From: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
To: Yinghai Lu <yinghai@kernel.org>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>,
Shuah Khan <shuahkhan@gmail.com>,
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: Wed, 9 Jan 2013 08:24:09 -0500 [thread overview]
Message-ID: <20130109132409.GB18395@phenom.dumpdata.com> (raw)
In-Reply-To: <CAE9FiQVomDtc0P9B7aTDBrm7kC+i=GYokBmT54w-VBwZsafb5g@mail.gmail.com>
On Tue, Jan 08, 2013 at 05:12:02PM -0800, Yinghai Lu wrote:
> On Tue, Jan 8, 2013 at 5:07 PM, Yinghai Lu <yinghai@kernel.org> wrote:
> > On Tue, Jan 8, 2013 at 4:58 PM, Eric W. Biederman <ebiederm@xmission.com> wrote:
> >
> >>
> >> So instead we need to say?
> >>
> >> + if (no_iotlb_memory)
> >> + panic("Cannot allocate SWIOTLB buffer");
> >> +
> >>
> >> Which is just making the panic a little later than it used to be and
> >> seems completely reasonable.
> >
> > yes, looks some driver just use map_single without checking results.
>
> update one.
Please make it inline.
Anyhow, comments:
> Subject: [PATCH] x86: Don't panic if can not alloc buffer for swiotlb
>
> Normal boot path on system with iommu support:
> swiotlb buffer will be allocated early at first and then try to initialize
> iommu, if iommu for intel or amd could setup properly, swiotlb buffer
Intel or AMD
> will be freed.
>
> The early allocating is with bootmem, and could panic when we try to use
> kdump with buffer above 4G only, or with memmap to limit mem under 4G.
Can you provide the memmap example in here, please.
>
> According to Eric, add _nopanic version and no_iotlb_memory to fail
> map single later if swiotlb is still needed.
>
> -v2: don't pass nopanic, and use -ENOMEM return value according to Eric.
> panic early instead of using swiotlb_full to panic...according to Eric/Konrad.
>
> Suggested-by: Eric W. Biederman <ebiederm@xmission.com>
> Signed-off-by: Yinghai Lu <yinghai@kernel.org>
> Cc: Konrad Rzeszutek Wilk <konrad.wilk@oracle.com>
> Cc: Joerg Roedel <joro@8bytes.org>
>
> ---
> arch/x86/kernel/pci-swiotlb.c | 2 -
> include/linux/swiotlb.h | 1
> lib/swiotlb.c | 54 +++++++++++++++++++++++++++++++++++-------
> 3 files changed, 47 insertions(+), 10 deletions(-)
>
> Index: linux-2.6/include/linux/swiotlb.h
> ===================================================================
> --- linux-2.6.orig/include/linux/swiotlb.h
> +++ linux-2.6/include/linux/swiotlb.h
> @@ -22,6 +22,7 @@ extern int swiotlb_force;
> */
> #define IO_TLB_SHIFT 11
>
> +void swiotlb_init_nopanic(int verbose);
No need for that, lets just use the swiotlb_init version.
> extern void swiotlb_init(int verbose);
> extern void swiotlb_init_with_tbl(char *tlb, unsigned long nslabs, int verbose);
> extern unsigned long swiotlb_nr_tbl(void);
> Index: linux-2.6/lib/swiotlb.c
> ===================================================================
> --- linux-2.6.orig/lib/swiotlb.c
> +++ linux-2.6/lib/swiotlb.c
> @@ -122,11 +122,18 @@ static dma_addr_t swiotlb_virt_to_bus(st
> return phys_to_dma(hwdev, virt_to_phys(address));
> }
>
> +static bool no_iotlb_memory;
> +
> void swiotlb_print_info(void)
> {
> unsigned long bytes = io_tlb_nslabs << IO_TLB_SHIFT;
> unsigned char *vstart, *vend;
>
> + if (no_iotlb_memory) {
> + printk(KERN_INFO "software IO TLB: No low mem\n");
KERN_ERROR or KERN_WARN
> + return;
> + }
> +
> vstart = phys_to_virt(io_tlb_start);
> vend = phys_to_virt(io_tlb_end);
>
> @@ -136,7 +143,8 @@ void swiotlb_print_info(void)
> bytes >> 20, vstart, vend - 1);
> }
>
> -void __init swiotlb_init_with_tbl(char *tlb, unsigned long nslabs, int verbose)
> +static int __init __swiotlb_init_with_tbl(char *tlb, unsigned long nslabs,
> + int verbose)
Collapse it in. Meaning make swiotlb_init_with_tbl return an int. This will
require altering the Xen version of Xen-SWIOTLB but that should be fairly easy.
> {
> void *v_overflow_buffer;
> unsigned long i, bytes;
> @@ -150,9 +158,10 @@ void __init swiotlb_init_with_tbl(char *
> /*
> * Get the overflow emergency buffer
> */
> - v_overflow_buffer = alloc_bootmem_low_pages(PAGE_ALIGN(io_tlb_overflow));
> + v_overflow_buffer = alloc_bootmem_low_pages_nopanic(
> + PAGE_ALIGN(io_tlb_overflow));
> if (!v_overflow_buffer)
> - panic("Cannot allocate SWIOTLB overflow buffer!\n");
> + return -ENOMEM;
>
> io_tlb_overflow_buffer = __pa(v_overflow_buffer);
>
> @@ -169,14 +178,22 @@ void __init swiotlb_init_with_tbl(char *
>
> if (verbose)
> swiotlb_print_info();
> +
> + return 0;
> +}
> +
> +void __init swiotlb_init_with_tbl(char *tlb, unsigned long nslabs, int verbose)
> +{
> + if (__swiotlb_init_with_tbl(tlb, nslabs, verbose) == -ENOMEM)
> + panic("Cannot allocate SWIOTLB buffer");
And that way we can get rid of this.
> }
>
> /*
> * Statically reserve bounce buffer space and initialize bounce buffer data
> * structures for the software IO TLB used to implement the DMA API.
> */
> -static void __init
> -swiotlb_init_with_default_size(size_t default_size, int verbose)
> +static __init
> +int swiotlb_init_with_default_size(size_t default_size, int verbose)
> {
> unsigned char *vstart;
> unsigned long bytes;
> @@ -191,17 +208,33 @@ swiotlb_init_with_default_size(size_t de
> /*
> * Get IO TLB memory from the low pages
> */
> - vstart = alloc_bootmem_low_pages(PAGE_ALIGN(bytes));
> + vstart = alloc_bootmem_low_pages_nopanic(PAGE_ALIGN(bytes));
> if (!vstart)
> - panic("Cannot allocate SWIOTLB buffer");
> + return -ENOMEM;
>
> - swiotlb_init_with_tbl(vstart, io_tlb_nslabs, verbose);
> + if (__swiotlb_init_with_tbl(vstart, io_tlb_nslabs, verbose) == -ENOMEM) {
> + free_bootmem(io_tlb_start,
> + PAGE_ALIGN(io_tlb_nslabs << IO_TLB_SHIFT));
> + return -ENOMEM;
> + }
> +
> + return 0;
> +}
> +
> +void __init
> +swiotlb_init_nopanic(int verbose)
> +{
> + /* default to 64MB */
> + if (swiotlb_init_with_default_size(64 * (1<<20), verbose) == -ENOMEM)
> + pr_warn("Cannot allocate SWIOTLB buffer");
> }
And this
>
> void __init
> swiotlb_init(int verbose)
> {
> - swiotlb_init_with_default_size(64 * (1<<20), verbose); /* default to 64MB */
> + /* default to 64MB */
> + if (swiotlb_init_with_default_size(64 * (1<<20), verbose) == -ENOMEM)
> + panic("Cannot allocate SWIOTLB buffer");
> }
And just make this function return 'int'
>
> /*
> @@ -405,6 +438,9 @@ phys_addr_t swiotlb_tbl_map_single(struc
> unsigned long offset_slots;
> unsigned long max_slots;
>
> + if (no_iotlb_memory)
> + panic("Cannot allocate SWIOTLB buffer");
> +
> mask = dma_get_seg_boundary(hwdev);
>
> tbl_dma_addr &= mask;
> Index: linux-2.6/arch/x86/kernel/pci-swiotlb.c
> ===================================================================
> --- linux-2.6.orig/arch/x86/kernel/pci-swiotlb.c
> +++ linux-2.6/arch/x86/kernel/pci-swiotlb.c
> @@ -91,7 +91,7 @@ IOMMU_INIT(pci_swiotlb_detect_4gb,
> void __init pci_swiotlb_init(void)
> {
> if (swiotlb) {
> - swiotlb_init(0);
> + swiotlb_init_nopanic(0);
> dma_ops = &swiotlb_dma_ops;
> }
> }
next prev parent reply other threads:[~2013-01-09 13:24 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
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 [this message]
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=20130109132409.GB18395@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