstm Richard Henderson
2021-08-20 11:45 ` Peter Maydell
2021-08-18 21:29 ` [PATCH v3 08/14] tcg/arm: Simplify usage of encode_imm Richard Henderson
2021-08-20 11:50 ` Peter Maydell
2021-08-18 21:29 ` [PATCH v3 09/14] tcg/arm: Drop inline markers Richard Henderson
2021-08-18 22:02 ` Philippe Mathieu-Daudé
2021-08-18 21:29 ` [PATCH v3 10/14] tcg/arm: Give enum arm_cond_code_e a typedef and use it Richard Henderson
2021-08-18 22:04 ` Philippe Mathieu-Daudé
2021-08-18 21:29 ` [PATCH v3 11/14] tcg/arm: More use of the ARMInsn enum Richard Henderson
2021-08-18 22:04 ` Philippe Mathieu-Daudé
2021-08-18 21:29 ` [PATCH v3 12/14] tcg/arm: More use of the TCGReg enum Richard Henderson
2021-08-18 22:05 ` Philippe Mathieu-Daudé
2021-08-18 21:29 ` [PATCH v3 13/14] tcg/arm: Reserve a register for guest_base Richard Henderson
2021-08-20 12:03 ` Peter Maydell
2021-08-20 18:47 ` Richard Henderson
2021-08-21 10:38 ` Peter Maydell
2021-08-18 21:29 ` [PATCH v3 14/14] tcg/arm: Support raising sigbus for user-only Richard Henderson
2021-08-20 13:56 ` Peter Maydell
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20210818212912.396794-7-richard.henderson@linaro.org \
--to=richard.henderson@linaro.org \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).