Re: [Qemu-devel] [RFC v6 02/14] softmmu: Add new TLB_EXCL flag

[Alex Bennée]

alvise rigo <address@hidden> writes:

> On Tue, Jan 5, 2016 at 5:10 PM, Alex Bennée <address@hidden> wrote:
>>
>> Alvise Rigo <address@hidden> writes:
>>
>>> Add a new TLB flag to force all the accesses made to a page to follow
>>> the slow-path.
>>>
>>> In the case we remove a TLB entry marked as EXCL, we unset the
>>> corresponding exclusive bit in the bitmap.
>>>
>>> Suggested-by: Jani Kokkonen <address@hidden>
>>> Suggested-by: Claudio Fontana <address@hidden>
>>> Signed-off-by: Alvise Rigo <address@hidden>
>>> ---
>>>  cputlb.c                |  38 +++++++++++++++-
>>>  include/exec/cpu-all.h  |   8 ++++
>>>  include/exec/cpu-defs.h |   1 +
>>>  include/qom/cpu.h       |  14 ++++++
>>>  softmmu_template.h      | 114 
>>> ++++++++++++++++++++++++++++++++++++++----------
>>>  5 files changed, 152 insertions(+), 23 deletions(-)
>>>
>>> diff --git a/cputlb.c b/cputlb.c
>>> index bf1d50a..7ee0c89 100644
>>> --- a/cputlb.c
>>> +++ b/cputlb.c
>>> @@ -394,6 +394,16 @@ void tlb_set_page_with_attrs(CPUState *cpu, 
>>> target_ulong vaddr,
>>>      env->tlb_v_table[mmu_idx][vidx] = *te;
>>>      env->iotlb_v[mmu_idx][vidx] = env->iotlb[mmu_idx][index];
>>>
>>> +    if (unlikely(!(te->addr_write & TLB_MMIO) && (te->addr_write &
>>> TLB_EXCL))) {
>>
>> Why do we care about TLB_MMIO flags here? Does it actually happen? Would
>> bad things happen if we enforced exclusivity for an MMIO write? Do the
>> other flags matter?
>
> In the previous version of the patch series it came out that the
> accesses to MMIO regions have to be supported since, for instance, the
> GDB stub relies on them.
> The last two patches actually finalize the MMIO support.
>
>>
>> There should be a comment as to why MMIO is mentioned I think.
>
> OK.
>
>>
>>> +        /* We are removing an exclusive entry, set the page to dirty. This
>>> +         * is not be necessary if the vCPU has performed both SC and LL. */
>>> +        hwaddr hw_addr = (env->iotlb[mmu_idx][index].addr & 
>>> TARGET_PAGE_MASK) +
>>> +                                          (te->addr_write & 
>>> TARGET_PAGE_MASK);
>>> +        if (!cpu->ll_sc_context) {
>>> +            cpu_physical_memory_unset_excl(hw_addr, cpu->cpu_index);
>>> +        }
>>> +    }
>>> +
>>>      /* refill the tlb */
>>>      env->iotlb[mmu_idx][index].addr = iotlb - vaddr;
>>>      env->iotlb[mmu_idx][index].attrs = attrs;
>>> @@ -419,7 +429,15 @@ void tlb_set_page_with_attrs(CPUState *cpu, 
>>> target_ulong vaddr,
>>>                                                     + xlat)) {
>>>              te->addr_write = address | TLB_NOTDIRTY;
>>>          } else {
>>> -            te->addr_write = address;
>>> +            if (!(address & TLB_MMIO) &&
>>> +                cpu_physical_memory_atleast_one_excl(section->mr->ram_addr
>>> +                                                           + xlat)) {
>>> +                /* There is at least one vCPU that has flagged the address 
>>> as
>>> +                 * exclusive. */
>>> +                te->addr_write = address | TLB_EXCL;
>>> +            } else {
>>> +                te->addr_write = address;
>>> +            }
>>>          }
>>>      } else {
>>>          te->addr_write = -1;
>>> @@ -471,6 +489,24 @@ tb_page_addr_t get_page_addr_code(CPUArchState *env1, 
>>> target_ulong addr)
>>>      return qemu_ram_addr_from_host_nofail(p);
>>>  }
>>>
>>> +/* For every vCPU compare the exclusive address and reset it in case of a
>>> + * match. Since only one vCPU is running at once, no lock has to be held to
>>> + * guard this operation. */
>>> +static inline void lookup_and_reset_cpus_ll_addr(hwaddr addr, hwaddr size)
>>> +{
>>> +    CPUState *cpu;
>>> +
>>> +    CPU_FOREACH(cpu) {
>>> +        if (cpu->excl_protected_range.begin != EXCLUSIVE_RESET_ADDR &&
>>> +            ranges_overlap(cpu->excl_protected_range.begin,
>>> +                           cpu->excl_protected_range.end -
>>> +                           cpu->excl_protected_range.begin,
>>> +                           addr, size)) {
>>> +            cpu->excl_protected_range.begin = EXCLUSIVE_RESET_ADDR;
>>> +        }
>>> +    }
>>> +}
>>> +
>>>  #define MMUSUFFIX _mmu
>>>
>>>  #define SHIFT 0
>>> diff --git a/include/exec/cpu-all.h b/include/exec/cpu-all.h
>>> index 83b1781..f8d8feb 100644
>>> --- a/include/exec/cpu-all.h
>>> +++ b/include/exec/cpu-all.h
>>> @@ -277,6 +277,14 @@ CPUArchState *cpu_copy(CPUArchState *env);
>>>  #define TLB_NOTDIRTY    (1 << 4)
>>>  /* Set if TLB entry is an IO callback.  */
>>>  #define TLB_MMIO        (1 << 5)
>>> +/* Set if TLB entry references a page that requires exclusive access.  */
>>> +#define TLB_EXCL        (1 << 6)
>>> +
>>> +/* Do not allow a TARGET_PAGE_MASK which covers one or more bits defined
>>> + * above. */
>>> +#if TLB_EXCL >= TARGET_PAGE_SIZE
>>> +#error TARGET_PAGE_MASK covering the low bits of the TLB virtual address
>>> +#endif
>>>
>>>  void dump_exec_info(FILE *f, fprintf_function cpu_fprintf);
>>>  void dump_opcount_info(FILE *f, fprintf_function cpu_fprintf);
>>> diff --git a/include/exec/cpu-defs.h b/include/exec/cpu-defs.h
>>> index 5093be2..b34d7ae 100644
>>> --- a/include/exec/cpu-defs.h
>>> +++ b/include/exec/cpu-defs.h
>>> @@ -27,6 +27,7 @@
>>>  #include <inttypes.h>
>>>  #include "qemu/osdep.h"
>>>  #include "qemu/queue.h"
>>> +#include "qemu/range.h"
>>>  #include "tcg-target.h"
>>>  #ifndef CONFIG_USER_ONLY
>>>  #include "exec/hwaddr.h"
>>> diff --git a/include/qom/cpu.h b/include/qom/cpu.h
>>> index 51a1323..c6bb6b6 100644
>>> --- a/include/qom/cpu.h
>>> +++ b/include/qom/cpu.h
>>> @@ -29,6 +29,7 @@
>>>  #include "qemu/queue.h"
>>>  #include "qemu/thread.h"
>>>  #include "qemu/typedefs.h"
>>> +#include "qemu/range.h"
>>>
>>>  typedef int (*WriteCoreDumpFunction)(const void *buf, size_t size,
>>>                                       void *opaque);
>>> @@ -210,6 +211,9 @@ struct kvm_run;
>>>  #define TB_JMP_CACHE_BITS 12
>>>  #define TB_JMP_CACHE_SIZE (1 << TB_JMP_CACHE_BITS)
>>>
>>> +/* Atomic insn translation TLB support. */
>>> +#define EXCLUSIVE_RESET_ADDR ULLONG_MAX
>>> +
>>>  /**
>>>   * CPUState:
>>>   * @cpu_index: CPU index (informative).
>>> @@ -329,6 +333,16 @@ struct CPUState {
>>>       */
>>>      bool throttle_thread_scheduled;
>>>
>>> +    /* Used by the atomic insn translation backend. */
>>> +    int ll_sc_context;
>>> +    /* vCPU current exclusive addresses range.
>>> +     * The address is set to EXCLUSIVE_RESET_ADDR if the vCPU is not.
>>> +     * in the middle of a LL/SC. */
>>> +    struct Range excl_protected_range;
>>> +    /* Used to carry the SC result but also to flag a normal (legacy)
>>> +     * store access made by a stcond (see softmmu_template.h). */
>>> +    int excl_succeeded;
>>
>> It might be clearer if excl_succeeded was defined as a bool?
>
> Yes, that might be a good idea.
>
>>
>>>      /* Note that this is accessed at the start of every TB via a negative
>>>         offset from AREG0.  Leave this field at the end so as to make the
>>>         (absolute value) offset as small as possible.  This reduces code
>>> diff --git a/softmmu_template.h b/softmmu_template.h
>>> index 6803890..24d29b7 100644
>>> --- a/softmmu_template.h
>>> +++ b/softmmu_template.h
>>> @@ -395,19 +395,54 @@ void helper_le_st_name(CPUArchState *env, 
>>> target_ulong addr, DATA_TYPE val,
>>>          tlb_addr = env->tlb_table[mmu_idx][index].addr_write;
>>>      }
>>>
>>> -    /* Handle an IO access.  */
>>> +    /* Handle an IO access or exclusive access.  */
>>>      if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
>>> -        CPUIOTLBEntry *iotlbentry;
>>> -        if ((addr & (DATA_SIZE - 1)) != 0) {
>>> -            goto do_unaligned_access;
>>> +        CPUIOTLBEntry *iotlbentry = &env->iotlb[mmu_idx][index];
>>> +
>>> +        if ((tlb_addr & ~TARGET_PAGE_MASK) == TLB_EXCL) {
>>> +            CPUState *cpu = ENV_GET_CPU(env);
>>> +            /* The slow-path has been forced since we are writing to
>>> +             * exclusive-protected memory. */
>>> +            hwaddr hw_addr = (iotlbentry->addr & TARGET_PAGE_MASK) + addr;
>>> +
>>> +            /* The function lookup_and_reset_cpus_ll_addr could have reset 
>>> the
>>> +             * exclusive address. Fail the SC in this case.
>>> +             * N.B.: Here excl_succeeded == 0 means that helper_le_st_name 
>>> has
>>> +             * not been called by a softmmu_llsc_template.h. */
>>
>> Could this be better worded (along with bool-ising) as:
>>
>> "excl_succeeded is set by helper_le_st_name (softmmu_llsc_template)."
>>
>> But having said that grepping for helper_le_st_name I see that's defined
>> in softmmu_template.h so now the comments has confused me.
>
> I see now that the comment refers to softmmu_llsc_template that will
> be created later on. Please consider this fixed.
> In any case excl_succeeded, as the name suggests, is used by
> helper_stcond_name to know if the exclusive access went well.
> However, it is also used by softmmu_template to know whether we came
> from softmmu_llsc_template or not. This behaviour is pointed out in a
> comment is softmmu_llsc_template.
>
>>
>> It also might be worth mentioning the subtly that exclusive addresses
>> are based on the real hwaddr (hence the iotlb lookup?).
>
> OK.
>
>>
>>> +            if (cpu->excl_succeeded) {
>>> +                if (cpu->excl_protected_range.begin != hw_addr) {
>>> +                    /* The vCPU is SC-ing to an unprotected address. */
>>> +                    cpu->excl_protected_range.begin = EXCLUSIVE_RESET_ADDR;
>>> +                    cpu->excl_succeeded = 0;
>>
>> cpu->excl_succeeded = false;
>>
>>> +
>>> +                    return;
>>> +                }
>>> +
>>> +                cpu_physical_memory_unset_excl(hw_addr, cpu->cpu_index);
>>> +            }
>>> +
>>> +            haddr = addr + env->tlb_table[mmu_idx][index].addend;
>>> +        #if DATA_SIZE == 1
>>> +            glue(glue(st, SUFFIX), _p)((uint8_t *)haddr, val);
>>> +        #else
>>> +            glue(glue(st, SUFFIX), _le_p)((uint8_t *)haddr, val);
>>> +        #endif
>>
>> Why the special casing for byte access? Isn't this something the glue +
>> SUFFIX magic is meant to sort out?
>
> For the byte access the byte ordering is irrelevant, in fact there is
> only one version of stb_p.

I'm just wondering why this little detail isn't hidden in the
st_SUFFIX_le_p helpers rather than having to be explicit here.

>
>>
>>> +
>>> +            lookup_and_reset_cpus_ll_addr(hw_addr, DATA_SIZE);
>>> +
>>> +            return;
>>> +        } else {
>>> +            if ((addr & (DATA_SIZE - 1)) != 0) {
>>> +                goto do_unaligned_access;
>>> +            }
>>> +            iotlbentry = &env->iotlb[mmu_idx][index];
>>
>> Are we re-loading the TLB entry here?
>
> Indeed, that should not be there (anyhow it will go away when we
> refactor this helper in patches 10,11,12).
>
>>
>>> +
>>> +            /* ??? Note that the io helpers always read data in the target
>>> +               byte ordering.  We should push the LE/BE request down into 
>>> io.  */
>>> +            val = TGT_LE(val);
>>> +            glue(io_write, SUFFIX)(env, iotlbentry, val, addr,
>>> retaddr);
>>
>> What happens if the software does and exclusive operation on a io
>> address?
>
> At this stage of the patch series such operations are not supported.
> Should I add an hw_error in case a software tries to do that?

If it is a known limitation then I think it might be worth it.

> As
> written above, patches 13 and 14 add the missing pieces to support
> exclusive operations to MMIO regions.
>
>>
>>> +            return;
>>>          }
>>> -        iotlbentry = &env->iotlb[mmu_idx][index];
>>> -
>>> -        /* ??? Note that the io helpers always read data in the target
>>> -           byte ordering.  We should push the LE/BE request down into io.  
>>> */
>>> -        val = TGT_LE(val);
>>> -        glue(io_write, SUFFIX)(env, iotlbentry, val, addr, retaddr);
>>> -        return;
>>>      }
>>>
>>>      /* Handle slow unaligned access (it spans two pages or IO).  */
>>> @@ -475,19 +510,54 @@ void helper_be_st_name(CPUArchState *env, 
>>> target_ulong addr, DATA_TYPE val,
>>>          tlb_addr = env->tlb_table[mmu_idx][index].addr_write;
>>>      }
>>>
>>> -    /* Handle an IO access.  */
>>> +    /* Handle an IO access or exclusive access.  */
>>
>> Hmm there looks like a massive amount of duplication (not your fault, it
>> was like that when you got here ;-) but maybe this can be re-factored
>> away somehow?
>
> That's true. This is why patches 10,11,12 try to alleviate this
> problem by making the code a little bit more compact and readable.

Although I was actually comparing the two helpers on the final tree
state and there was loads of duplication still. If there is general
re-factoring outside of the LL/SC case it would be worth having them at
the start of the patch series so those improvements can get pulled into
the tree and the LL/SC code is simpler when applied.

>
> Regards,
> alvise
>
>>
>>>      if (unlikely(tlb_addr & ~TARGET_PAGE_MASK)) {
>>> -        CPUIOTLBEntry *iotlbentry;
>>> -        if ((addr & (DATA_SIZE - 1)) != 0) {
>>> -            goto do_unaligned_access;
>>> +        CPUIOTLBEntry *iotlbentry = &env->iotlb[mmu_idx][index];
>>> +
>>> +        if ((tlb_addr & ~TARGET_PAGE_MASK) == TLB_EXCL) {
>>> +            CPUState *cpu = ENV_GET_CPU(env);
>>> +            /* The slow-path has been forced since we are writing to
>>> +             * exclusive-protected memory. */
>>> +            hwaddr hw_addr = (iotlbentry->addr & TARGET_PAGE_MASK) + addr;
>>> +
>>> +            /* The function lookup_and_reset_cpus_ll_addr could have reset 
>>> the
>>> +             * exclusive address. Fail the SC in this case.
>>> +             * N.B.: Here excl_succeeded == 0 means that helper_le_st_name 
>>> has
>>> +             * not been called by a softmmu_llsc_template.h. */
>>> +            if (cpu->excl_succeeded) {
>>> +                if (cpu->excl_protected_range.begin != hw_addr) {
>>> +                    /* The vCPU is SC-ing to an unprotected address. */
>>> +                    cpu->excl_protected_range.begin = EXCLUSIVE_RESET_ADDR;
>>> +                    cpu->excl_succeeded = 0;
>>> +
>>> +                    return;
>>> +                }
>>> +
>>> +                cpu_physical_memory_unset_excl(hw_addr, cpu->cpu_index);
>>> +            }
>>> +
>>> +            haddr = addr + env->tlb_table[mmu_idx][index].addend;
>>> +        #if DATA_SIZE == 1
>>> +            glue(glue(st, SUFFIX), _p)((uint8_t *)haddr, val);
>>> +        #else
>>> +            glue(glue(st, SUFFIX), _le_p)((uint8_t *)haddr, val);
>>> +        #endif
>>> +
>>> +            lookup_and_reset_cpus_ll_addr(hw_addr, DATA_SIZE);
>>> +
>>> +            return;
>>> +        } else {
>>> +            if ((addr & (DATA_SIZE - 1)) != 0) {
>>> +                goto do_unaligned_access;
>>> +            }
>>> +            iotlbentry = &env->iotlb[mmu_idx][index];
>>> +
>>> +            /* ??? Note that the io helpers always read data in the target
>>> +               byte ordering.  We should push the LE/BE request down into 
>>> io.  */
>>> +            val = TGT_BE(val);
>>> +            glue(io_write, SUFFIX)(env, iotlbentry, val, addr, retaddr);
>>> +            return;
>>>          }
>>> -        iotlbentry = &env->iotlb[mmu_idx][index];
>>> -
>>> -        /* ??? Note that the io helpers always read data in the target
>>> -           byte ordering.  We should push the LE/BE request down into io.  
>>> */
>>> -        val = TGT_BE(val);
>>> -        glue(io_write, SUFFIX)(env, iotlbentry, val, addr, retaddr);
>>> -        return;
>>>      }
>>>
>>>      /* Handle slow unaligned access (it spans two pages or IO).  */
>>
>>
>> --
>> Alex Bennée


--
Alex Bennée