[Xen-ia64-devel] Re: [Xen-devel] [PATCH 0/5] dump-core take 2:
On Thu, Jan 18, 2007 at 07:13:17AM +, Keir Fraser wrote: On 18/1/07 6:52 am, Isaku Yamahata [EMAIL PROTECTED] wrote: Subject: [PATCH 1/5] dump-core take 2: XENMEM_set_memory_map hypercall Subject: [PATCH 2/5] dump-core take 2: libxc: xc_domain memmap functions Should be able to work without these. We need to be able to support ballooning anyway, so it's not as if every E820_RAM region will necessarily be entirely populated with memory. What you need is a max_pfn value and then iterate 0...max_pfn-1 and try to map each page. If the mapping fails then there is no underlying memory. The tools could give a suitable max_pfn or we could add a hypercall to get it from Xen. max_pfn isn't sufficient. Memory may be sparse on ia64 so that iterating on [0, max_pfn - 1] isn't practical. It would take too long time. Mempry map is also necessary to avoid dumping I/O regions of a driver domain. Subject: [PATCH 3/5] dump-core take 2: libxc: add xc_domain_tranlate_gpfn() Why? x86 moved to always mapping HVM memory by GPFN. Can ia64 do the same? IA64 uses GPFN for both domU and HVM. It is used to just in order to check whether each GPFNs have underlying memory. Not in order to get MFN. It can be replaced with trying to map and checking the result, however I want to know it _before_ dumping pages to create program headers. Is there any cheaper way than trying to map each PFNs? Subject: [PATCH 5/5] dump-core take 2: elf formatify and added PFN-GMFN table Shouldn't dump zero pages. Hence we need PFN-GMFN info even for HVM guests -- absence of PFN-GMFN pair, or GMFN==INVALID_MFN, could represent a RAM hole more cheaply than 4kB of zeroes. Otherwise PFN=GMFN. I'm not sure I understand. The posted patch doesn't dump a page which doesn't have underlying memory. By checking program header's physical address and size (Elf_Phdr.{p_paddr, p_filesz}), we can know whether a given GPFN is present or not. -- yamahata ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
[Xen-ia64-devel][Patch] Clean patch (clean VLIDATE_VT)
It's not be defined, so clean it. Good good study,day day up ! ^_^ -Wing(zhang xin) OTC,Intel Corporation clean_VALIDATE_VT.patch Description: clean_VALIDATE_VT.patch ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
RE: [Xen-ia64-devel] Re: [Xen-devel] [PATCH 0/5] dump-core take 2:
From: Isaku Yamahata Sent: 2007年1月18日 16:33 Should be able to work without these. We need to be able to support ballooning anyway, so it's not as if every E820_RAM region will necessarily be entirely populated with memory. What you need is a max_pfn value and then iterate 0...max_pfn-1 and try to map each page. If the mapping fails then there is no underlying memory. The tools could give a suitable max_pfn or we could add a hypercall to get it from Xen. max_pfn isn't sufficient. Memory may be sparse on ia64 so that iterating on [0, max_pfn - 1] isn't practical. It would take too long time. Mempry map is also necessary to avoid dumping I/O regions of a driver domain. Yeah, memory map may be sparse on ia64, but, only at physical level. You can always present a compact pseudo physical layout to a domain, despite of sparse or not in real physical.:-) BTW, is it possible to save memmap into xenstore, so that multiple user components can communicate such info directly without xen's intervention? Thanks, Kevin ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
[Xen-ia64-devel] Xen/IA64 Healthiness Report -Cset#13465
Xen/IA64 Healthiness Report All 16 cases pass testing. Testing Environment: Platform: Tiger4 Processor: Itanium 2 Processor Logic Processors number: 8 (2 processors with Due Core) Service OS: RHEL4u3 IA64 SMP with 2 vcpus 1G memory VTI Guest OS: RHEL4u2 RHEL4u3 XenU Guest OS: RHEL4u2 Xen IA64 Unstable tree: 13465:58637a0a7c7e Xen Schedule: credit VTI Guest Firmware Flash.fd.2006.12.01 MD5: 09a224270416036a8b4e6f8496e97854 Summary Test Report: - Total cases: 16 Passed:16 Failed: 0 Case Name Status Case Description Four_SMPVTI_Coexistpass 4 VTI (mem=256, vcpus=2) Two_UP_VTI_Co pass 2 UP_VTI (mem=256) One_UP_VTIpass1 UP_VTI (mem=256) One_UP_XenU pass1 UP_xenU(mem=256) SMPVTI_LTPpassVTI (vcpus=4, mem=512) run LTP SMPVTI_and_SMPXenU pass 1 VTI + 1 xenU (mem=256 vcpus=2) Two_SMPXenU_Coexistpass2 xenU (mem=256, vcpus=2) One_SMPVTI_4096M pass 1 VTI (vcpus=2, mem=4096M) SMPVTI_Network pass 1 VTI (mem=256,vcpu=2) and 'ping' SMPXenU_Networkpass 1 XenU (vcpus=2) and 'ping' One_SMP_XenU pass 1 SMP xenU (vcpus=2) One_SMP_VTIpass 1 SMP VTI (vcpus=2) SMPVTI_Kernel_Build pass VTI (vcpus=4) and do Kernel Build Four_SMPVTI_Coexist pass 4 VTI domains( mem=256, vcpu=2) SMPVTI_Windows pass SMPVTI windows(vcpu=2) SMPWin_SMPVTI_SMPxenU pass SMPVTI Linux/Windows XenU UPVTI_Kernel_Build pass 1 UP VTI and do kernel build Notes: - The last stable changeset: - 13465:58637a0a7c7e Thanks, Yongkang ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
RE: [Xen-ia64-devel] Re: [Xen-devel] [PATCH 0/5] dump-core take 2:
From: Isaku Yamahata [mailto:[EMAIL PROTECTED] Sent: 2007年1月18日 17:25 On Thu, Jan 18, 2007 at 04:54:05PM +0800, Tian, Kevin wrote: Yeah, memory map may be sparse on ia64, but, only at physical level. You can always present a compact pseudo physical layout to a domain, despite of sparse or not in real physical.:-) That's right. Xen/ia64 does so now for paravirtualized domain except dom0. There is an unsolved issue. If much memory (e.g. 4GB) is given to a driver domain, the domain can't access I/O. At least the I/O area must be avoided somehow, thus paravirtualized domain's memory map may become sparse (in the future when the issue is solved). Yes, if I/O regions are very sparse, so does memory map for driver domain. BTW, is it possible to save memmap into xenstore, so that multiple user components can communicate such info directly without xen's intervention? Do you have any usage in mind? Case like your above requirement, case like qemu, and even case like save/restore... anyway, to me there's no need to let Xen aware of the domain memmap. Domain image builder constructs the memmap based on its configuration, and then just notify xen to allocate pages for appropriate regions or setup mapping for assigned MMIO ranges. If builder also saves memmap to xenstore, you don't need above hypercall then. Just an alternative... Thanks, Kevin ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
RE: [Xen-ia64-devel][PATCH]Implement eager save, lazy restore FPalgorithm
Alex Williamson write on 2007年1月17日 5:12: On Tue, 2007-01-16 at 12:52 +0800, Xu, Anthony wrote: This patch Implements eager save, lazy restore FP algorithm, This patch reduces the cost of VCPU schedule. Hi Anthony, In theory, this is nice. In practice, I get segfaults doing a kernel build in a domVT with multiple VT domains running. Thanks, Alex Hi Alex, I didn't consider VCPU migration in previous patch. This patch fixed this. Tested by running KB on two VTI domain with two VCPUs each. --anthony lazy_fp2.patch Description: lazy_fp2.patch ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
Re: [Xen-ia64-devel] [Patch] allocate all memory to dom0
On Thu, 2007-01-18 at 13:06 +0900, Akio Takebe wrote: Hi, If we don't specify dom0_mem, we can use all memory on dom0 with this patch. I change alloc_dom0() to alloc_dom0_size(), and alloc_dom0_size() is static function. Hi Akio, I don't think we're able to do this yet. One of my test systems is an rx6600 w/ 96G of memory. With this patch, I get a _very_ long pause during Xen boot (30 seconds), followed by an endless loop of page allocation failures: (XEN) Domain0 EFI passthrough: ACPI 2.0=0x3fdce000 SMBIOS=0x3e52a000 [long pause here] (XEN) Cannot handle page request order 0! (XEN) Cannot handle page request order 0! (XEN) Cannot handle page request order 0! (XEN) Cannot handle page request order 0! (XEN) Cannot handle page request order 0! ... The most memory I can specify for dom0_mem and still boot seems to be around 87G. Obviously failing to boot is bad, but the long pause while we're individually mapping every page for dom0 is also a usability issue. We either need to see about optimizing this code segment to reduce the delay to something acceptable, or add some kind of forward progress indicator. Thanks, Alex -- Alex Williamson HP Open Source Linux Org. ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
RE: [Xen-ia64-devel][PATCH] don't panic dom0 if there is not enoughmemory to create guest
I am sorry I miss the case which VHPT allocated successful but TLB allocated failed. This patch is a new version for that. Good good study,day day up ! ^_^ -Wing(zhang xin) OTC,Intel Corporation From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Zhang, Xing Z Sent: 2007年1月18日 18:01 To: xen-ia64-devel@lists.xensource.com Subject: [Xen-ia64-devel][PATCH] don't panic dom0 if there is not enoughmemory to create guest When there is not enough memory to create a domain, we not need panic domain0. Just prevent it from crating. Signed-off-by, Zhang Xin [EMAIL PROTECTED] Good good study,day day up ! ^_^ -Wing(zhang xin) OTC,Intel Corporation ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
RE: [Xen-ia64-devel][PATCH] don't panic dom0 if there is not enoughmemory to create guest
Sorry, forgot to attach the patch. Good good study,day day up ! ^_^ -Wing(zhang xin) OTC,Intel Corporation From: Zhang, Xing Z Sent: 2007年1月19日 9:59 To: Zhang, Xing Z; xen-ia64-devel@lists.xensource.com Subject: RE: [Xen-ia64-devel][PATCH] don't panic dom0 if there is not enoughmemory to create guest I am sorry I miss the case which VHPT allocated successful but TLB allocated failed. This patch is a new version for that. Good good study,day day up ! ^_^ -Wing(zhang xin) OTC,Intel Corporation From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Zhang, Xing Z Sent: 2007年1月18日 18:01 To: xen-ia64-devel@lists.xensource.com Subject: [Xen-ia64-devel][PATCH] don't panic dom0 if there is not enoughmemory to create guest When there is not enough memory to create a domain, we not need panic domain0. Just prevent it from crating. Signed-off-by, Zhang Xin [EMAIL PROTECTED] Good good study,day day up ! ^_^ -Wing(zhang xin) OTC,Intel Corporation donnot_panic_dom0_2.patch Description: donnot_panic_dom0_2.patch ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
Re: [Xen-ia64-devel] [Patch] allocate all memory to dom0
Hi, Alex Thank you for your comments. I agree, my patch can be applied after fixing .config(of discontigous issue), fixing and optimizing them, and so on. Also, as Mark said, xen-ia64 may not need to allocate all memory and cpu to dom0. IA64 machine is very large system, and dom0 don't need large machine environment. Best Regards, Akio Takebe On Thu, 2007-01-18 at 13:06 +0900, Akio Takebe wrote: Hi, If we don't specify dom0_mem, we can use all memory on dom0 with this patch. I change alloc_dom0() to alloc_dom0_size(), and alloc_dom0_size() is static function. Hi Akio, I don't think we're able to do this yet. One of my test systems is an rx6600 w/ 96G of memory. With this patch, I get a _very_ long pause during Xen boot (30 seconds), followed by an endless loop of page allocation failures: (XEN) Domain0 EFI passthrough: ACPI 2.0=0x3fdce000 SMBIOS=0x3e52a000 [long pause here] (XEN) Cannot handle page request order 0! (XEN) Cannot handle page request order 0! (XEN) Cannot handle page request order 0! (XEN) Cannot handle page request order 0! (XEN) Cannot handle page request order 0! ... The most memory I can specify for dom0_mem and still boot seems to be around 87G. Obviously failing to boot is bad, but the long pause while we're individually mapping every page for dom0 is also a usability issue. We either need to see about optimizing this code segment to reduce the delay to something acceptable, or add some kind of forward progress indicator. Thanks, Alex -- Alex Williamson HP Open Source Linux Org. ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
Re: [Xen-ia64-devel] Re: [Xen-devel] [PATCH 0/5] dump-core take 2:
Kevin == Tian, Kevin [EMAIL PROTECTED] writes: Kevin Yeah, memory map may be sparse on ia64, but, only at physical Kevin level. You can always present a compact pseudo physical layout Kevin to a domain, despite of sparse or not in real physical.:-) BTW, Kevin is it possible to save memmap into xenstore, so that multiple Kevin user components can communicate such info directly without Kevin xen's intervention? Providing a fake linear memory map like that is totally broken, it means the domU operating system will not be able to benefit from NUMA information and do appropriate scheduling. The domU pages needs to be placed in the metaphysical memory zones that match their physical zone to get this right. We can provide a virtual linear map for special cases, like to support lesser operating systems that can't handle real computers, but the general case needs to be that pages go into the metaphysical zone that matches their real physical zone. This is applicable to any NUMA system, not just ia64 systems, so with x86_64 becoming mainstream they will need it there too. A linear scan of the pfn list is just wrong, one should never do that. Cheers, Jes ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
[Xen-ia64-devel] [Patch 0/2] INIT handler improvement
Hi, I fix a bug and I improve CallTrace at the time of pushing INIT button. This improve to support calltrace of all vcpu. [1/2] fix a bug init_handler_platform [2/2] improve calltarce Best Regards, Akio Takebe ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
[Xen-ia64-devel] [Patch 0/1] fix a bug init_handler_platform
Hi, This patch fix breaking switch stack. unw_init_running() in INIT handler the save switch stack pointer. The swith stack is not guaranteed after returning unw_init_running(). But because INIT handler of Xen-ia64 call some functions after unw_init_running(), the switch stack is broken by the functions. Signed-off-by: Akio Takebe [EMAIL PROTECTED] Best Regards, Akio Takebe fix_break_sw_init_hanlder.patch Description: Binary data ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
[Xen-ia64-devel] [Patch 2/2] improve calltarce
Hi, I improve CallTrace at the time of pushing INIT button. This improve to support calltrace of all vcpu. Signed-off-by: Akio Takebe [EMAIL PROTECTED] Best Regards, Akio Takebe improvement_init_handler.patch Description: Binary data ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
Re: [Xen-ia64-devel] Xen crash when creating VTI in some machines.
Hi Yongkang, Hmmm... It's the strange issue. I think so, too. I'm trying to reproduce the strange issue. But I'm not able to meet the strange issue at this stage. Currently, the creation of HVM domain does not call the pervcpu_vhpt_free() by my patch(changeset:13434). Which function is f7b30080? Best regards, Kan Hi all, This is a strange issue, because some machines do not meet this problem, including our daily regular testing machine. But we found Xen might crash when doing the operation of creating VTI in some other machines. Serial console would keep report: ... (XEN) [f4080300] pervcpu_vhpt_free+0x30/0x50 (XEN) sp=f7bffe00 bsp= f7bf93e8 (XEN) [f7b30080] ??? (XEN) sp=f7bffe00 bsp= f7bf93e8 ... We found this issue at least in 13438 and 13465. Does anybody meet this issue too? Best Regards, Yongkang (Kangkang) モタソオ ---text/plain--- ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
RE: [Xen-ia64-devel] Xen crash when creating VTI in some machines.
Hi Wing and Yongkang, Oh! I am sorry that it was caused by my patch. I will review my patch. Best regards, Kan Hi Kan: f7b30080 is an address belong to free_domheap_pages(). I found if reversed your patch, the issue will not meet. Could you give a review to your patch, maybe there is something lost. Good good study,day day up ! ^_^ -Wing(zhang xin) OTC,Intel Corporation -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Masaki Kanno Sent: 2007年1月19日 13:13 To: You, Yongkang; xen-ia64-devel Subject: Re: [Xen-ia64-devel] Xen crash when creating VTI in some machines. Hi Yongkang, Hmmm... It's the strange issue. I think so, too. I'm trying to reproduce the strange issue. But I'm not able to meet the strange issue at this stage. Currently, the creation of HVM domain does not call the pervcpu_vhpt_free() by my patch(changeset:13434). Which function is f7b30080? Best regards, Kan Hi all, This is a strange issue, because some machines do not meet this problem, including our daily regular testing machine. But we found Xen might crash when doing the operation of creating VTI in some other machines. Serial console would keep report: ... (XEN) [f4080300] pervcpu_vhpt_free+0x30/0x50 (XEN) sp=f7bffe00 bsp= f7bf93e8 (XEN) [f7b30080] ??? (XEN) sp=f7bffe00 bsp= f7bf93e8 ... We found this issue at least in 13438 and 13465. Does anybody meet this issue too? Best Regards, Yongkang (Kangkang) モタソオ ---text/plain--- ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel ___ Xen-ia64-devel mailing list Xen-ia64-devel@lists.xensource.com http://lists.xensource.com/xen-ia64-devel
[Xen-ia64-devel] Weekly benchmark results [ww03]
Hi, I report a benchmark result of this week. In LTP, clone06 and sockioctl01 failed. Both cases passed when testing manually. sockioctl01's fail caused by a bad configuration (/dev/tty0 is missing). I will fix it. clone06 fails because the environment variable TERM is unset. In such cases, clone06 causes SEGV. TEST ENVIRONMENT Machine : Tiger4 Kernel : 2.6.16.33-xen Changeset: 13438:43115ffc6635 (xen-ia64-unstable) Dom0 OS : RHEL4 U2 (2P) DomU OS : RHEL4 U2 (8P) #DomU: 1 Scheduler: credit Device : tap (tap:aio) TEST RESULT unixbench4.1.0: Pass bonnie++-1.03 : Pass ltp-full-20061121 : 3/831 FAIL (see attachment) iozone3_191 : Pass lmbench-3.0-a5: Pass Best regards, KUWAMURA and Fujitsu members Test Start Time: Thu Jan 18 17:18:04 2007 - Testcase Result Exit Value -- -- abort01PASS 0 accept01 PASS 0 access01 PASS 0 access02 PASS 0 access03 PASS 0 access04 PASS 0 access05 PASS 0 acct01 PASS 0 acct02 PASS 0 adjtimex01 PASS 0 adjtimex02 PASS 0 alarm01PASS 0 alarm02PASS 0 alarm03PASS 0 alarm04PASS 0 alarm05PASS 0 alarm06PASS 0 alarm07PASS 0 asyncio02 PASS 0 bind01 PASS 0 bind02 PASS 0 brk01 PASS 0 capget01 PASS 0 capget02 PASS 0 capset01 PASS 0 capset02 PASS 0 chdir01PASS 0 chdir01A PASS 0 chdir02PASS 0 chdir03PASS 0 chdir04PASS 0 chmod01PASS 0 chmod01A PASS 0 chmod02PASS 0 chmod03PASS 0 chmod04PASS 0 chmod05PASS 0 chmod06PASS 0 chmod07PASS 0 chown01PASS 0 chown02PASS 0 chown03PASS 0 chown04PASS 0 chown05PASS 0 chroot01 PASS 0 chroot02 PASS 0 chroot03 PASS 0 chroot04 PASS 0 clone01PASS 0 clone02PASS 0 clone03PASS 0 clone04PASS 0 clone05PASS 0 clone06FAIL 2 clone07PASS 0 close01PASS 0 close02PASS 0 close08PASS 0 confstr01 PASS 0 connect01 PASS 0 creat01PASS 0 creat03PASS 0 creat04PASS 0 creat05PASS 0 creat06PASS 0 creat07PASS 0 creat08PASS 0 creat09PASS 0 dup01 PASS 0 dup02 PASS 0 dup03 PASS 0 dup04 PASS 0 dup05 PASS 0 dup06 PASS 0 dup07 PASS 0 dup201 PASS 0 dup202 PASS 0 dup203 PASS 0 dup204 PASS 0 dup205 PASS 0 execl01PASS 0 execle01 PASS 0 execlp01