Re: [PULL 21/30] target/arm: use official org.gnu.gdb.aarch64.sve layout for registers

[Claudio Fontana]

Hi Alex,

after updating to latest master today, I am getting the following error with

make check-tcg

qemu-system-aarch64: -gdb 
unix:path=/tmp/tmp9ru5tgk8qemu-gdbstub/gdbstub.socket,server: info: QEMU 
waiting for connection on: 
disconnected:unix:/tmp/tmp9ru5tgk8qemu-gdbstub/gdbstub.socket,server
warning: while parsing target description (at line 47): Vector "svevhf" 
references undefined type "ieee_half"
warning: Could not load XML target description; ignoring
qemu-system-aarch64: QEMU: Terminated via GDBstub

Seems to indicate it is "ieee_half" -related?

Thanks,

Claudio

On 1/15/21 2:08 PM, Alex Bennée wrote:
> While GDB can work with any XML description given to it there is
> special handling for SVE registers on the GDB side which makes the
> users life a little better. The changes aren't that major and all the
> registers save the $vg reported the same. All that changes is:
> 
>   - report org.gnu.gdb.aarch64.sve
>   - use gdb nomenclature for names and types
>   - minor re-ordering of the types to match reference
>   - re-enable ieee_half (as we know gdb supports it now)
>   - $vg is now a 64 bit int
>   - check $vN and $zN aliasing in test
> 
> Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
> Reviewed-by: Luis Machado <luis.machado@linaro.org>
> Message-Id: <20210108224256.2321-11-alex.bennee@linaro.org>
> 
> diff --git a/target/arm/gdbstub.c b/target/arm/gdbstub.c
> index 866595b4f1..a8fff2a3d0 100644
> --- a/target/arm/gdbstub.c
> +++ b/target/arm/gdbstub.c
> @@ -195,22 +195,17 @@ static const struct TypeSize vec_lanes[] = {
>      { "uint128", 128, 'q', 'u' },
>      { "int128", 128, 'q', 's' },
>      /* 64 bit */
> +    { "ieee_double", 64, 'd', 'f' },
>      { "uint64", 64, 'd', 'u' },
>      { "int64", 64, 'd', 's' },
> -    { "ieee_double", 64, 'd', 'f' },
>      /* 32 bit */
> +    { "ieee_single", 32, 's', 'f' },
>      { "uint32", 32, 's', 'u' },
>      { "int32", 32, 's', 's' },
> -    { "ieee_single", 32, 's', 'f' },
>      /* 16 bit */
> +    { "ieee_half", 16, 'h', 'f' },
>      { "uint16", 16, 'h', 'u' },
>      { "int16", 16, 'h', 's' },
> -    /*
> -     * TODO: currently there is no reliable way of telling
> -     * if the remote gdb actually understands ieee_half so
> -     * we don't expose it in the target description for now.
> -     * { "ieee_half", 16, 'h', 'f' },
> -     */
>      /* bytes */
>      { "uint8", 8, 'b', 'u' },
>      { "int8", 8, 'b', 's' },
> @@ -223,17 +218,16 @@ int arm_gen_dynamic_svereg_xml(CPUState *cs, int 
> base_reg)
>      GString *s = g_string_new(NULL);
>      DynamicGDBXMLInfo *info = &cpu->dyn_svereg_xml;
>      g_autoptr(GString) ts = g_string_new("");
> -    int i, bits, reg_width = (cpu->sve_max_vq * 128);
> +    int i, j, bits, reg_width = (cpu->sve_max_vq * 128);
>      info->num = 0;
>      g_string_printf(s, "<?xml version=\"1.0\"?>");
>      g_string_append_printf(s, "<!DOCTYPE target SYSTEM \"gdb-target.dtd\">");
> -    g_string_append_printf(s, "<feature name=\"org.qemu.gdb.aarch64.sve\">");
> +    g_string_append_printf(s, "<feature name=\"org.gnu.gdb.aarch64.sve\">");
>  
>      /* First define types and totals in a whole VL */
>      for (i = 0; i < ARRAY_SIZE(vec_lanes); i++) {
>          int count = reg_width / vec_lanes[i].size;
> -        g_string_printf(ts, "vq%d%c%c", count,
> -                        vec_lanes[i].sz, vec_lanes[i].suffix);
> +        g_string_printf(ts, "svev%c%c", vec_lanes[i].sz, 
> vec_lanes[i].suffix);
>          g_string_append_printf(s,
>                                 "<vector id=\"%s\" type=\"%s\" 
> count=\"%d\"/>",
>                                 ts->str, vec_lanes[i].gdb_type, count);
> @@ -243,39 +237,37 @@ int arm_gen_dynamic_svereg_xml(CPUState *cs, int 
> base_reg)
>       * signed and potentially float versions of each size from 128 to
>       * 8 bits.
>       */
> -    for (bits = 128; bits >= 8; bits /= 2) {
> -        int count = reg_width / bits;
> -        g_string_append_printf(s, "<union id=\"vq%dn\">", count);
> -        for (i = 0; i < ARRAY_SIZE(vec_lanes); i++) {
> -            if (vec_lanes[i].size == bits) {
> -                g_string_append_printf(s, "<field name=\"%c\" 
> type=\"vq%d%c%c\"/>",
> -                                       vec_lanes[i].suffix,
> -                                       count,
> -                                       vec_lanes[i].sz, vec_lanes[i].suffix);
> +    for (bits = 128, i = 0; bits >= 8; bits /= 2, i++) {
> +        const char suf[] = { 'q', 'd', 's', 'h', 'b' };
> +        g_string_append_printf(s, "<union id=\"svevn%c\">", suf[i]);
> +        for (j = 0; j < ARRAY_SIZE(vec_lanes); j++) {
> +            if (vec_lanes[j].size == bits) {
> +                g_string_append_printf(s, "<field name=\"%c\" 
> type=\"svev%c%c\"/>",
> +                                       vec_lanes[j].suffix,
> +                                       vec_lanes[j].sz, vec_lanes[j].suffix);
>              }
>          }
>          g_string_append(s, "</union>");
>      }
>      /* And now the final union of unions */
> -    g_string_append(s, "<union id=\"vq\">");
> -    for (bits = 128; bits >= 8; bits /= 2) {
> -        int count = reg_width / bits;
> -        for (i = 0; i < ARRAY_SIZE(vec_lanes); i++) {
> -            if (vec_lanes[i].size == bits) {
> -                g_string_append_printf(s, "<field name=\"%c\" 
> type=\"vq%dn\"/>",
> -                                       vec_lanes[i].sz, count);
> -                break;
> -            }
> -        }
> +    g_string_append(s, "<union id=\"svev\">");
> +    for (bits = 128, i = 0; bits >= 8; bits /= 2, i++) {
> +        const char suf[] = { 'q', 'd', 's', 'h', 'b' };
> +        g_string_append_printf(s, "<field name=\"%c\" type=\"svevn%c\"/>",
> +                               suf[i], suf[i]);
>      }
>      g_string_append(s, "</union>");
>  
> +    /* Finally the sve prefix type */
> +    g_string_append_printf(s,
> +                           "<vector id=\"svep\" type=\"uint8\" 
> count=\"%d\"/>",
> +                           reg_width / 8);
> +
>      /* Then define each register in parts for each vq */
>      for (i = 0; i < 32; i++) {
>          g_string_append_printf(s,
>                                 "<reg name=\"z%d\" bitsize=\"%d\""
> -                               " regnum=\"%d\" group=\"vector\""
> -                               " type=\"vq\"/>",
> +                               " regnum=\"%d\" type=\"svev\"/>",
>                                 i, reg_width, base_reg++);
>          info->num++;
>      }
> @@ -287,31 +279,22 @@ int arm_gen_dynamic_svereg_xml(CPUState *cs, int 
> base_reg)
>                             " regnum=\"%d\" group=\"float\""
>                             " type=\"int\"/>", base_reg++);
>      info->num += 2;
> -    /*
> -     * Predicate registers aren't so big they are worth splitting up
> -     * but we do need to define a type to hold the array of quad
> -     * references.
> -     */
> -    g_string_append_printf(s,
> -                           "<vector id=\"vqp\" type=\"uint16\" 
> count=\"%d\"/>",
> -                           cpu->sve_max_vq);
> +
>      for (i = 0; i < 16; i++) {
>          g_string_append_printf(s,
>                                 "<reg name=\"p%d\" bitsize=\"%d\""
> -                               " regnum=\"%d\" group=\"vector\""
> -                               " type=\"vqp\"/>",
> +                               " regnum=\"%d\" type=\"svep\"/>",
>                                 i, cpu->sve_max_vq * 16, base_reg++);
>          info->num++;
>      }
>      g_string_append_printf(s,
>                             "<reg name=\"ffr\" bitsize=\"%d\""
>                             " regnum=\"%d\" group=\"vector\""
> -                           " type=\"vqp\"/>",
> +                           " type=\"svep\"/>",
>                             cpu->sve_max_vq * 16, base_reg++);
>      g_string_append_printf(s,
>                             "<reg name=\"vg\" bitsize=\"64\""
> -                           " regnum=\"%d\" group=\"vector\""
> -                           " type=\"uint32\"/>",
> +                           " regnum=\"%d\" type=\"int\"/>",
>                             base_reg++);
>      info->num += 2;
>      g_string_append_printf(s, "</feature>");
> diff --git a/target/arm/helper.c b/target/arm/helper.c
> index 5ab3f5ace3..8a492465d6 100644
> --- a/target/arm/helper.c
> +++ b/target/arm/helper.c
> @@ -276,7 +276,7 @@ static int arm_gdb_get_svereg(CPUARMState *env, 
> GByteArray *buf, int reg)
>           * while the ZCR works in Vector Quads (VQ) which is 128bit chunks.
>           */
>          int vq = sve_zcr_len_for_el(env, arm_current_el(env)) + 1;
> -        return gdb_get_reg32(buf, vq * 2);
> +        return gdb_get_reg64(buf, vq * 2);
>      }
>      default:
>          /* gdbstub asked for something out our range */
> diff --git a/tests/tcg/aarch64/gdbstub/test-sve-ioctl.py 
> b/tests/tcg/aarch64/gdbstub/test-sve-ioctl.py
> index 972cf73c31..b9ef169c1a 100644
> --- a/tests/tcg/aarch64/gdbstub/test-sve-ioctl.py
> +++ b/tests/tcg/aarch64/gdbstub/test-sve-ioctl.py
> @@ -40,6 +40,17 @@ class TestBreakpoint(gdb.Breakpoint):
>          except gdb.error:
>              report(False, "checking zregs (out of range)")
>  
> +        # Check the aliased V registers are set and GDB has correctly
> +        # created them for us having recognised and handled SVE.
> +        try:
> +            for i in range(0, 16):
> +                val_z = gdb.parse_and_eval("$z0.b.u[%d]" % i)
> +                val_v = gdb.parse_and_eval("$v0.b.u[%d]" % i)
> +                report(int(val_z) == int(val_v),
> +                       "v0.b.u[%d] == z0.b.u[%d]" % (i, i))
> +        except gdb.error:
> +            report(False, "checking vregs (out of range)")
> +
>  
>  def run_test():
>      "Run through the tests one by one"
>