Re: [Qemu-devel] [PATCH for-2.11 06/27] x86: extract legacy cpu features format parser

[Eduardo Habkost]

On Fri, Aug 18, 2017 at 12:08:38PM +0200, Igor Mammedov wrote:
> Move cpu_model +-feat parsing into a separate file so that it
> could be reused later for parsing similar format of sparc target
> 
> Signed-off-by: Igor Mammedov <address@hidden>
> ---
> CC: Richard Henderson <address@hidden>
> CC: Eduardo Habkost <address@hidden>
> CC: Mark Cave-Ayland <address@hidden>
> CC: Artyom Tarasenko <address@hidden>
> CC: Philippe Mathieu-Daudé <address@hidden>
> ---
>  include/qom/cpu.h                     |   2 +
>  default-configs/i386-bsd-user.mak     |   1 +
>  default-configs/i386-linux-user.mak   |   1 +
>  default-configs/i386-softmmu.mak      |   1 +
>  default-configs/x86_64-bsd-user.mak   |   1 +
>  default-configs/x86_64-linux-user.mak |   1 +
>  default-configs/x86_64-softmmu.mak    |   1 +
>  target/i386/cpu.c                     | 125 +-------------------------
>  util/Makefile.objs                    |   1 +
>  util/legacy_cpu_features_parser.c     | 161 
> ++++++++++++++++++++++++++++++++++
>  10 files changed, 171 insertions(+), 124 deletions(-)
>  create mode 100644 util/legacy_cpu_features_parser.c
> 
[...]
> diff --git a/util/legacy_cpu_features_parser.c 
> b/util/legacy_cpu_features_parser.c
> new file mode 100644
> index 0000000..6b352a3
> --- /dev/null
> +++ b/util/legacy_cpu_features_parser.c
> @@ -0,0 +1,161 @@
> +/* Support for legacy -cpu cpu,features CLI option with +-feat syntax,
> + * used by x86/sparc targets
> + *
> + * Author: Andreas Färber <address@hidden>
> + * Author: Andre Przywara <address@hidden>
> + * Author: Eduardo Habkost <address@hidden>
> + * Author: Igor Mammedov <address@hidden>
> + * Author: Paolo Bonzini <address@hidden>
> + * Author: Markus Armbruster <address@hidden>

IANAL, but I believe a
  Copyright (c) <YEAR> <COPYRIGHT HOLDER>
line is needed here.

> + *
> + * This program is free software; you can redistribute it and/or modify
> + * it under the terms of the GNU General Public License as published by
> + * the Free Software Foundation; either version 2 of the License, or
> + * (at your option) any later version.
> +
> + * This program is distributed in the hope that it will be useful,
> + * but WITHOUT ANY WARRANTY; without even the implied warranty of
> + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> + * GNU General Public License for more details.
> +
> + * You should have received a copy of the GNU General Public License along
> + * with this program; if not, see <http://www.gnu.org/licenses/>.
> + */
> +
> +#include "qemu/osdep.h"
> +#include "qapi/error.h"
> +#include "qemu/cutils.h"
> +#include "qom/cpu.h"
> +#include "qemu/error-report.h"
> +#include "hw/qdev-properties.h"
> +
> +static inline void feat2prop(char *s)
> +{
> +    while ((s = strchr(s, '_'))) {
> +        *s = '-';
> +    }
> +}
> +
> +static gint compare_string(gconstpointer a, gconstpointer b)
> +{
> +    return g_strcmp0(a, b);
> +}
> +
> +static void
> +cpu_add_feat_as_prop(const char *typename, const char *name, const char *val)
> +{
> +    GlobalProperty *prop = g_new0(typeof(*prop), 1);
> +    prop->driver = typename;
> +    prop->property = g_strdup(name);
> +    prop->value = g_strdup(val);
> +    prop->errp = &error_fatal;
> +    qdev_prop_register_global(prop);
> +}
> +
> +/* DO NOT USE WITH NEW CODE
> + * Parse "+feature,-feature,feature=foo" CPU feature string
> + */
> +void cpu_legacy_parse_featurestr(const char *typename, char *features,
> +                                 Error **errp)
> +{
> +    /* Compatibily hack to maintain legacy +-feat semantic,
> +     * where +-feat overwrites any feature set by
> +     * feat=on|feat even if the later is parsed after +-feat
> +     * (i.e. "-x2apic,x2apic=on" will result in x2apic disabled)
> +     */
> +    GList *l, *plus_features = NULL, *minus_features = NULL;
> +    char *featurestr; /* Single 'key=value" string being parsed */
> +    static bool cpu_globals_initialized;
> +    bool ambiguous = false;
> +
> +    if (cpu_globals_initialized) {
> +        return;
> +    }
> +    cpu_globals_initialized = true;
> +
> +    if (!features) {
> +        return;
> +    }
> +
> +    for (featurestr = strtok(features, ",");
> +         featurestr;
> +         featurestr = strtok(NULL, ",")) {
> +        const char *name;
> +        const char *val = NULL;
> +        char *eq = NULL;
> +        char num[32];
> +
> +        /* Compatibility syntax: */
> +        if (featurestr[0] == '+') {
> +            plus_features = g_list_append(plus_features,
> +                                          g_strdup(featurestr + 1));
> +            continue;
> +        } else if (featurestr[0] == '-') {
> +            minus_features = g_list_append(minus_features,
> +                                           g_strdup(featurestr + 1));
> +            continue;
> +        }

These 6 lines of code (or something equivalent to them) are
supposed to be the only difference to the generic parsing
function.  I would simply make this feature (support for
[+-]feature) enabled by a CPUClass::plus_minus_features flag
handled by cpu_common_parse_features().

(But this can be done as a follow-up patch.)

> +
> +        eq = strchr(featurestr, '=');
> +        if (eq) {
> +            *eq++ = 0;
> +            val = eq;
> +        } else {
> +            val = "on";
> +        }
> +
> +        feat2prop(featurestr);
> +        name = featurestr;
> +
> +        if (g_list_find_custom(plus_features, name, compare_string)) {
> +            warn_report("Ambiguous CPU model string. "
> +                        "Don't mix both \"+%s\" and \"%s=%s\"",
> +                        name, name, val);
> +            ambiguous = true;
> +        }
> +        if (g_list_find_custom(minus_features, name, compare_string)) {
> +            warn_report("Ambiguous CPU model string. "
> +                        "Don't mix both \"-%s\" and \"%s=%s\"",
> +                        name, name, val);
> +            ambiguous = true;
> +        }
> +
> +        /* Special case: */
> +        if (!strcmp(name, "tsc-freq")) {
> +            int ret;
> +            uint64_t tsc_freq;
> +
> +            ret = qemu_strtosz_metric(val, NULL, &tsc_freq);
> +            if (ret < 0 || tsc_freq > INT64_MAX) {
> +                error_setg(errp, "bad numerical value %s", val);
> +                return;
> +            }
> +            snprintf(num, sizeof(num), "%" PRId64, tsc_freq);
> +            val = num;
> +            name = "tsc-frequency";
> +        }

This is x86-specific and should stay in x86-specific code.  It
can probably be handled by the tsc-freq setter.

> +
> +        cpu_add_feat_as_prop(typename, name, val);
> +    }
> +
> +    if (ambiguous) {
> +        warn_report("Compatibility of ambiguous CPU model "
> +                    "strings won't be kept on future QEMU versions");
> +    }

As noted in the review of the x86 patch that removes the
plus_features/minus_features static variables, this obsolete (and
confusing) property ordering misfeature should be removed before
we make this code generic and reuse it on other architectures.

> +
> +    for (l = plus_features; l; l = l->next) {
> +        const char *name = l->data;
> +        cpu_add_feat_as_prop(typename, name, "on");
> +    }
> +    if (plus_features) {
> +        g_list_free_full(plus_features, g_free);
> +    }
> +
> +    for (l = minus_features; l; l = l->next) {
> +        const char *name = l->data;
> +        cpu_add_feat_as_prop(typename, name, "off");
> +    }
> +    if (minus_features) {
> +        g_list_free_full(minus_features, g_free);
> +    }
> +}
> -- 
> 2.7.4
> 
> 

-- 
Eduardo