Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
Hi Andrew, On Wed, Apr 1, 2015 at 2:54 PM, Kevin Hilman khil...@kernel.org wrote: Andrew Morton a...@linux-foundation.org writes: On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier marc.zyng...@arm.com wrote: -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page, - unsigned long private, struct page *page, int force, - enum migrate_mode mode) +static noinline int unmap_and_move(new_page_t get_new_page, + free_page_t put_new_page, + unsigned long private, struct page *page, + int force, enum migrate_mode mode) { int rc = 0; int *result = NULL; Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of the parameters on the stack, which is not going to help performance either (not that this would be useful on 32bit ARM anyway...). Any chance you could make this dependent on some compiler detection mechanism? With my arm compiler (gcc-4.4.4) the patch makes no difference - unmap_and_move() isn't being inlined anyway. How does this look? Kevin, could you please retest? I might have fat-fingered something... Your patch on top of Geert's still compiles fine for me with gcc-4.7.3. However, I'm not sure how specific we can be on the versions. /me goes to test a few more compilers... OK... ICE: 4.7.1, 4.7.3, 4.8.3 OK: 4.6.3, 4.9.2, 4.9.3 The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. I see my fix in your mmots since last Thurs (4/2), but it's not in mmotm (last updated today) so today's linux-next still has the ICE for anything other than gcc-4.7.3. Just checking to see when you plan to update mmotm. Thanks, Kevin -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
On Tue, 7 Apr 2015 10:57:52 -0700 Kevin Hilman khil...@kernel.org wrote: The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. I see my fix in your mmots since last Thurs (4/2), but it's not in mmotm (last updated today) so today's linux-next still has the ICE for anything other than gcc-4.7.3. Just checking to see when you plan to update mmotm. It should all be there today? -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
Andrew Morton a...@linux-foundation.org writes: On Tue, 7 Apr 2015 10:57:52 -0700 Kevin Hilman khil...@kernel.org wrote: The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. I see my fix in your mmots since last Thurs (4/2), but it's not in mmotm (last updated today) so today's linux-next still has the ICE for anything other than gcc-4.7.3. Just checking to see when you plan to update mmotm. It should all be there today? Nope. In mmotm, only the original patch plus your first fix is there: $ curl -sO http://www.ozlabs.org/~akpm/mmotm/broken-out.tar.gz $ tar -tavf broken-out.tar.gz |grep gcc-473 -rw-r- akpm/eng 1838 2015-04-01 14:41 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473.patch -rw-r- akpm/eng 1309 2015-04-01 14:41 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix.patch but in mmots, the additional ptch from me, plus another comment fixup from you are also there: $ curl -sO http://www.ozlabs.org/~akpm/mmots/broken-out.tar.gz $ tar -tavf broken-out.tar.gz |grep gcc-473 -rw-r- akpm/eng 1882 2015-04-06 16:24 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473.patch -rw-r- akpm/eng 1271 2015-04-06 16:24 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix.patch -rw-r- akpm/eng 1382 2015-04-06 16:24 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix-fix.patch -rw-r- akpm/eng968 2015-04-06 16:24 broken-out/mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix-fix-fix.patch Kevin -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
On Tue, 07 Apr 2015 15:41:32 -0700 Kevin Hilman khil...@kernel.org wrote: Andrew Morton a...@linux-foundation.org writes: On Tue, 7 Apr 2015 10:57:52 -0700 Kevin Hilman khil...@kernel.org wrote: The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. I see my fix in your mmots since last Thurs (4/2), but it's not in mmotm (last updated today) so today's linux-next still has the ICE for anything other than gcc-4.7.3. Just checking to see when you plan to update mmotm. It should all be there today? Nope. huh, I swear I did an mmotm yesterday. Let me see if I can sort out the watchdog mess and produce something releasable... -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
Andrew Morton a...@linux-foundation.org writes: On Tue, 07 Apr 2015 16:27:44 -0700 Kevin Hilman khil...@kernel.org wrote: It should all be there today? Nope. huh, I swear I did an mmotm yesterday. Well, based on the timestamp of the mmotm dir on ozlabs, it appears you did. That's why I was confused why the gcc-473 patches from mmots aren't there. Things look a bit better now. Yup, I can confirm all 4 patches are there now. Things should be in good shape for the next -next. Thanks, Kevin -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
Andrew Morton a...@linux-foundation.org writes: On Tue, 07 Apr 2015 15:41:32 -0700 Kevin Hilman khil...@kernel.org wrote: Andrew Morton a...@linux-foundation.org writes: On Tue, 7 Apr 2015 10:57:52 -0700 Kevin Hilman khil...@kernel.org wrote: The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. I see my fix in your mmots since last Thurs (4/2), but it's not in mmotm (last updated today) so today's linux-next still has the ICE for anything other than gcc-4.7.3. Just checking to see when you plan to update mmotm. It should all be there today? Nope. huh, I swear I did an mmotm yesterday. Well, based on the timestamp of the mmotm dir on ozlabs, it appears you did. That's why I was confused why the gcc-473 patches from mmots aren't there. Let me see if I can sort out the watchdog mess and produce something releasable... OK, thanks. Kevin -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
On Tue, 07 Apr 2015 16:27:44 -0700 Kevin Hilman khil...@kernel.org wrote: It should all be there today? Nope. huh, I swear I did an mmotm yesterday. Well, based on the timestamp of the mmotm dir on ozlabs, it appears you did. That's why I was confused why the gcc-473 patches from mmots aren't there. Things look a bit better now. -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
Hi Russell, On Wed, Apr 1, 2015 at 11:59 PM, Russell King - ARM Linux li...@arm.linux.org.uk wrote: On Wed, Apr 01, 2015 at 02:54:59PM -0700, Kevin Hilman wrote: Your patch on top of Geert's still compiles fine for me with gcc-4.7.3. However, I'm not sure how specific we can be on the versions. /me goes to test a few more compilers... OK... ICE: 4.7.1, 4.7.3, 4.8.3 OK: 4.6.3, 4.9.2, 4.9.3 The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. Interesting. I'm using stock gcc 4.7.4 here, though I'm not building -next (only mainline + my tree + arm-soc) and it hasn't shown a problem yet. Mainline doesn't fail. I think we need to ask the question: is the bug in stock GCC or Linaro GCC? If it's not in stock GCC, then it's a GCC vendor problem :) Can you please try -next (e.g. next-20150320)? make bockw_defconfig make mm/migrate.o Thanks! Gr{oetje,eeting}s, Geert -- Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@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 -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
On Thu, Apr 2, 2015 at 12:12 PM, Lina Iyer lina.i...@linaro.org wrote: On Wed, Apr 01 2015 at 15:57 -0600, Kevin Hilman wrote: Andrew Morton a...@linux-foundation.org writes: On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier marc.zyng...@arm.com wrote: -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page, - unsigned long private, struct page *page, int force, - enum migrate_mode mode) +static noinline int unmap_and_move(new_page_t get_new_page, +free_page_t put_new_page, +unsigned long private, struct page *page, +int force, enum migrate_mode mode) { int rc = 0; int *result = NULL; Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of the parameters on the stack, which is not going to help performance either (not that this would be useful on 32bit ARM anyway...). Any chance you could make this dependent on some compiler detection mechanism? With my arm compiler (gcc-4.4.4) the patch makes no difference - unmap_and_move() isn't being inlined anyway. How does this look? Kevin, could you please retest? I might have fat-fingered something... Your patch on top of Geert's still compiles fine for me with gcc-4.7.3. However, I'm not sure how specific we can be on the versions. /me goes to test a few more compilers... OK... ICE: 4.7.1, 4.7.3, 4.8.3 OK: 4.6.3, 4.9.2, 4.9.3 The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. I see ICE on arm-linux-gnueabi-gcc (Ubuntu/Linaro 4.7.4-2ubuntu1) 4.7.4 Thanks for checking. I'm assuming my patch fixes it for your since that should catch any 4.7.x compiler. Kevin -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
On Thu, Apr 02 2015 at 15:12 -0600, Kevin Hilman wrote: On Thu, Apr 2, 2015 at 12:12 PM, Lina Iyer lina.i...@linaro.org wrote: On Wed, Apr 01 2015 at 15:57 -0600, Kevin Hilman wrote: Andrew Morton a...@linux-foundation.org writes: On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier marc.zyng...@arm.com wrote: -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page, - unsigned long private, struct page *page, int force, - enum migrate_mode mode) +static noinline int unmap_and_move(new_page_t get_new_page, +free_page_t put_new_page, +unsigned long private, struct page *page, +int force, enum migrate_mode mode) { int rc = 0; int *result = NULL; Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of the parameters on the stack, which is not going to help performance either (not that this would be useful on 32bit ARM anyway...). Any chance you could make this dependent on some compiler detection mechanism? With my arm compiler (gcc-4.4.4) the patch makes no difference - unmap_and_move() isn't being inlined anyway. How does this look? Kevin, could you please retest? I might have fat-fingered something... Your patch on top of Geert's still compiles fine for me with gcc-4.7.3. However, I'm not sure how specific we can be on the versions. /me goes to test a few more compilers... OK... ICE: 4.7.1, 4.7.3, 4.8.3 OK: 4.6.3, 4.9.2, 4.9.3 The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. I see ICE on arm-linux-gnueabi-gcc (Ubuntu/Linaro 4.7.4-2ubuntu1) 4.7.4 Thanks for checking. I'm assuming my patch fixes it for your since that should catch any 4.7.x compiler. Yes, thank you. This fixes it on 4.7.4 Kevin -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
On Wed, Apr 01 2015 at 15:57 -0600, Kevin Hilman wrote: Andrew Morton a...@linux-foundation.org writes: On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier marc.zyng...@arm.com wrote: -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page, - unsigned long private, struct page *page, int force, - enum migrate_mode mode) +static noinline int unmap_and_move(new_page_t get_new_page, + free_page_t put_new_page, + unsigned long private, struct page *page, + int force, enum migrate_mode mode) { int rc = 0; int *result = NULL; Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of the parameters on the stack, which is not going to help performance either (not that this would be useful on 32bit ARM anyway...). Any chance you could make this dependent on some compiler detection mechanism? With my arm compiler (gcc-4.4.4) the patch makes no difference - unmap_and_move() isn't being inlined anyway. How does this look? Kevin, could you please retest? I might have fat-fingered something... Your patch on top of Geert's still compiles fine for me with gcc-4.7.3. However, I'm not sure how specific we can be on the versions. /me goes to test a few more compilers... OK... ICE: 4.7.1, 4.7.3, 4.8.3 OK: 4.6.3, 4.9.2, 4.9.3 The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. I see ICE on arm-linux-gnueabi-gcc (Ubuntu/Linaro 4.7.4-2ubuntu1) 4.7.4 Kevin [1] diff --git a/mm/migrate.c b/mm/migrate.c index 25fd7f6291de..6e15ae3248e0 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -901,10 +901,10 @@ out: } /* - * gcc-4.7.3 on arm gets an ICE when inlining unmap_and_move(). Work around + * gcc 4.7 and 4.8 on arm gets an ICE when inlining unmap_and_move(). Work around * it. */ -#if GCC_VERSION == 40703 defined(CONFIG_ARM) +#if (GCC_VERSION = 40700 GCC_VERSION 40900) defined(CONFIG_ARM) #define ICE_noinline noinline #else #define ICE_noinline ___ linux-arm-kernel mailing list linux-arm-ker...@lists.infradead.org http://lists.infradead.org/mailman/listinfo/linux-arm-kernel -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
On 01/04/15 10:37, Geert Uytterhoeven wrote: Hi Kevin, On Tue, 31 Mar 2015, Kevin Hilman wrote: Ard Biesheuvel ard.biesheu...@linaro.org writes: Nope, that branch is already part of linux-next, and linux-next still fails to compile for 20+ defconfigs[1] Could you elaborate on the issue please? What is the error you are getting, and can you confirm that is is caused by ld choking on the linker script? If not, this is another error than the one we have been trying to fix It's definitely not linker script related. Using arm-linux-gnueabi-gcc (Ubuntu/Linaro 4.7.3-12ubuntu1) 4.7.3, here's the error when building for multi_v7_defconfig (full log available[2]): ../mm/migrate.c: In function 'migrate_pages': ../mm/migrate.c:1148:1: internal compiler error: in push_minipool_fix, at config/arm/arm.c:13101 Please submit a full bug report, with preprocessed source if appropriate. See file:///usr/share/doc/gcc-4.7/README.Bugs for instructions. Preprocessed source stored into /tmp/ccO1Nz1m.out file, please attach this to your bugreport. make[2]: *** [mm/migrate.o] Error 1 make[2]: Target `__build' not remade because of errors. make[1]: *** [mm] Error 2 build bisect points to commit 21f992084aeb[3], but that doesn't revert cleanly so I haven't got any further than that yet. I installed gcc-arm-linux-gnueabi (4:4.7.2-1 from Ubuntu 14.04 LTS) and could reproduce the ICE. I came up with the workaround below. Does this work for you? From 7ebe83316eaf1952e55a76754ce7a5832e461b8c Mon Sep 17 00:00:00 2001 From: Geert Uytterhoeven geert+rene...@glider.be Date: Wed, 1 Apr 2015 11:22:51 +0200 Subject: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit With gcc version 4.7.3 (Ubuntu/Linaro 4.7.3-12ubuntu1) : mm/migrate.c: In function ‘migrate_pages’: mm/migrate.c:1148:1: internal compiler error: in push_minipool_fix, at config/arm/arm.c:13500 Please submit a full bug report, with preprocessed source if appropriate. See file:///usr/share/doc/gcc-4.7/README.Bugs for instructions. Preprocessed source stored into /tmp/ccPoM1tr.out file, please attach this to your bugreport. make[1]: *** [mm/migrate.o] Error 1 make: *** [mm/migrate.o] Error 2 Mark unmap_and_move() (which is used in a single place only) noinline to work around this compiler bug. Reported-by: Kevin Hilman khil...@kernel.org Signed-off-by: Geert Uytterhoeven geert+rene...@glider.be --- mm/migrate.c | 7 --- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/mm/migrate.c b/mm/migrate.c index 114602a68111d809..98f8574456c2010c 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -904,9 +904,10 @@ out: * Obtain the lock on page, remove all ptes and migrate the page * to the newly allocated page in newpage. */ -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page, - unsigned long private, struct page *page, int force, - enum migrate_mode mode) +static noinline int unmap_and_move(new_page_t get_new_page, +free_page_t put_new_page, +unsigned long private, struct page *page, +int force, enum migrate_mode mode) { int rc = 0; int *result = NULL; Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of the parameters on the stack, which is not going to help performance either (not that this would be useful on 32bit ARM anyway...). Any chance you could make this dependent on some compiler detection mechanism? Thanks, M. -- Jazz is not dead. It just smells funny... -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier marc.zyng...@arm.com wrote: -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page, - unsigned long private, struct page *page, int force, - enum migrate_mode mode) +static noinline int unmap_and_move(new_page_t get_new_page, + free_page_t put_new_page, + unsigned long private, struct page *page, + int force, enum migrate_mode mode) { int rc = 0; int *result = NULL; Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of the parameters on the stack, which is not going to help performance either (not that this would be useful on 32bit ARM anyway...). Any chance you could make this dependent on some compiler detection mechanism? With my arm compiler (gcc-4.4.4) the patch makes no difference - unmap_and_move() isn't being inlined anyway. How does this look? Kevin, could you please retest? I might have fat-fingered something... --- a/mm/migrate.c~mm-migrate-mark-unmap_and_move-noinline-to-avoid-ice-in-gcc-473-fix +++ a/mm/migrate.c @@ -901,10 +901,20 @@ out: } /* + * gcc-4.7.3 on arm gets an ICE when inlining unmap_and_move(). Work around + * it. + */ +#if GCC_VERSION == 40703 defined(CONFIG_ARM) +#define ICE_noinline noinline +#else +#define ICE_noinline +#endif + +/* * Obtain the lock on page, remove all ptes and migrate the page * to the newly allocated page in newpage. */ -static noinline int unmap_and_move(new_page_t get_new_page, +static ICE_noinline int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page, unsigned long private, struct page *page, int force, enum migrate_mode mode) _ -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
Geert Uytterhoeven ge...@linux-m68k.org writes: [...] build bisect points to commit 21f992084aeb[3], but that doesn't revert cleanly so I haven't got any further than that yet. I installed gcc-arm-linux-gnueabi (4:4.7.2-1 from Ubuntu 14.04 LTS) and could reproduce the ICE. I came up with the workaround below. Awesome, thanks! Does this work for you? Yes, that patch works well and fixes the regression. Build results for all the defconfigs here: http://kernelci.org/build/khilman/kernel/v4.0-rc6-8294-g2ef3958cc27e/ and the remaining issues arent' realted to this ICE. From 7ebe83316eaf1952e55a76754ce7a5832e461b8c Mon Sep 17 00:00:00 2001 From: Geert Uytterhoeven geert+rene...@glider.be Date: Wed, 1 Apr 2015 11:22:51 +0200 Subject: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3 MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit With gcc version 4.7.3 (Ubuntu/Linaro 4.7.3-12ubuntu1) : mm/migrate.c: In function ‘migrate_pages’: mm/migrate.c:1148:1: internal compiler error: in push_minipool_fix, at config/arm/arm.c:13500 Please submit a full bug report, with preprocessed source if appropriate. See file:///usr/share/doc/gcc-4.7/README.Bugs for instructions. Preprocessed source stored into /tmp/ccPoM1tr.out file, please attach this to your bugreport. make[1]: *** [mm/migrate.o] Error 1 make: *** [mm/migrate.o] Error 2 Mark unmap_and_move() (which is used in a single place only) noinline to work around this compiler bug. Reported-by: Kevin Hilman khil...@kernel.org Signed-off-by: Geert Uytterhoeven geert+rene...@glider.be Tested-by: Kevin Hilman khil...@linaro.org --- mm/migrate.c | 7 --- 1 file changed, 4 insertions(+), 3 deletions(-) diff --git a/mm/migrate.c b/mm/migrate.c index 114602a68111d809..98f8574456c2010c 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -904,9 +904,10 @@ out: * Obtain the lock on page, remove all ptes and migrate the page * to the newly allocated page in newpage. */ -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page, - unsigned long private, struct page *page, int force, - enum migrate_mode mode) +static noinline int unmap_and_move(new_page_t get_new_page, +free_page_t put_new_page, +unsigned long private, struct page *page, +int force, enum migrate_mode mode) { int rc = 0; int *result = NULL; -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
Andrew Morton a...@linux-foundation.org writes: On Wed, 01 Apr 2015 10:47:49 +0100 Marc Zyngier marc.zyng...@arm.com wrote: -static int unmap_and_move(new_page_t get_new_page, free_page_t put_new_page, - unsigned long private, struct page *page, int force, - enum migrate_mode mode) +static noinline int unmap_and_move(new_page_t get_new_page, + free_page_t put_new_page, + unsigned long private, struct page *page, + int force, enum migrate_mode mode) { int rc = 0; int *result = NULL; Ouch. That's really ugly. And on 32bit ARM, we end-up spilling half of the parameters on the stack, which is not going to help performance either (not that this would be useful on 32bit ARM anyway...). Any chance you could make this dependent on some compiler detection mechanism? With my arm compiler (gcc-4.4.4) the patch makes no difference - unmap_and_move() isn't being inlined anyway. How does this look? Kevin, could you please retest? I might have fat-fingered something... Your patch on top of Geert's still compiles fine for me with gcc-4.7.3. However, I'm not sure how specific we can be on the versions. /me goes to test a few more compilers... OK... ICE: 4.7.1, 4.7.3, 4.8.3 OK: 4.6.3, 4.9.2, 4.9.3 The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. Kevin [1] diff --git a/mm/migrate.c b/mm/migrate.c index 25fd7f6291de..6e15ae3248e0 100644 --- a/mm/migrate.c +++ b/mm/migrate.c @@ -901,10 +901,10 @@ out: } /* - * gcc-4.7.3 on arm gets an ICE when inlining unmap_and_move(). Work around + * gcc 4.7 and 4.8 on arm gets an ICE when inlining unmap_and_move(). Work around * it. */ -#if GCC_VERSION == 40703 defined(CONFIG_ARM) +#if (GCC_VERSION = 40700 GCC_VERSION 40900) defined(CONFIG_ARM) #define ICE_noinline noinline #else #define ICE_noinline -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [PATCH] mm/migrate: Mark unmap_and_move() noinline to avoid ICE in gcc 4.7.3
On Wed, Apr 01, 2015 at 02:54:59PM -0700, Kevin Hilman wrote: Your patch on top of Geert's still compiles fine for me with gcc-4.7.3. However, I'm not sure how specific we can be on the versions. /me goes to test a few more compilers... OK... ICE: 4.7.1, 4.7.3, 4.8.3 OK: 4.6.3, 4.9.2, 4.9.3 The diff below[2] on top of yours compiles fine here and at least covers the compilers I *know* to trigger the ICE. Interesting. I'm using stock gcc 4.7.4 here, though I'm not building -next (only mainline + my tree + arm-soc) and it hasn't shown a problem yet. I think we need to ask the question: is the bug in stock GCC or Linaro GCC? If it's not in stock GCC, then it's a GCC vendor problem :) -- FTTC broadband for 0.8mile line: currently at 10.5Mbps down 400kbps up according to speedtest.net. -- To unsubscribe from this list: send the line unsubscribe linux-omap in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html