Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1645671 > unrolled thread

[PATCH] kconfig: always use user input symbols

Started byTycho Andersen <tycho@docker.com>
First post2017-05-19 17:10 +0200
Last post2017-05-19 20:20 +0200
Articles 3 — 2 participants

Back to article view | Back to linux.kernel


Contents

  [PATCH] kconfig: always use user input symbols Tycho Andersen <tycho@docker.com> - 2017-05-19 17:10 +0200
    Re: [PATCH] kconfig: always use user input symbols Geert Uytterhoeven <geert@linux-m68k.org> - 2017-05-19 19:30 +0200
      Re: [PATCH] kconfig: always use user input symbols Tycho Andersen <tycho@docker.com> - 2017-05-19 20:20 +0200

#1645671 — [PATCH] kconfig: always use user input symbols

FromTycho Andersen <tycho@docker.com>
Date2017-05-19 17:10 +0200
Subject[PATCH] kconfig: always use user input symbols
Message-ID<tIRHr-3XA-13@gated-at.bofh.it>
...regardless of visibility.

When a symbol that is not visible by default (e.g. PNFS_FLEXFILE_LAYOUT)
has a default value, it is impossible to set the value to something not the
default:

~/packages/linux render-symbol-inputs grep FLEXFILE .config
CONFIG_PNFS_FLEXFILE_LAYOUT=y
~/packages/linux render-symbol-inputs make oldconfig
scripts/kconfig/conf  --oldconfig Kconfig
~/packages/linux render-symbol-inputs grep FLEXFILE .config
CONFIG_PNFS_FLEXFILE_LAYOUT=m

There are two reasons for this: the symbol's user input value is only
considered when it is visible (hunks 2 and 3), and the user values are
explicitly ignored (hunk 1) if the symbols are not visible.

It's not clear to me why hunk 1 exists. I'm sure it solve some problem, but
I'm not sure why we would ever want to discard user input values, and
causes a problem exactly as the comment describes.

Signed-off-by: Tycho Andersen <tycho@docker.com>
CC: Roman Zippel <zippel@linux-m68k.org>
---
I don't know much about how kconfig works, I just noticed this when trying
to do some automatic merging of configs. My main goal is to build a kernel
without modules, but after a merge and running `make oldconfig`, some
symbols always come back set as modules.
---
 scripts/kconfig/confdata.c |  7 -------
 scripts/kconfig/symbol.c   | 32 ++++++++++++++++++--------------
 2 files changed, 18 insertions(+), 21 deletions(-)

diff --git a/scripts/kconfig/confdata.c b/scripts/kconfig/confdata.c
index 297b079..3537090 100644
--- a/scripts/kconfig/confdata.c
+++ b/scripts/kconfig/confdata.c
@@ -448,13 +448,6 @@ int conf_read(const char *name)
 
 	for_all_symbols(i, sym) {
 		if (sym_has_value(sym) && !sym_is_choice_value(sym)) {
-			/* Reset values of generates values, so they'll appear
-			 * as new, if they should become visible, but that
-			 * doesn't quite work if the Kconfig and the saved
-			 * configuration disagree.
-			 */
-			if (sym->visible == no && !conf_unsaved)
-				sym->flags &= ~SYMBOL_DEF_USER;
 			switch (sym->type) {
 			case S_STRING:
 			case S_INT:
diff --git a/scripts/kconfig/symbol.c b/scripts/kconfig/symbol.c
index 20136ff..27fca39 100644
--- a/scripts/kconfig/symbol.c
+++ b/scripts/kconfig/symbol.c
@@ -386,17 +386,19 @@ void sym_calc_value(struct symbol *sym)
 			prop = sym_get_choice_prop(sym);
 			newval.tri = (prop_get_symbol(prop)->curr.val == sym) ? yes : no;
 		} else {
-			if (sym->visible != no) {
-				/* if the symbol is visible use the user value
-				 * if available, otherwise try the default value
-				 */
+			/* If the user defined a value, let's not ignore it,
+			 * even if the symbol is not visible.
+			 */
+			if (sym_has_value(sym)) {
 				sym->flags |= SYMBOL_WRITE;
-				if (sym_has_value(sym)) {
-					newval.tri = EXPR_AND(sym->def[S_DEF_USER].tri,
-							      sym->visible);
-					goto calc_newval;
-				}
+				newval.tri = sym->def[S_DEF_USER].tri;
+				goto calc_newval;
 			}
+
+			if (sym->visible != no)
+				sym->flags |= SYMBOL_WRITE;
+
+			/* otherwise, let's use the default */
 			if (sym->rev_dep.tri != no)
 				sym->flags |= SYMBOL_WRITE;
 			if (!sym_is_choice(sym)) {
@@ -433,13 +435,15 @@ void sym_calc_value(struct symbol *sym)
 	case S_STRING:
 	case S_HEX:
 	case S_INT:
-		if (sym->visible != no) {
+		if (sym_has_value(sym)) {
 			sym->flags |= SYMBOL_WRITE;
-			if (sym_has_value(sym)) {
-				newval.val = sym->def[S_DEF_USER].val;
-				break;
-			}
+			newval.val = sym->def[S_DEF_USER].val;
+			goto calc_newval;
 		}
+
+		if (sym->visible != no)
+			sym->flags |= SYMBOL_WRITE;
+
 		prop = sym_get_default_prop(sym);
 		if (prop) {
 			struct symbol *ds = prop_get_symbol(prop);
-- 
2.9.3

[toc] | [next] | [standalone]


#1645780

FromGeert Uytterhoeven <geert@linux-m68k.org>
Date2017-05-19 19:30 +0200
Message-ID<tITSW-5kd-1@gated-at.bofh.it>
In reply to#1645671
Hi Tycho,

On Fri, May 19, 2017 at 5:08 PM, Tycho Andersen <tycho@docker.com> wrote:
> ...regardless of visibility.
>
> When a symbol that is not visible by default (e.g. PNFS_FLEXFILE_LAYOUT)
> has a default value, it is impossible to set the value to something not the
> default:
>
> ~/packages/linux render-symbol-inputs grep FLEXFILE .config
> CONFIG_PNFS_FLEXFILE_LAYOUT=y
> ~/packages/linux render-symbol-inputs make oldconfig
> scripts/kconfig/conf  --oldconfig Kconfig
> ~/packages/linux render-symbol-inputs grep FLEXFILE .config
> CONFIG_PNFS_FLEXFILE_LAYOUT=m
>
> There are two reasons for this: the symbol's user input value is only
> considered when it is visible (hunks 2 and 3), and the user values are
> explicitly ignored (hunk 1) if the symbols are not visible.
>
> It's not clear to me why hunk 1 exists. I'm sure it solve some problem, but
> I'm not sure why we would ever want to discard user input values, and
> causes a problem exactly as the comment describes.

This is intentional.  If a symbol is not visible, it's not meant to be changed
by the user, as doing so may break things (at build or runtime).

E.g. running "make oldconfig" after a kernel upgrade may retain the old
state of a variable, which is now invalid.  Ignoring invisible symbols
avoids this.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

[toc] | [prev] | [next] | [standalone]


#1645830

FromTycho Andersen <tycho@docker.com>
Date2017-05-19 20:20 +0200
Message-ID<tIUFk-5T5-23@gated-at.bofh.it>
In reply to#1645780
Hi Geert,

On Fri, May 19, 2017 at 07:29:05PM +0200, Geert Uytterhoeven wrote:
> Hi Tycho,
> 
> On Fri, May 19, 2017 at 5:08 PM, Tycho Andersen <tycho@docker.com> wrote:
> > ...regardless of visibility.
> >
> > When a symbol that is not visible by default (e.g. PNFS_FLEXFILE_LAYOUT)
> > has a default value, it is impossible to set the value to something not the
> > default:
> >
> > ~/packages/linux render-symbol-inputs grep FLEXFILE .config
> > CONFIG_PNFS_FLEXFILE_LAYOUT=y
> > ~/packages/linux render-symbol-inputs make oldconfig
> > scripts/kconfig/conf  --oldconfig Kconfig
> > ~/packages/linux render-symbol-inputs grep FLEXFILE .config
> > CONFIG_PNFS_FLEXFILE_LAYOUT=m
> >
> > There are two reasons for this: the symbol's user input value is only
> > considered when it is visible (hunks 2 and 3), and the user values are
> > explicitly ignored (hunk 1) if the symbols are not visible.
> >
> > It's not clear to me why hunk 1 exists. I'm sure it solve some problem, but
> > I'm not sure why we would ever want to discard user input values, and
> > causes a problem exactly as the comment describes.
> 
> This is intentional.  If a symbol is not visible, it's not meant to be changed
> by the user, as doing so may break things (at build or runtime).
> 
> E.g. running "make oldconfig" after a kernel upgrade may retain the old
> state of a variable, which is now invalid.  Ignoring invisible symbols
> avoids this.

Makes sense. Thanks!

Tycho

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web