| Message ID | 1460948988-7477-2-git-send-email-wangkefeng.wang@huawei.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
On 18/04/16 04:09, Kefeng Wang wrote: > Each line with single pr_cont() in Virtual kernel memory layout, > or the dump of the kernel memory layout in dmesg is not aligned > when PRINTK_TIME enabled, due to the missing time stamps. Looks good to me, but I can't reproduce the miss-aligned output on v4.6-rc4, I get: > [ 0.000000] .text : 0xffffff8008080000 - 0xffffff8008887000 ( 8220 KB) > [ 0.000000] .rodata : 0xffffff8008887000 - 0xffffff8008bdc000 ( 3412 KB) > [ 0.000000] .init : 0xffffff8008bdc000 - 0xffffff8008cc3000 ( 924 KB) > [ 0.000000] .data : 0xffffff8008cc3000 - 0xffffff8008dd7200 ( 1105 KB) without this patch, but it evidently happens for you. Do you have any special build options or kernel version? I'd like to understand if there is some kconfig/earlycon option that affects this! Thanks, James
On 18 April 2016 at 13:16, James Morse <james.morse@arm.com> wrote: > On 18/04/16 04:09, Kefeng Wang wrote: >> Each line with single pr_cont() in Virtual kernel memory layout, >> or the dump of the kernel memory layout in dmesg is not aligned >> when PRINTK_TIME enabled, due to the missing time stamps. > > Looks good to me, but I can't reproduce the miss-aligned output on v4.6-rc4, I get: >> [ 0.000000] .text : 0xffffff8008080000 - 0xffffff8008887000 ( 8220 KB) >> [ 0.000000] .rodata : 0xffffff8008887000 - 0xffffff8008bdc000 ( 3412 KB) >> [ 0.000000] .init : 0xffffff8008bdc000 - 0xffffff8008cc3000 ( 924 KB) >> [ 0.000000] .data : 0xffffff8008cc3000 - 0xffffff8008dd7200 ( 1105 KB) > > without this patch, but it evidently happens for you. Do you have any special > build options or kernel version? I'd like to understand if there is some > kconfig/earlycon option that affects this! > Interestingly, for me it only occurs when running dmesg, and not when observing the output from the console directly.
Hi Ard, On 18/04/16 12:21, Ard Biesheuvel wrote: > On 18 April 2016 at 13:16, James Morse <james.morse@arm.com> wrote: >> On 18/04/16 04:09, Kefeng Wang wrote: >>> Each line with single pr_cont() in Virtual kernel memory layout, >>> or the dump of the kernel memory layout in dmesg is not aligned >>> when PRINTK_TIME enabled, due to the missing time stamps. >> >> Looks good to me, but I can't reproduce the miss-aligned output on v4.6-rc4, I get: >>> [ 0.000000] .text : 0xffffff8008080000 - 0xffffff8008887000 ( 8220 KB) >>> [ 0.000000] .rodata : 0xffffff8008887000 - 0xffffff8008bdc000 ( 3412 KB) >>> [ 0.000000] .init : 0xffffff8008bdc000 - 0xffffff8008cc3000 ( 924 KB) >>> [ 0.000000] .data : 0xffffff8008cc3000 - 0xffffff8008dd7200 ( 1105 KB) >> >> without this patch, but it evidently happens for you. Do you have any special >> build options or kernel version? I'd like to understand if there is some >> kconfig/earlycon option that affects this! >> > > Interestingly, for me it only occurs when running dmesg, and not when > observing the output from the console directly. Hmmm, after some more digging: dmesg on Ubuntu 15.04 miss-aligns this, but Ubuntu 14.04 doesn't.... Regardless, the patch fixes it, (and it looks the same on both ubuntu versions): Tested-by: James Morse <james.morse@arm.com> From the comment above 'pr_cont()' in include/linux/printk.h: > * Like KERN_CONT, pr_cont() should only be used when continuing > * a line with no newline ('\n') enclosed. Otherwise it defaults > * back to KERN_DEFAULT. arm64 is probably confusing dmesg with its use of pr_cont("...\n"). Thanks, James 15.04: util-linux 2.25.2-4ubuntu3 14.04: util-linux 2.20.1-5.1ubuntu20
On 2016/4/18 21:13, James Morse wrote: > Hi Ard, > > On 18/04/16 12:21, Ard Biesheuvel wrote: >> On 18 April 2016 at 13:16, James Morse <james.morse@arm.com> wrote: >>> On 18/04/16 04:09, Kefeng Wang wrote: >>>> Each line with single pr_cont() in Virtual kernel memory layout, >>>> or the dump of the kernel memory layout in dmesg is not aligned >>>> when PRINTK_TIME enabled, due to the missing time stamps. >>> >>> Looks good to me, but I can't reproduce the miss-aligned output on v4.6-rc4, I get: >>>> [ 0.000000] .text : 0xffffff8008080000 - 0xffffff8008887000 ( 8220 KB) >>>> [ 0.000000] .rodata : 0xffffff8008887000 - 0xffffff8008bdc000 ( 3412 KB) >>>> [ 0.000000] .init : 0xffffff8008bdc000 - 0xffffff8008cc3000 ( 924 KB) >>>> [ 0.000000] .data : 0xffffff8008cc3000 - 0xffffff8008dd7200 ( 1105 KB) >>> >>> without this patch, but it evidently happens for you. Do you have any special >>> build options or kernel version? I'd like to understand if there is some >>> kconfig/earlycon option that affects this! >>> >> >> Interestingly, for me it only occurs when running dmesg, and not when >> observing the output from the console directly. > > Hmmm, after some more digging: dmesg on Ubuntu 15.04 miss-aligns this, but > Ubuntu 14.04 doesn't.... > > Regardless, the patch fixes it, (and it looks the same on both ubuntu versions): > Tested-by: James Morse <james.morse@arm.com> > Thanks :) > >>From the comment above 'pr_cont()' in include/linux/printk.h: >> * Like KERN_CONT, pr_cont() should only be used when continuing >> * a line with no newline ('\n') enclosed. Otherwise it defaults >> * back to KERN_DEFAULT. > > arm64 is probably confusing dmesg with its use of pr_cont("...\n"). After check the util-linux, find this commit(b45c3da2e1581 tags/v2.26-rc2~86) commit b45c3da2e15814ca0c5b2ed2608de34df60af4da Author: Ingo Br<C3><BC>ckl <ib@wupperonline.de> Date: Wed Jan 14 19:36:55 2015 +0100 dmesg: Enable proper indent for messages containing line breaks Messages containing line breaks somehow look broken when additional time, facility or level information is displayed, because they partly appear in time/facility/level column. Indent them accordingly. Signed-off-by: Karel Zak <kzak@redhat.com> After with this patch, the dmesg shows below, [ 0.000000] Virtual kernel memory layout: [ 0.000000] modules : 0xffffff8000000000 - 0xffffff8008000000 ( 128 MB) [ 0.000000] vmalloc : 0xffffff8008000000 - 0xffffffbdbfff0000 ( 246 GB) [ 0.000000] .text : 0xffffff8008080000 - 0xffffff8008755000 ( 6996 KB) .rodata : 0xffffff8008755000 - 0xffffff8008a3d000 ( 2976 KB) .init : 0xffffff8008a3d000 - 0xffffff8008b09000 ( 816 KB) .data : 0xffffff8008b09000 - 0xffffff8008ba6800 ( 630 KB) [ 0.000000] vmemmap : 0xffffffbdc0000000 - 0xffffffbfc0000000 ( 8 GB maximum) 0xffffffbdc0000000 - 0xffffffbdc8000000 ( 128 MB actual) [ 0.000000] fixed : 0xffffffbffe7fd000 - 0xffffffbffec00000 ( 4108 KB) [ 0.000000] PCI I/O : 0xffffffbffee00000 - 0xffffffbfffe00000 ( 16 MB) [ 0.000000] memory : 0xffffffc000000000 - 0xffffffc200000000 ( 8192 MB) BRs, Kefeng > > > Thanks, > > James > > > 15.04: util-linux 2.25.2-4ubuntu3 > 14.04: util-linux 2.20.1-5.1ubuntu20 > > . >
diff --git a/arch/arm64/mm/init.c b/arch/arm64/mm/init.c index ea989d8..ce2e400 100644 --- a/arch/arm64/mm/init.c +++ b/arch/arm64/mm/init.c @@ -371,19 +371,18 @@ void __init mem_init(void) MLM(MODULES_VADDR, MODULES_END)); pr_cont(" vmalloc : 0x%16lx - 0x%16lx (%6ld GB)\n", MLG(VMALLOC_START, VMALLOC_END)); - pr_cont(" .text : 0x%p" " - 0x%p" " (%6ld KB)\n" - " .rodata : 0x%p" " - 0x%p" " (%6ld KB)\n" - " .init : 0x%p" " - 0x%p" " (%6ld KB)\n" - " .data : 0x%p" " - 0x%p" " (%6ld KB)\n", - MLK_ROUNDUP(_text, __start_rodata), - MLK_ROUNDUP(__start_rodata, _etext), - MLK_ROUNDUP(__init_begin, __init_end), + pr_cont(" .text : 0x%p" " - 0x%p" " (%6ld KB)\n", + MLK_ROUNDUP(_text, __start_rodata)); + pr_cont(" .rodata : 0x%p" " - 0x%p" " (%6ld KB)\n", + MLK_ROUNDUP(__start_rodata, _etext)); + pr_cont(" .init : 0x%p" " - 0x%p" " (%6ld KB)\n", + MLK_ROUNDUP(__init_begin, __init_end)); + pr_cont(" .data : 0x%p" " - 0x%p" " (%6ld KB)\n", MLK_ROUNDUP(_sdata, _edata)); #ifdef CONFIG_SPARSEMEM_VMEMMAP - pr_cont(" vmemmap : 0x%16lx - 0x%16lx (%6ld GB maximum)\n" - " 0x%16lx - 0x%16lx (%6ld MB actual)\n", - MLG(VMEMMAP_START, - VMEMMAP_START + VMEMMAP_SIZE), + pr_cont(" vmemmap : 0x%16lx - 0x%16lx (%6ld GB maximum)\n", + MLG(VMEMMAP_START, VMEMMAP_START + VMEMMAP_SIZE)); + pr_cont(" 0x%16lx - 0x%16lx (%6ld MB actual)\n", MLM((unsigned long)phys_to_page(memblock_start_of_DRAM()), (unsigned long)virt_to_page(high_memory))); #endif
Each line with single pr_cont() in Virtual kernel memory layout, or the dump of the kernel memory layout in dmesg is not aligned when PRINTK_TIME enabled, due to the missing time stamps. Signed-off-by: Kefeng Wang <wangkefeng.wang@huawei.com> --- arch/arm64/mm/init.c | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-)