qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
From: "Alex Bennée" <alex.bennee@linaro.org>
To: Richard Henderson <richard.henderson@linaro.org>
Cc: pbonzini@redhat.com, qemu-devel@nongnu.org, stefanha@redhat.com,
	david@redhat.com
Subject: Re: [PATCH v4 06/16] cputlb: Introduce TLB_BSWAP
Date: Tue, 24 Sep 2019 19:25:26 +0100	[thread overview]
Message-ID: <87zhita6w9.fsf@linaro.org> (raw)
In-Reply-To: <20190923230004.9231-7-richard.henderson@linaro.org>


Richard Henderson <richard.henderson@linaro.org> writes:

> Handle bswap on ram directly in load/store_helper.  This fixes a
> bug with the previous implementation in that one cannot use the
> I/O path for RAM.
>
> Fixes: a26fc6f5152b47f1
> Reviewed-by: David Hildenbrand <david@redhat.com>
> Signed-off-by: Richard Henderson <richard.henderson@linaro.org>
> ---
>  include/exec/cpu-all.h |  4 ++-
>  accel/tcg/cputlb.c     | 62 ++++++++++++++++++++++--------------------
>  2 files changed, 36 insertions(+), 30 deletions(-)
>
> diff --git a/include/exec/cpu-all.h b/include/exec/cpu-all.h
> index e0c8dc540c..d148bded35 100644
> --- a/include/exec/cpu-all.h
> +++ b/include/exec/cpu-all.h
> @@ -335,12 +335,14 @@ CPUArchState *cpu_copy(CPUArchState *env);
>  #define TLB_MMIO            (1 << (TARGET_PAGE_BITS_MIN - 3))
>  /* Set if TLB entry contains a watchpoint.  */
>  #define TLB_WATCHPOINT      (1 << (TARGET_PAGE_BITS_MIN - 4))
> +/* Set if TLB entry requires byte swap.  */
> +#define TLB_BSWAP           (1 << (TARGET_PAGE_BITS_MIN - 5))
>
>  /* Use this mask to check interception with an alignment mask
>   * in a TCG backend.
>   */
>  #define TLB_FLAGS_MASK \
> -    (TLB_INVALID_MASK | TLB_NOTDIRTY | TLB_MMIO | TLB_WATCHPOINT)
> +    (TLB_INVALID_MASK | TLB_NOTDIRTY | TLB_MMIO | TLB_WATCHPOINT | TLB_BSWAP)
>
>  /**
>   * tlb_hit_page: return true if page aligned @addr is a hit against the
> diff --git a/accel/tcg/cputlb.c b/accel/tcg/cputlb.c
> index 430ba4a69d..f634edb4f4 100644
> --- a/accel/tcg/cputlb.c
> +++ b/accel/tcg/cputlb.c
> @@ -737,8 +737,7 @@ void tlb_set_page_with_attrs(CPUState *cpu, target_ulong vaddr,
>          address |= TLB_INVALID_MASK;
>      }
>      if (attrs.byte_swap) {
> -        /* Force the access through the I/O slow path.  */
> -        address |= TLB_MMIO;
> +        address |= TLB_BSWAP;
>      }
>      if (!memory_region_is_ram(section->mr) &&
>          !memory_region_is_romd(section->mr)) {
> @@ -901,10 +900,6 @@ static uint64_t io_readx(CPUArchState *env, CPUIOTLBEntry *iotlbentry,
>      bool locked = false;
>      MemTxResult r;
>
> -    if (iotlbentry->attrs.byte_swap) {
> -        op ^= MO_BSWAP;
> -    }
> -
>      section = iotlb_to_section(cpu, iotlbentry->addr, iotlbentry->attrs);
>      mr = section->mr;
>      mr_offset = (iotlbentry->addr & TARGET_PAGE_MASK) + addr;
> @@ -947,10 +942,6 @@ static void io_writex(CPUArchState *env, CPUIOTLBEntry *iotlbentry,
>      bool locked = false;
>      MemTxResult r;
>
> -    if (iotlbentry->attrs.byte_swap) {
> -        op ^= MO_BSWAP;
> -    }
> -
>      section = iotlb_to_section(cpu, iotlbentry->addr, iotlbentry->attrs);
>      mr = section->mr;
>      mr_offset = (iotlbentry->addr & TARGET_PAGE_MASK) + addr;
> @@ -1133,8 +1124,8 @@ void *probe_access(CPUArchState *env, target_ulong addr, int size,
>                               wp_access, retaddr);
>      }
>
> -    if (tlb_addr & (TLB_NOTDIRTY | TLB_MMIO)) {
> -        /* I/O access */
> +    /* Reject I/O access, or other required slow-path.  */
> +    if (tlb_addr & (TLB_NOTDIRTY | TLB_MMIO | TLB_BSWAP)) {
>          return NULL;
>      }
>
> @@ -1344,6 +1335,7 @@ load_helper(CPUArchState *env, target_ulong addr, TCGMemOpIdx oi,
>      /* Handle anything that isn't just a straight memory access.  */
>      if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
>          CPUIOTLBEntry *iotlbentry;
> +        bool need_swap;
>
>          /* For anything that is unaligned, recurse through full_load.  */
>          if ((addr & (size - 1)) != 0) {
> @@ -1357,17 +1349,22 @@ load_helper(CPUArchState *env, target_ulong addr, TCGMemOpIdx oi,
>              /* On watchpoint hit, this will longjmp out.  */
>              cpu_check_watchpoint(env_cpu(env), addr, size,
>                                   iotlbentry->attrs, BP_MEM_READ, retaddr);
> -
> -            /* The backing page may or may not require I/O.  */
> -            tlb_addr &= ~TLB_WATCHPOINT;
> -            if ((tlb_addr & ~TARGET_PAGE_MASK) == 0) {
> -                goto do_aligned_access;
> -            }
>          }
>
           /* We don't apply MO_BSWAP to op here because we want to
            * ensure the compiler can always unfold and dead-code away
            * the final load_memop in the fast path. If you try the
            * you will find the assert will get you ;-)
            */

?

Otherwise:

Reviewed-by: Alex Bennée <alex.bennee@linaro.org>

> +        need_swap = size > 1 && (tlb_addr & TLB_BSWAP);
> +
>          /* Handle I/O access.  */
> -        return io_readx(env, iotlbentry, mmu_idx, addr,
> -                        retaddr, access_type, op);
> +        if (likely(tlb_addr & TLB_MMIO)) {
> +            return io_readx(env, iotlbentry, mmu_idx, addr, retaddr,
> +                            access_type, op ^ (need_swap * MO_BSWAP));
> +        }
> +
> +        haddr = (void *)((uintptr_t)addr + entry->addend);
> +
> +        if (unlikely(need_swap)) {
> +            return load_memop(haddr, op ^ MO_BSWAP);
> +        }
> +        return load_memop(haddr, op);
>      }
>
>      /* Handle slow unaligned access (it spans two pages or IO).  */
> @@ -1394,7 +1391,6 @@ load_helper(CPUArchState *env, target_ulong addr, TCGMemOpIdx oi,
>          return res & MAKE_64BIT_MASK(0, size * 8);
>      }
>
> - do_aligned_access:
>      haddr = (void *)((uintptr_t)addr + entry->addend);
>      return load_memop(haddr, op);
>  }
> @@ -1592,6 +1588,7 @@ store_helper(CPUArchState *env, target_ulong addr, uint64_t val,
>      /* Handle anything that isn't just a straight memory access.  */
>      if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
>          CPUIOTLBEntry *iotlbentry;
> +        bool need_swap;
>
>          /* For anything that is unaligned, recurse through byte stores.  */
>          if ((addr & (size - 1)) != 0) {
> @@ -1605,16 +1602,24 @@ store_helper(CPUArchState *env, target_ulong addr, uint64_t val,
>              /* On watchpoint hit, this will longjmp out.  */
>              cpu_check_watchpoint(env_cpu(env), addr, size,
>                                   iotlbentry->attrs, BP_MEM_WRITE, retaddr);
> -
> -            /* The backing page may or may not require I/O.  */
> -            tlb_addr &= ~TLB_WATCHPOINT;
> -            if ((tlb_addr & ~TARGET_PAGE_MASK) == 0) {
> -                goto do_aligned_access;
> -            }
>          }
>
> +        need_swap = size > 1 && (tlb_addr & TLB_BSWAP);
> +
>          /* Handle I/O access.  */
> -        io_writex(env, iotlbentry, mmu_idx, val, addr, retaddr, op);
> +        if (likely(tlb_addr & (TLB_MMIO | TLB_NOTDIRTY))) {
> +            io_writex(env, iotlbentry, mmu_idx, val, addr, retaddr,
> +                      op ^ (need_swap * MO_BSWAP));
> +            return;
> +        }
> +
> +        haddr = (void *)((uintptr_t)addr + entry->addend);
> +
> +        if (unlikely(need_swap)) {
> +            store_memop(haddr, val, op ^ MO_BSWAP);
> +        } else {
> +            store_memop(haddr, val, op);
> +        }
>          return;
>      }
>
> @@ -1683,7 +1688,6 @@ store_helper(CPUArchState *env, target_ulong addr, uint64_t val,
>          return;
>      }
>
> - do_aligned_access:
>      haddr = (void *)((uintptr_t)addr + entry->addend);
>      store_memop(haddr, val, op);
>  }


--
Alex Bennée


  reply	other threads:[~2019-09-24 18:27 UTC|newest]

Thread overview: 49+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-09-23 22:59 [PATCH v4 00/16] Move rom and notdirty handling to cputlb Richard Henderson
2019-09-23 22:59 ` [PATCH v4 01/16] exec: Use TARGET_PAGE_BITS_MIN for TLB flags Richard Henderson
2019-09-24 13:53   ` Alex Bennée
2019-09-23 22:59 ` [PATCH v4 02/16] cputlb: Disable __always_inline__ without optimization Richard Henderson
2019-09-24 13:56   ` Alex Bennée
2019-09-23 22:59 ` [PATCH v4 03/16] qemu/compiler.h: Add optimize_away Richard Henderson
2019-09-24  7:47   ` David Hildenbrand
2019-09-24 17:27     ` Richard Henderson
2019-09-24 17:29       ` David Hildenbrand
2019-09-24 15:47   ` Alex Bennée
2019-09-23 22:59 ` [PATCH v4 04/16] cputlb: Use optimize_away in load/store_helpers Richard Henderson
2019-09-24  7:47   ` David Hildenbrand
2019-09-24 15:47   ` Alex Bennée
2019-09-23 22:59 ` [PATCH v4 05/16] cputlb: Split out load/store_memop Richard Henderson
2019-09-24  7:48   ` David Hildenbrand
2019-09-24 15:51   ` Alex Bennée
2019-09-23 22:59 ` [PATCH v4 06/16] cputlb: Introduce TLB_BSWAP Richard Henderson
2019-09-24 18:25   ` Alex Bennée [this message]
2019-09-25 17:36     ` Richard Henderson
2019-09-23 22:59 ` [PATCH v4 07/16] exec: Adjust notdirty tracing Richard Henderson
2019-09-24 21:53   ` Alex Bennée
2019-09-23 22:59 ` [PATCH v4 08/16] cputlb: Move ROM handling from I/O path to TLB path Richard Henderson
2019-09-25  0:16   ` Alex Bennée
2019-09-25  6:59     ` David Hildenbrand
2019-09-25 16:01       ` Alex Bennée
2019-09-25 17:01       ` Richard Henderson
2019-09-23 22:59 ` [PATCH v4 09/16] cputlb: Move NOTDIRTY " Richard Henderson
2019-09-25 16:06   ` Alex Bennée
2019-09-23 22:59 ` [PATCH v4 10/16] cputlb: Partially inline memory_region_section_get_iotlb Richard Henderson
2019-09-24  7:59   ` David Hildenbrand
2019-09-25 17:55     ` Richard Henderson
2019-09-25 19:40       ` David Hildenbrand
2019-09-25 16:12   ` Alex Bennée
2019-09-23 22:59 ` [PATCH v4 11/16] cputlb: Merge and move memory_notdirty_write_{prepare, complete} Richard Henderson
2019-09-24  8:04   ` [PATCH v4 11/16] cputlb: Merge and move memory_notdirty_write_{prepare,complete} David Hildenbrand
2019-09-25 16:15   ` Alex Bennée
2019-09-23 23:00 ` [PATCH v4 12/16] cputlb: Handle TLB_NOTDIRTY in probe_access Richard Henderson
2019-09-24  8:05   ` David Hildenbrand
2019-09-25 16:21   ` Alex Bennée
2019-09-23 23:00 ` [PATCH v4 13/16] cputlb: Remove cpu->mem_io_vaddr Richard Henderson
2019-09-25 16:22   ` Alex Bennée
2019-09-23 23:00 ` [PATCH v4 14/16] cputlb: Remove tb_invalidate_phys_page_range is_cpu_write_access Richard Henderson
2019-09-25 16:23   ` Alex Bennée
2019-09-23 23:00 ` [PATCH v4 15/16] cputlb: Pass retaddr to tb_invalidate_phys_page_fast Richard Henderson
2019-09-25 16:28   ` Alex Bennée
2019-09-23 23:00 ` [PATCH v4 16/16] cputlb: Pass retaddr to tb_check_watchpoint Richard Henderson
2019-09-25 16:30   ` Alex Bennée
2019-09-25 18:52 ` [PATCH v4 00/16] Move rom and notdirty handling to cputlb Mark Cave-Ayland
2019-09-25 18:54   ` Mark Cave-Ayland

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=87zhita6w9.fsf@linaro.org \
    --to=alex.bennee@linaro.org \
    --cc=david@redhat.com \
    --cc=pbonzini@redhat.com \
    --cc=qemu-devel@nongnu.org \
    --cc=richard.henderson@linaro.org \
    --cc=stefanha@redhat.com \
    /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).