Re: [PATCH v2 4/6] RISC-V: Detect unaligned vector accesses supported.

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Sun, Jun 23, 2024 at 10:34 PM Andy Chiu <andy.chiu@xxxxxxxxxx> wrote:
>
> On Fri, Jun 21, 2024 at 2:52 AM Evan Green <evan@xxxxxxxxxxxx> wrote:
> >
> > On Thu, Jun 13, 2024 at 12:17 PM Jesse Taube <jesse@xxxxxxxxxxxx> wrote:
> > >
> > > Run a unaligned vector access to test if the system supports
> > > vector unaligned access. Add the result to a new key in hwprobe.
> > > This is useful for usermode to know if vector misaligned accesses are
> > > supported and if they are faster or slower than equivalent byte accesses.
> > >
> > > Signed-off-by: Jesse Taube <jesse@xxxxxxxxxxxx>
> > > ---
> > > V1 -> V2:
> > >  - Add Kconfig options
> > >  - Add insn_is_vector
> > >  - Add handle_vector_misaligned_load
> > >  - Fix build
> > >  - Seperate vector from scalar misaligned access
> > >  - This patch was almost completely rewritten
> > > ---
> > >  arch/riscv/Kconfig                         |  41 +++++++
> > >  arch/riscv/include/asm/cpufeature.h        |   7 +-
> > >  arch/riscv/include/asm/entry-common.h      |  11 --
> > >  arch/riscv/include/asm/hwprobe.h           |   2 +-
> > >  arch/riscv/include/asm/vector.h            |   1 +
> > >  arch/riscv/include/uapi/asm/hwprobe.h      |   5 +
> > >  arch/riscv/kernel/Makefile                 |   4 +-
> > >  arch/riscv/kernel/sys_hwprobe.c            |  41 +++++++
> > >  arch/riscv/kernel/traps_misaligned.c       | 119 ++++++++++++++++++++-
> > >  arch/riscv/kernel/unaligned_access_speed.c |   9 +-
> > >  arch/riscv/kernel/vector.c                 |   2 +-
> > >  11 files changed, 221 insertions(+), 21 deletions(-)
> > >
> > > diff --git a/arch/riscv/Kconfig b/arch/riscv/Kconfig
> > > index b94176e25be1..f12df0ca6c18 100644
> > > --- a/arch/riscv/Kconfig
> > > +++ b/arch/riscv/Kconfig
> > > @@ -723,6 +723,12 @@ config RISCV_MISALIGNED
> > >         help
> > >           Embed support for emulating misaligned loads and stores.
> > >
> > > +config RISCV_VECTOR_MISALIGNED
> > > +       bool
> > > +       depends on RISCV_ISA_V
> > > +       help
> > > +         Enable detecting support for vector misaligned loads and stores.
> > > +
> > >  choice
> > >         prompt "Unaligned Accesses Support"
> > >         default RISCV_PROBE_UNALIGNED_ACCESS
> > > @@ -774,6 +780,41 @@ config RISCV_EFFICIENT_UNALIGNED_ACCESS
> > >
> > >  endchoice
> > >
> > > +choice
> > > +       prompt "Vector unaligned Accesses Support"
> > > +       depends on RISCV_ISA_V
> > > +       default RISCV_PROBE_VECTOR_UNALIGNED_ACCESS
> > > +       help
> > > +         This determines the level of support for vector unaligned accesses. This
> > > +         information is used by the kernel to perform optimizations. It is also
> > > +         exposed to user space via the hwprobe syscall. The hardware will be
> > > +         probed at boot by default.
> > > +
> > > +config RISCV_DETECT_VECTOR_UNALIGNED_ACCESS
> > > +       bool "Detect support for vector unaligned accesses"
> > > +       select RISCV_VECTOR_MISALIGNED
> > > +       help
> > > +         During boot, the kernel will detect if the system supports vector
> > > +         unaligned accesses.
> > > +
> > > +config RISCV_PROBE_VECTOR_UNALIGNED_ACCESS
> > > +       bool "Probe speed of vector unaligned accesses"
> > > +       select RISCV_VECTOR_MISALIGNED
> > > +       help
> > > +         During boot, the kernel will run a series of tests to determine the
> > > +         speed of vector unaligned accesses if they are supported. This probing
> > > +         will dynamically determine the speed of vector unaligned accesses on
> > > +         the underlying system if they are supported.
> > > +
> > > +config CONFIG_RISCV_UNALIGNED_ACCESS_UNSUPPORTED
> > > +       bool "Assume the system does not support vector unaligned memory accesses"
> > > +       help
> > > +         Assume that the system does not support vector unaligned memory accesses.
> > > +         The kernel and userspace programs may run them successfully on systems
> > > +         that do support vector unaligned memory accesses.
> > > +
> > > +endchoice
> > > +
> > >  endmenu # "Platform type"
> > >
> > >  menu "Kernel features"
> > > diff --git a/arch/riscv/include/asm/cpufeature.h b/arch/riscv/include/asm/cpufeature.h
> > > index 347805446151..d0ea5921ab20 100644
> > > --- a/arch/riscv/include/asm/cpufeature.h
> > > +++ b/arch/riscv/include/asm/cpufeature.h
> > > @@ -33,8 +33,8 @@ extern struct riscv_isainfo hart_isa[NR_CPUS];
> > >
> > >  void riscv_user_isa_enable(void);
> > >
> > > -#if defined(CONFIG_RISCV_MISALIGNED)
> > >  bool check_unaligned_access_emulated_all_cpus(void);
> > > +#if defined(CONFIG_RISCV_MISALIGNED)
> > >  void unaligned_emulation_finish(void);
> > >  bool unaligned_ctl_available(void);
> > >  DECLARE_PER_CPU(long, misaligned_access_speed);
> > > @@ -45,6 +45,11 @@ static inline bool unaligned_ctl_available(void)
> > >  }
> > >  #endif
> > >
> > > +bool check_vector_unaligned_access_emulated_all_cpus(void);
> > > +#if defined(CONFIG_RISCV_VECTOR_MISALIGNED)
> > > +DECLARE_PER_CPU(long, vector_misaligned_access);
> > > +#endif
> > > +
> > >  #if defined(CONFIG_RISCV_PROBE_UNALIGNED_ACCESS)
> > >  DECLARE_STATIC_KEY_FALSE(fast_unaligned_access_speed_key);
> > >
> > > diff --git a/arch/riscv/include/asm/entry-common.h b/arch/riscv/include/asm/entry-common.h
> > > index 2293e535f865..7b32d2b08bb6 100644
> > > --- a/arch/riscv/include/asm/entry-common.h
> > > +++ b/arch/riscv/include/asm/entry-common.h
> > > @@ -25,18 +25,7 @@ static inline void arch_exit_to_user_mode_prepare(struct pt_regs *regs,
> > >  void handle_page_fault(struct pt_regs *regs);
> > >  void handle_break(struct pt_regs *regs);
> > >
> > > -#ifdef CONFIG_RISCV_MISALIGNED
> > >  int handle_misaligned_load(struct pt_regs *regs);
> > >  int handle_misaligned_store(struct pt_regs *regs);
> > > -#else
> > > -static inline int handle_misaligned_load(struct pt_regs *regs)
> > > -{
> > > -       return -1;
> > > -}
> > > -static inline int handle_misaligned_store(struct pt_regs *regs)
> > > -{
> > > -       return -1;
> > > -}
> > > -#endif
> > >
> > >  #endif /* _ASM_RISCV_ENTRY_COMMON_H */
> > > diff --git a/arch/riscv/include/asm/hwprobe.h b/arch/riscv/include/asm/hwprobe.h
> > > index 150a9877b0af..ef01c182af2b 100644
> > > --- a/arch/riscv/include/asm/hwprobe.h
> > > +++ b/arch/riscv/include/asm/hwprobe.h
> > > @@ -8,7 +8,7 @@
> > >
> > >  #include <uapi/asm/hwprobe.h>
> > >
> > > -#define RISCV_HWPROBE_MAX_KEY 7
> > > +#define RISCV_HWPROBE_MAX_KEY 8
> > >
> > >  static inline bool riscv_hwprobe_key_is_valid(__s64 key)
> > >  {
> > > diff --git a/arch/riscv/include/asm/vector.h b/arch/riscv/include/asm/vector.h
> > > index be7d309cca8a..99b0f91db9ee 100644
> > > --- a/arch/riscv/include/asm/vector.h
> > > +++ b/arch/riscv/include/asm/vector.h
> > > @@ -21,6 +21,7 @@
> > >
> > >  extern unsigned long riscv_v_vsize;
> > >  int riscv_v_setup_vsize(void);
> > > +bool insn_is_vector(u32 insn_buf);
> > >  bool riscv_v_first_use_handler(struct pt_regs *regs);
> > >  void kernel_vector_begin(void);
> > >  void kernel_vector_end(void);
> > > diff --git a/arch/riscv/include/uapi/asm/hwprobe.h b/arch/riscv/include/uapi/asm/hwprobe.h
> > > index 023b7771d1b7..2fee870e41bb 100644
> > > --- a/arch/riscv/include/uapi/asm/hwprobe.h
> > > +++ b/arch/riscv/include/uapi/asm/hwprobe.h
> > > @@ -75,6 +75,11 @@ struct riscv_hwprobe {
> > >  #define                RISCV_HWPROBE_MISALIGNED_MASK           (7 << 0)
> > >  #define RISCV_HWPROBE_KEY_ZICBOZ_BLOCK_SIZE    6
> > >  #define RISCV_HWPROBE_KEY_MISALIGNED_PERF      7
> > > +#define RISCV_HWPROBE_KEY_VEC_MISALIGNED_PERF  8
> > > +#define                RISCV_HWPROBE_VEC_MISALIGNED_UNKNOWN            0
> > > +#define                RISCV_HWPROBE_VEC_MISALIGNED_SLOW               2
> > > +#define                RISCV_HWPROBE_VEC_MISALIGNED_FAST               3
> > > +#define                RISCV_HWPROBE_VEC_MISALIGNED_UNSUPPORTED        4
> > >  /* Increase RISCV_HWPROBE_MAX_KEY when adding items. */
> > >
> > >  /* Flags */
> > > diff --git a/arch/riscv/kernel/Makefile b/arch/riscv/kernel/Makefile
> > > index 5b243d46f4b1..62ac19c029f1 100644
> > > --- a/arch/riscv/kernel/Makefile
> > > +++ b/arch/riscv/kernel/Makefile
> > > @@ -62,8 +62,8 @@ obj-y += probes/
> > >  obj-y  += tests/
> > >  obj-$(CONFIG_MMU) += vdso.o vdso/
> > >
> > > -obj-$(CONFIG_RISCV_MISALIGNED) += traps_misaligned.o
> > > -obj-$(CONFIG_RISCV_MISALIGNED) += unaligned_access_speed.o
> > > +obj-y  += traps_misaligned.o
> > > +obj-y  += unaligned_access_speed.o
> > >  obj-$(CONFIG_RISCV_PROBE_UNALIGNED_ACCESS)     += copy-unaligned.o
> > >
> > >  obj-$(CONFIG_FPU)              += fpu.o
> > > diff --git a/arch/riscv/kernel/sys_hwprobe.c b/arch/riscv/kernel/sys_hwprobe.c
> > > index e910e2971984..c40df314058b 100644
> > > --- a/arch/riscv/kernel/sys_hwprobe.c
> > > +++ b/arch/riscv/kernel/sys_hwprobe.c
> > > @@ -194,6 +194,43 @@ static u64 hwprobe_misaligned(const struct cpumask *cpus)
> > >  }
> > >  #endif
> > >
> > > +#ifdef CONFIG_RISCV_VECTOR_MISALIGNED
> > > +static u64 hwprobe_vec_misaligned(const struct cpumask *cpus)
> > > +{
> > > +       int cpu;
> > > +       u64 perf = -1ULL;
> > > +
> > > +       if (IS_ENABLED(CONFIG_RISCV_UNALIGNED_ACCESS_UNSUPPORTED))
> > > +               return RISCV_HWPROBE_VEC_MISALIGNED_UNSUPPORTED;
> > > +
> > > +       /* Return if supported or not even if speed wasn't probed */
> > > +       for_each_cpu(cpu, cpus) {
> > > +               int this_perf = per_cpu(vector_misaligned_access, cpu);
> > > +
> > > +               if (perf == -1ULL)
> > > +                       perf = this_perf;
> > > +
> > > +               if (perf != this_perf) {
> > > +                       perf = RISCV_HWPROBE_VEC_MISALIGNED_UNKNOWN;
> > > +                       break;
> > > +               }
> > > +       }
> > > +
> > > +       if (perf == -1ULL)
> > > +               return RISCV_HWPROBE_VEC_MISALIGNED_UNKNOWN;
> > > +
> > > +       return perf;
> > > +}
> > > +#else
> > > +static u64 hwprobe_vec_misaligned(const struct cpumask *cpus)
> > > +{
> > > +       if (IS_ENABLED(CONFIG_RISCV_UNALIGNED_ACCESS_UNSUPPORTED))
> > > +               return RISCV_HWPROBE_VEC_MISALIGNED_UNSUPPORTED;
> > > +
> > > +       return RISCV_HWPROBE_VEC_MISALIGNED_UNKNOWN;
> > > +}
> > > +#endif
> > > +
> > >  static void hwprobe_one_pair(struct riscv_hwprobe *pair,
> > >                              const struct cpumask *cpus)
> > >  {
> > > @@ -222,6 +259,10 @@ static void hwprobe_one_pair(struct riscv_hwprobe *pair,
> > >                 pair->value = hwprobe_misaligned(cpus);
> > >                 break;
> > >
> > > +       case RISCV_HWPROBE_KEY_VEC_MISALIGNED_PERF:
> > > +               pair->value = hwprobe_vec_misaligned(cpus);
> > > +               break;
> > > +
> > >         case RISCV_HWPROBE_KEY_ZICBOZ_BLOCK_SIZE:
> > >                 pair->value = 0;
> > >                 if (hwprobe_ext0_has(cpus, RISCV_HWPROBE_EXT_ZICBOZ))
> > > diff --git a/arch/riscv/kernel/traps_misaligned.c b/arch/riscv/kernel/traps_misaligned.c
> > > index 8fadbe00dd62..6f0264a8c9de 100644
> > > --- a/arch/riscv/kernel/traps_misaligned.c
> > > +++ b/arch/riscv/kernel/traps_misaligned.c
> > > @@ -16,6 +16,7 @@
> > >  #include <asm/entry-common.h>
> > >  #include <asm/hwprobe.h>
> > >  #include <asm/cpufeature.h>
> > > +#include <asm/vector.h>
> > >
> > >  #define INSN_MATCH_LB                  0x3
> > >  #define INSN_MASK_LB                   0x707f
> > > @@ -322,12 +323,37 @@ union reg_data {
> > >         u64 data_u64;
> > >  };
> > >
> > > -static bool unaligned_ctl __read_mostly;
> > > -
> > >  /* sysctl hooks */
> > >  int unaligned_enabled __read_mostly = 1;       /* Enabled by default */
> > >
> > > -int handle_misaligned_load(struct pt_regs *regs)
> > > +#ifdef CONFIG_RISCV_VECTOR_MISALIGNED
> > > +static int handle_vector_misaligned_load(struct pt_regs *regs)
> > > +{
> > > +       unsigned long epc = regs->epc;
> > > +       unsigned long insn;
> > > +
> > > +       if (get_insn(regs, epc, &insn))
> > > +               return -1;
> > > +
> > > +       /* Only return 0 when in check_vector_unaligned_access_emulated */
> > > +       if (*this_cpu_ptr(&vector_misaligned_access) == RISCV_HWPROBE_VEC_MISALIGNED_UNKNOWN) {
> > > +               *this_cpu_ptr(&vector_misaligned_access) = RISCV_HWPROBE_VEC_MISALIGNED_UNSUPPORTED;
> > > +               regs->epc = epc + INSN_LEN(insn);
> > > +               return 0;
> > > +       }
> > > +
> > > +       /* If vector instruction we don't emulate it yet */
> > > +       regs->epc = epc;
> > > +       return -1;
> > > +}
> > > +#else
> > > +static int handle_vector_misaligned_load(struct pt_regs *regs)
> > > +{
> > > +       return -1;
> > > +}
> > > +#endif
> > > +
> > > +static int handle_scalar_misaligned_load(struct pt_regs *regs)
> > >  {
> > >         union reg_data val;
> > >         unsigned long epc = regs->epc;
> > > @@ -435,7 +461,7 @@ int handle_misaligned_load(struct pt_regs *regs)
> > >         return 0;
> > >  }
> > >
> > > -int handle_misaligned_store(struct pt_regs *regs)
> > > +static int handle_scalar_misaligned_store(struct pt_regs *regs)
> > >  {
> > >         union reg_data val;
> > >         unsigned long epc = regs->epc;
> > > @@ -526,6 +552,85 @@ int handle_misaligned_store(struct pt_regs *regs)
> > >         return 0;
> > >  }
> > >
> > > +int handle_misaligned_load(struct pt_regs *regs)
> > > +{
> > > +       unsigned long epc = regs->epc;
> > > +       unsigned long insn;
> > > +
> > > +       if (IS_ENABLED(CONFIG_RISCV_VECTOR_MISALIGNED)) {
> > > +               if (get_insn(regs, epc, &insn))
> > > +                       return -1;
> > > +
> > > +               if (insn_is_vector(insn))
> > > +                       return handle_vector_misaligned_load(regs);
> > > +       }
> > > +
> > > +       if (IS_ENABLED(CONFIG_RISCV_MISALIGNED))
> > > +               return handle_scalar_misaligned_load(regs);
> > > +
> > > +       return -1;
> > > +}
> > > +
> > > +int handle_misaligned_store(struct pt_regs *regs)
> > > +{
> > > +       if (IS_ENABLED(CONFIG_RISCV_MISALIGNED))
> > > +               return handle_scalar_misaligned_store(regs);
> > > +
> > > +       return -1;
> > > +}
> > > +
> > > +#ifdef CONFIG_RISCV_VECTOR_MISALIGNED
> > > +static void check_vector_unaligned_access_emulated(struct work_struct *unused)
> > > +{
> > > +       long *mas_ptr = this_cpu_ptr(&vector_misaligned_access);
> > > +       unsigned long tmp_var;
> > > +
> > > +       *mas_ptr = RISCV_HWPROBE_VEC_MISALIGNED_UNKNOWN;
> > > +
> > > +       local_irq_enable();
> >
> > Generally if a function is called with interrupts disabled there's a
> > reason for it, like the system will crash if an interrupt fires during
> > execution of this region. I haven't researched this, but to feel
> > comfortable I'd want to know why interrupts were disabled on entry
> > here, why it's safe to enable them now, and why it's safe to return
> > from the function with them still enabled.
> >
> > I'm guessing this was added because may_use_simd() was blowing up for
> > you without it. If that's the case, I think we'll need to reconcile
> > that in a different way. From a quick glance at kernel_mode_vector.c,
> > I was originally thinking may_use_simd() enforces this because there's
>
> Current nesting support of kernel mode vector only allows softirq
> context to use v context nesting on top of kernel thread's v context.
> I did not expect the use of v in the interrupt handler but I am open
> to discussions if it is needed.
>
> The reason why may_use_simd() checks for irq_disabled() is that
> local_bh_enable later in kernel_vector_end() expects running with irq
> on. Although the success path of preemptible v does not require
> !irq_disabled(), the kernel falls back to non-preemptible v
> (get_cpu_vector_context) in some cases. First is the nesting softirq
> case, or a rare case where an exception handler uses v in a kernel
> fault that also uses v. The second is when memory allocation for
> kernel_vstate fails. The fall back uses the non-preemptible v, so it
> is linked to bh_*.
>
> We can get rid of the last case by failing the launch of a kernel
> thread when the allocation fails.

Thanks for the context Andy, that fills in a couple of "wait why" gaps
I had in my brain while reading that code.

My understanding is Jesse skated around this issue by using
schedule_on_each_cpu(), which gets the function running on each cpu in
a nicer worker context where kernel_vector_begin() will work as is.
This local_irq_enable() was a leftover from when she was originally
experimenting with on_each_cpu(), which invokes the callback from a
harsher IPI context. So I think we can just remove the
local_irq_enable() and not have to modify the kernel V context code.
-Evan





[Index of Archives]     [Device Tree Compilter]     [Device Tree Spec]     [Linux Driver Backports]     [Video for Linux]     [Linux USB Devel]     [Linux PCI Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Yosemite Backpacking]


  Powered by Linux