From: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
To: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Shuah Khan <shuahkhan@gmail.com>, Yinghai Lu <yinghai@kernel.org>,
Thomas Gleixner <tglx@linutronix.de>, Ingo Molnar <mingo@elte.hu>,
"H. Peter Anvin" <hpa@zytor.com>,
Andrew Morton <akpm@linux-foundation.org>,
Borislav Petkov <bp@alien8.de>, Jan Kiszka <jan.kiszka@web.de>,
Jason Wessel <jason.wessel@windriver.com>,
linux-kernel@vger.kernel.org, Joerg Roedel <joro@8bytes.org>
Subject: Re: [PATCH v7u1 26/31] x86: Don't enable swiotlb if there is not enough ram for it
Date: Mon, 7 Jan 2013 21:48:30 -0500 [thread overview]
Message-ID: <20130108024830.GE9865@phenom.dumpdata.com> (raw)
In-Reply-To: <8762382z2c.fsf@xmission.com>
On Mon, Jan 07, 2013 at 06:22:51PM -0800, Eric W. Biederman wrote:
> Shuah Khan <shuahkhan@gmail.com> writes:
>
> > On Mon, Jan 7, 2013 at 8:26 AM, Konrad Rzeszutek Wilk
> > <konrad.wilk@oracle.com> wrote:
> >> On Fri, Jan 04, 2013 at 02:10:25PM -0800, Yinghai Lu wrote:
> >>> On Fri, Jan 4, 2013 at 1:02 PM, Shuah Khan <shuahkhan@gmail.com> wrote:
> >>> > Pani'cing the system doesn't sound like a good option to me in this
> >>> > case. This change to disable swiotlb is made for kdump. However, with
> >>> > this change several system fail to boot, unless crashkernel_low=72M is
> >>> > specified.
> >>>
> >>> this patchset is new feature to put second kdump kernel above 4G.
> >>>
> >>> >
> >>> > I would the say the right approach to solve this would be to not
> >>> > change the current pci_swiotlb_detect_override() behavior and treat
> >>> > swiotlb =1 upon entry equivalent to swiotlb_force set.
> >>>
> >>> that will make intel system have to take crashkernel_low=72M too.
> >>> otherwise intel system will get panic during swiotlb allocation.
> >>
> >> Two things:
> >>
> >> 1). You need to wrap the 'is_enough_..' in CONFIG_KEXEC, which means
> >> that the function needs to go in a header file.
> >> 2). The check for 1MB is suspect. Why only 1MB? You mentioned it is
> >> b/c of crashkernel_low=72M (which I am not seeing in v3.8 kernel-parameters.txt?
> >> Is that part of your mega-patchset?). Anyhow, there seems to be a disconnect -
> >> what if the user supplied crashkernel_low=27M? Perhaps the 'is_enough'
> >> should also parse the bootparams to double-check that there is enough
> >> low-mem space? But then if the kernel grows then 72M might not be enough -
> >> you might need 82M with 3.9.
> >>
> >> Perhaps a better way for this is to do:
> >> 1). Change 'is_enough' to check only for 4MB.
> >> 2). When booting as kexec, the SWIOTLB would only use 4MB instead of 64MB?
> >>
> >> Or, we could also use the post-late SWIOTLB initialization similiary to how it was
> >> done on ia64. This would mean that the AMD VI code would just call the
> >> .. something like this - NOT tested or even compile tested:
> >>
> >> diff --git a/drivers/iommu/amd_iommu.c b/drivers/iommu/amd_iommu.c
> >> index c1c74e0..e7fa8f7 100644
> >> --- a/drivers/iommu/amd_iommu.c
> >> +++ b/drivers/iommu/amd_iommu.c
> >> @@ -3173,6 +3173,24 @@ int __init amd_iommu_init_dma_ops(void)
> >> if (unhandled && max_pfn > MAX_DMA32_PFN) {
> >> /* There are unhandled devices - initialize swiotlb for them */
> >> swiotlb = 1;
> >> + /* Late (so no bootmem allocator) usage and only if the early SWIOTLB
> >> + * hadn't been allocated (which can happen on kexec kernels booted
> >> + * above 4GB). */
> >> + if (!swiotlb_nr_tbl()) {
> >> + int retry = 3;
> >> + int mb_size = 64;
> >> + int rc = 0;
> >> +retry_me:
> >> + if (retry < 0)
> >> + panic("We tried setting %dMB for SWIOTLB but got -ENOMEM", mb_size << 1);
> >> + rc = swiotlb_late_init_with_default_size(mb_size * (1<<20));
> >> + if (rc) {
> >> + retry --;
> >> + mb_size >> 1;
> >> + goto retry_me;
> >> + }
> >> + dma_ops = &swiotlb_dma_ops;
> >> + }
> >> }
> >>
> >> amd_iommu_stats_init();
> >>
> >> And then the early SWIOTLB initialization for 64MB can fail and we are still OK.
> >>>
> >
> > Yinghai/Konrad,
> >
> > Did more testing. btw this patch depends on your [v7u1,25/31]
> > memblock: add memblock_mem_size(). Here are the test results:
> >
> > 1. When there is not enough memory: (enough_mem_for_swiotlb() returns false)
> > system will panic in amd_iommu_init_dma_ops().
> >
> > 2. When there is enough memory: (enough_mem_for_swiotlb() returns true):
> > swiotlb is reserved
> > pci_swiotlb_late_init() leaves the buffer allocated since swiotlb=1
> > with that getting changed in amd_iommu_init_dma_ops().
> >
> > I agree with Konrad that the logic should be wrapped in CONFIG_KEXEC.
>
> If enough_mem_for_swiotlb needs to be conditional on CONFIG_KEXEC the
> code is architected wrong. None of this logic has anything to do with
> kexec except that the kexec path is one way to get this condition to
> happen. Especially since the kexec'd kernel where this condition occurs
> does not need kexec support built in.
Fair enough - with the 'memmap' command line options one can trigger
this.
>
> Yinghai I sat down and read your patch and the approach you are taking
> is totally wrong.
>
> The problem is that swiotlb_init() in lib/swiotlb.c does not know how to
> fail without panic'ing the system.
>
> Which leaves two valid approaches.
> - Create a variant of swiotlb_init that can fail for use on x86 and
> handle the failure.
As an safe-fail step we could retry with an smaller size until a fit is found.
> - Delay initializing the swiotlb until someone actually needs a mapping
> from it.
So late init the SWIOTLB and perhaps have multiple "segments" of 4MB
of SWIOTLB that can grow as we exhaust its memory. Could work.
>
> Delaying the initialization of the swiotlb is out because the code
> needs an early memory allocation to get a large chunk of contiguous
> memory for the bounce buffers.
Or it can use the late init, but with a smaller chunk of memory.
>
> Which means the panics that occurr in swiotlb_init() need to be delayed
> until someone something actually needs bounce buffers from the swiotlb.
>
> Although arguably what should actually happen instead of panic() is that
> swiotlb_map_single should simply fail early when it was not possible to
> preallocate bounce buffers.
This sounds like a Catch-22. Fail early implies that it would have to do
this when using the bootmem allocator. But the swiotlb_map_single is not
called at that time - it is called _after_ the bootmem allocator has been
de-activated. Actually it is called pretty late - when built-in PCI devices
start off or when 'udev' starts scanning the PCI bus and loading modules.
I think I am misunderstanding you - could you clarify please?
next prev parent reply other threads:[~2013-01-08 2:49 UTC|newest]
Thread overview: 199+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-01-04 0:48 [PATCH v7u1 00/31] x86, boot, 64bit: Add support for loading ramdisk and bzImage above 4G Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 01/31] x86, mm: Fix page table early allocation offset checking Yinghai Lu
2013-01-04 7:17 ` Borislav Petkov
2013-01-04 21:50 ` Yinghai Lu
2013-01-05 13:05 ` Borislav Petkov
2013-01-15 12:27 ` Stefano Stabellini
2013-01-04 0:48 ` [PATCH v7u1 02/31] x86, 64bit, mm: make pgd next calculation consistent with pud/pmd Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 03/31] x86, realmode: set real_mode permissions early Yinghai Lu
2013-01-04 20:15 ` Borislav Petkov
2013-01-04 20:58 ` Yinghai Lu
2013-01-04 21:04 ` Borislav Petkov
2013-01-04 22:13 ` Yinghai Lu
2013-01-05 13:25 ` Borislav Petkov
2013-01-07 12:40 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 04/31] x86, 64bit, mm: add generic kernel/ident mapping helper Yinghai Lu
2013-01-04 21:19 ` Borislav Petkov
2013-01-04 22:19 ` Yinghai Lu
2013-01-05 13:21 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 05/31] x86, 64bit: copy zero-page early Yinghai Lu
2013-01-07 15:53 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 06/31] x86, 64bit, realmode: use init_level4_pgt to set trapmoline_pgt directly Yinghai Lu
2013-01-04 17:18 ` Sakkinen, Jarkko
2013-01-04 22:01 ` Yinghai Lu
2013-01-05 9:59 ` Sakkinen, Jarkko
2013-01-07 15:54 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 07/31] x86, realmode: Separate real_mode reserve and setup Yinghai Lu
2013-01-04 17:18 ` Sakkinen, Jarkko
2013-01-07 15:54 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 08/31] x86, 64bit: early #PF handler set page table Yinghai Lu
2013-01-07 15:55 ` Borislav Petkov
2013-01-10 1:56 ` Yinghai Lu
2013-01-10 12:19 ` Borislav Petkov
2013-01-10 17:05 ` Yinghai Lu
2013-01-10 20:27 ` Borislav Petkov
2013-01-12 22:04 ` H. Peter Anvin
2013-01-04 0:48 ` [PATCH v7u1 09/31] x86, 64bit: #PF handler set page to cover 2M only Yinghai Lu
2013-01-09 22:57 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 10/31] x86, 64bit: Don't set max_pfn_mapped wrong value early on native path Yinghai Lu
2013-01-11 12:13 ` Borislav Petkov
2013-01-11 16:42 ` Yinghai Lu
2013-01-11 16:52 ` Borislav Petkov
2013-01-15 13:48 ` Stefano Stabellini
2013-01-15 15:22 ` Konrad Rzeszutek Wilk
2013-01-15 15:59 ` Stefano Stabellini
2013-01-15 16:37 ` Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 11/31] x86: Merge early_reserve_initrd for 32bit and 64bit Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 12/31] x86: add get_ramdisk_image/size() Yinghai Lu
2013-01-07 15:56 ` Borislav Petkov
2013-01-10 1:53 ` Yinghai Lu
2013-01-10 12:13 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 13/31] x86, boot: add get_cmd_line_ptr() Yinghai Lu
2013-01-07 15:56 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 14/31] x86, boot: move checking of cmd_line_ptr out of common path Yinghai Lu
2013-01-07 16:00 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 15/31] x86, boot: pass cmd_line_ptr with unsigned long instead Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 16/31] x86, boot: move verify_cpu.S and no_longmode down Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 17/31] x86, boot: Move lldt/ltr out of 64bit code section Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 18/31] x86, kexec: remove 1024G limitation for kexec buffer on 64bit Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 19/31] x86, kexec: set ident mapping for kernel that is above max_pfn Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 20/31] x86, kexec: replace ident_mapping_init and init_level4_page Yinghai Lu
2013-01-04 21:01 ` Borislav Petkov
2013-01-04 22:04 ` Yinghai Lu
2013-01-05 13:24 ` Borislav Petkov
2013-01-10 1:26 ` Yinghai Lu
2013-01-10 11:59 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 21/31] x86, kexec: only set ident mapping for ram Yinghai Lu
2013-01-13 12:56 ` Borislav Petkov
2013-01-14 5:46 ` Yinghai Lu
2013-01-14 9:53 ` Borislav Petkov
2013-01-14 18:17 ` Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 22/31] x86, boot: add fields to support load bzImage and ramdisk above 4G Yinghai Lu
2013-01-13 21:41 ` Borislav Petkov
2013-01-14 5:37 ` Yinghai Lu
2013-01-14 9:43 ` Borislav Petkov
2013-01-14 23:06 ` Yinghai Lu
2013-01-14 17:49 ` H. Peter Anvin
2013-01-14 18:57 ` Yinghai Lu
2013-01-14 18:59 ` H. Peter Anvin
2013-01-14 19:19 ` Yinghai Lu
2013-01-14 19:50 ` Yinghai Lu
2013-01-14 19:56 ` H. Peter Anvin
2013-01-14 20:05 ` Yinghai Lu
2013-01-15 6:17 ` Yinghai Lu
2013-01-15 15:50 ` Borislav Petkov
2013-01-15 16:03 ` Yinghai Lu
2013-01-15 16:48 ` Borislav Petkov
2013-01-15 18:43 ` Yinghai Lu
2013-01-15 19:49 ` Borislav Petkov
2013-01-15 20:16 ` Yinghai Lu
2013-01-15 20:28 ` Borislav Petkov
2013-01-14 20:05 ` Borislav Petkov
2013-01-14 20:14 ` Yinghai Lu
2013-01-14 20:26 ` Borislav Petkov
2013-01-14 22:38 ` Yinghai Lu
2013-01-14 23:11 ` Borislav Petkov
2013-01-15 1:04 ` Yinghai Lu
2013-01-14 23:10 ` H. Peter Anvin
2013-01-14 23:21 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 23/31] x86, boot: update comments about entries for 64bit image Yinghai Lu
2013-01-14 11:20 ` Borislav Petkov
2013-01-14 18:35 ` Yinghai Lu
2013-01-14 18:37 ` Yinghai Lu
2013-01-14 18:46 ` Borislav Petkov
2013-01-14 20:01 ` Yinghai Lu
2013-01-14 18:43 ` Borislav Petkov
2013-01-04 0:48 ` [PATCH v7u1 24/31] x86, boot: Not need to check setup_header version for setup_data Yinghai Lu
2013-01-14 11:26 ` Borislav Petkov
2013-01-14 17:37 ` H. Peter Anvin
2013-01-14 18:04 ` Borislav Petkov
2013-01-14 18:42 ` H. Peter Anvin
2013-01-04 0:48 ` [PATCH v7u1 25/31] memblock: add memblock_mem_size() Yinghai Lu
2013-01-14 20:42 ` H. Peter Anvin
2013-01-14 22:28 ` Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 26/31] x86: Don't enable swiotlb if there is not enough ram for it Yinghai Lu
2013-01-04 16:05 ` Konrad Rzeszutek Wilk
2013-01-04 19:57 ` Yinghai Lu
2013-01-04 17:50 ` Shuah Khan
2013-01-04 20:34 ` Yinghai Lu
2013-01-04 21:02 ` Shuah Khan
2013-01-04 22:10 ` Yinghai Lu
2013-01-04 22:26 ` Shuah Khan
2013-01-04 22:34 ` Yinghai Lu
2013-01-04 22:47 ` Eric W. Biederman
2013-01-04 22:56 ` Shuah Khan
2013-01-04 23:00 ` Yinghai Lu
2013-01-04 23:21 ` Shuah Khan
2013-01-04 23:55 ` Yinghai Lu
2013-01-05 2:02 ` Shuah Khan
2013-01-05 4:10 ` Yinghai Lu
2013-01-05 22:04 ` Shuah Khan
2013-01-04 22:58 ` Yinghai Lu
2013-01-07 15:26 ` Konrad Rzeszutek Wilk
2013-01-07 17:02 ` Shuah Khan
2013-01-07 19:29 ` Konrad Rzeszutek Wilk
2013-01-08 2:22 ` Eric W. Biederman
2013-01-08 2:48 ` Konrad Rzeszutek Wilk [this message]
2013-01-08 3:03 ` Eric W. Biederman
2013-01-08 3:01 ` Yinghai Lu
2013-01-08 3:13 ` Eric W. Biederman
2013-01-08 3:50 ` Yinghai Lu
2013-01-08 23:40 ` Yinghai Lu
2013-01-09 0:04 ` Eric W. Biederman
2013-01-09 0:43 ` Konrad Rzeszutek Wilk
2013-01-09 0:56 ` Yinghai Lu
2013-01-09 0:58 ` Eric W. Biederman
2013-01-09 1:07 ` Yinghai Lu
2013-01-09 1:12 ` Yinghai Lu
2013-01-09 2:31 ` Eric W. Biederman
2013-01-09 13:24 ` Konrad Rzeszutek Wilk
2013-01-09 17:27 ` Yinghai Lu
2013-01-09 18:01 ` Shuah Khan
2013-01-09 19:13 ` Yinghai Lu
2013-01-09 21:00 ` Eric W. Biederman
2013-01-09 21:15 ` Yinghai Lu
2013-01-10 23:07 ` Yinghai Lu
2013-01-10 23:15 ` Eric W. Biederman
2013-01-10 23:55 ` Yinghai Lu
2013-01-11 16:35 ` Konrad Rzeszutek Wilk
2013-01-11 16:52 ` Yinghai Lu
2013-01-11 17:49 ` Yinghai Lu
2013-01-15 6:19 ` Yinghai Lu
2013-01-18 15:55 ` Konrad Rzeszutek Wilk
2013-01-24 15:39 ` Konrad Rzeszutek Wilk
2013-01-24 16:51 ` Shuah Khan
2013-01-24 19:22 ` Shuah Khan
2013-01-24 21:50 ` Yinghai Lu
2013-01-29 2:27 ` Shuah Khan
2013-01-29 3:44 ` Yinghai Lu
2013-01-31 19:28 ` Shuah Khan
2013-01-31 19:35 ` H. Peter Anvin
2013-01-09 13:12 ` Konrad Rzeszutek Wilk
2013-01-07 20:32 ` Yinghai Lu
2013-01-07 21:30 ` Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 27/31] x86, kdump: remove crashkernel range find limit for 64bit Yinghai Lu
2013-01-14 15:43 ` Borislav Petkov
2013-01-14 18:18 ` Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 28/31] x86: add Crash kernel low reservation Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 29/31] x86: Merge early kernel reserve for 32bit and 64bit Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 30/31] x86, 64bit, mm: Mark data/bss/brk to nx Yinghai Lu
2013-01-04 0:48 ` [PATCH v7u1 31/31] x86, 64bit, mm: hibernate use generic mapping_init Yinghai Lu
2013-01-04 11:43 ` Rafael J. Wysocki
2013-01-04 21:59 ` Yinghai Lu
2013-01-04 22:07 ` Rafael J. Wysocki
2013-01-04 7:09 ` [PATCH v7u1 00/31] x86, boot, 64bit: Add support for loading ramdisk and bzImage above 4G Borislav Petkov
2013-01-04 21:44 ` Yinghai Lu
2013-01-14 20:45 ` H. Peter Anvin
2013-01-14 22:44 ` Yinghai Lu
2013-01-14 23:16 ` H. Peter Anvin
2013-01-14 23:39 ` David Woodhouse
2013-01-14 23:50 ` H. Peter Anvin
2013-01-15 0:12 ` David Woodhouse
2013-01-15 12:19 ` Stefano Stabellini
2013-01-15 16:43 ` Yinghai Lu
2013-01-15 19:28 ` Yinghai Lu
2013-01-16 11:32 ` Stefano Stabellini
2013-01-16 17:31 ` Yinghai Lu
2013-01-16 17:38 ` H. Peter Anvin
2013-01-16 18:20 ` Yinghai Lu
2013-01-17 2:35 ` Yinghai Lu
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20130108024830.GE9865@phenom.dumpdata.com \
--to=konrad.wilk@oracle.com \
--cc=akpm@linux-foundation.org \
--cc=bp@alien8.de \
--cc=ebiederm@xmission.com \
--cc=hpa@zytor.com \
--cc=jan.kiszka@web.de \
--cc=jason.wessel@windriver.com \
--cc=joro@8bytes.org \
--cc=linux-kernel@vger.kernel.org \
--cc=mingo@elte.hu \
--cc=shuahkhan@gmail.com \
--cc=tglx@linutronix.de \
--cc=yinghai@kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
Powered by JetHome