diff for duplicates of <200802131617.58646.ossthema@de.ibm.com> diff --git a/a/1.txt b/N1/1.txt index be97cbb..02afc45 100644 --- a/a/1.txt +++ b/N1/1.txt @@ -15,50 +15,44 @@ to form a base for the eHEA memory add / remove concept discussion: Explanation of the current eHEA memory add / remove concept: Constraints imposed by HW / FW: -=2D eHEA has own MMU -=2D eHEA =A0Memory Regions (MRs) are used by the eHEA MMU =A0to translate v= -irtual -=A0 addresses to absolute addresses (like DMA mapped memory on a PCI bus) -=2D The number of MRs is limited (not enough to have one MR per packet) -=2D Registration of MRs is comparativley slow as done via slow firmware call +- eHEA has own MMU +- eHEA Memory Regions (MRs) are used by the eHEA MMU to translate virtual + addresses to absolute addresses (like DMA mapped memory on a PCI bus) +- The number of MRs is limited (not enough to have one MR per packet) +- Registration of MRs is comparativley slow as done via slow firmware call (H_CALL) -=2D MRs can have a maximum size of the memory available under linux -=2D MRs cover a contiguous virtual memory block (no holes) +- MRs can have a maximum size of the memory available under linux +- MRs cover a contiguous virtual memory block (no holes) Because of this there is just one big MR that covers entire kernel memory. We also need a mapping table from kernel addresses to this contiguous "virtual memory IO space" (here called ehea_bmap). -=2D When memory is added / removed to LPAR (and linux), the MR has to be up= -dated. -=A0 This can only be done by destroying and recreating the MR. There is no = -H_CALL -=A0 to modify MR size. To find holes in the linux kernel memory layout we h= -ave to -=A0 iterate over the memory sections for recreating a ehea_bmap +- When memory is added / removed to LPAR (and linux), the MR has to be updated. + This can only be done by destroying and recreating the MR. There is no H_CALL + to modify MR size. To find holes in the linux kernel memory layout we have to + iterate over the memory sections for recreating a ehea_bmap (otherwise MR would be bigger then available memory causing the registration to fail) -=2D DLPAR userspace tools, kernel, driver, firmware and HMC are involved in= - that -=A0 process on System p +- DLPAR userspace tools, kernel, driver, firmware and HMC are involved in that + process on System p Memory add: version without a external memory notifier call -=2D new memory used in a transfer_xmit will result in a "ehea_bmap +- new memory used in a transfer_xmit will result in a "ehea_bmap translation miss", which triggers a rebuild and reregistration -=A0 of the ehea_bmap based on the current kernel memory setup. -=2D advantage: the number of MR rebuilds is reduced significantly compared = -to + of the ehea_bmap based on the current kernel memory setup. +- advantage: the number of MR rebuilds is reduced significantly compared to a rebuild for each 16MB chunk of memory added. Memory add: version with external notifier call: -=2D We still need a ehea_bmap (whatever structure it has) +- We still need a ehea_bmap (whatever structure it has) Memory remove with notifier: -=2D We have to rebuild the ehea_bmap instantly to remove the pages that are +- We have to rebuild the ehea_bmap instantly to remove the pages that are no longer available. Without doing that, the firmware (pHYP) cannot remove - that memory from the LPAR. As we don't know if or how many additional=20 - sections are to be removed before the DLPAR user space tool tells the=20 + that memory from the LPAR. As we don't know if or how many additional + sections are to be removed before the DLPAR user space tool tells the firmware to remove the memory, we can't wait with the rebuild. @@ -66,33 +60,32 @@ Our current understanding about the current Memory Hotplug System are (please correct me if I'm wrong): -=2D depends on sparse mem -=2D only whole memory sections are added / removed -=2D for each section a memory resource is registered +- depends on sparse mem +- only whole memory sections are added / removed +- for each section a memory resource is registered -=46rom the driver side we need: -=2D some kind of memory notification mechanism. -=A0 For memory add we can live without any external memory notification +>From the driver side we need: +- some kind of memory notification mechanism. + For memory add we can live without any external memory notification event. For memory remove we do need an external trigger (see explanation above). -=2D a way to iterate over all kernel pages and a way to detect holes in the +- a way to iterate over all kernel pages and a way to detect holes in the kernel memory layout in order to build up our own ehea_bmap. Memory notification trigger: -=2D These triggers exist, an exported "register_memory_notifier" / -=A0 "unregister_memory_notifier" would work in this scheme +- These triggers exist, an exported "register_memory_notifier" / + "unregister_memory_notifier" would work in this scheme -=46unctions to use while building ehea_bmap + MRs: -=2D Use either the functions that are used by the memory hotplug system as +Functions to use while building ehea_bmap + MRs: +- Use either the functions that are used by the memory hotplug system as well, that means using the section defines + functions (section_nr_to_pfn, -=A0 pfn_valid) -=2D Use currently other not exported functions in kernel/resource.c, like - walk_memory_resource (where we would still need the maximum possible numb= -er + pfn_valid) +- Use currently other not exported functions in kernel/resource.c, like + walk_memory_resource (where we would still need the maximum possible number of pages NR_MEM_SECTIONS) -=2D Maybe some kind of new interface? +- Maybe some kind of new interface? What would you suggest? diff --git a/a/content_digest b/N1/content_digest index 0ebf518..5daad6a 100644 --- a/a/content_digest +++ b/N1/content_digest @@ -4,17 +4,17 @@ "Subject\0Re: [PATCH] drivers/base: export gpl (un)register_memory_notifier\0" "Date\0Wed, 13 Feb 2008 16:17:57 +0100\0" "To\0linuxppc-dev@ozlabs.org\0" - "Cc\0Thomas Klein <tklein@de.ibm.com>" + "Cc\0Dave Hansen <haveblue@us.ibm.com>" + Thomas Klein <tklein@de.ibm.com> Themann Jan-Bernd <themann@de.ibm.com> netdev <netdev@vger.kernel.org> - Dave Hansen <haveblue@us.ibm.com> apw <apw@uk.ibm.com> linux-kernel <linux-kernel@vger.kernel.org> + Thomas Klein <osstklei@de.ibm.com> Christoph Raisch <RAISCH@de.ibm.com> Badari Pulavarty <pbadari@us.ibm.com> - Greg KH <greg@kroah.com> - " Thomas Klein <osstklei@de.ibm.com>\0" + " Greg KH <greg@kroah.com>\0" "\00:1\0" "b\0" "Hi Dave,\n" @@ -34,50 +34,44 @@ "Explanation of the current eHEA memory add / remove concept:\n" "\n" "Constraints imposed by HW / FW:\n" - "=2D eHEA has own MMU\n" - "=2D eHEA =A0Memory Regions (MRs) are used by the eHEA MMU =A0to translate v=\n" - "irtual\n" - "=A0 addresses to absolute addresses (like DMA mapped memory on a PCI bus)\n" - "=2D The number of MRs is limited (not enough to have one MR per packet)\n" - "=2D Registration of MRs is comparativley slow as done via slow firmware call\n" + "- eHEA has own MMU\n" + "- eHEA \302\240Memory Regions (MRs) are used by the eHEA MMU \302\240to translate virtual\n" + "\302\240 addresses to absolute addresses (like DMA mapped memory on a PCI bus)\n" + "- The number of MRs is limited (not enough to have one MR per packet)\n" + "- Registration of MRs is comparativley slow as done via slow firmware call\n" "(H_CALL)\n" - "=2D MRs can have a maximum size of the memory available under linux\n" - "=2D MRs cover a contiguous virtual memory block (no holes)\n" + "- MRs can have a maximum size of the memory available under linux\n" + "- MRs cover a contiguous virtual memory block (no holes)\n" "\n" "Because of this there is just one big MR that covers entire kernel memory.\n" "We also need a mapping table from kernel addresses to this\n" "contiguous \"virtual memory IO space\" (here called ehea_bmap).\n" "\n" - "=2D When memory is added / removed to LPAR (and linux), the MR has to be up=\n" - "dated.\n" - "=A0 This can only be done by destroying and recreating the MR. There is no =\n" - "H_CALL\n" - "=A0 to modify MR size. To find holes in the linux kernel memory layout we h=\n" - "ave to\n" - "=A0 iterate over the memory sections for recreating a ehea_bmap\n" + "- When memory is added / removed to LPAR (and linux), the MR has to be updated.\n" + "\302\240 This can only be done by destroying and recreating the MR. There is no H_CALL\n" + "\302\240 to modify MR size. To find holes in the linux kernel memory layout we have to\n" + "\302\240 iterate over the memory sections for recreating a ehea_bmap\n" " (otherwise MR would be bigger then available memory causing the\n" " registration to fail)\n" "\n" - "=2D DLPAR userspace tools, kernel, driver, firmware and HMC are involved in=\n" - " that\n" - "=A0 process on System p\n" + "- DLPAR userspace tools, kernel, driver, firmware and HMC are involved in that\n" + "\302\240 process on System p\n" "\n" "Memory add: version without a external memory notifier call\n" - "=2D new memory used in a transfer_xmit will result in a \"ehea_bmap\n" + "- new memory used in a transfer_xmit will result in a \"ehea_bmap\n" " translation miss\", which triggers a rebuild and reregistration\n" - "=A0 of the ehea_bmap based on the current kernel memory setup.\n" - "=2D advantage: the number of MR rebuilds is reduced significantly compared =\n" - "to\n" + "\302\240 of the ehea_bmap based on the current kernel memory setup.\n" + "- advantage: the number of MR rebuilds is reduced significantly compared to\n" " a rebuild for each 16MB chunk of memory added.\n" "\n" "Memory add: version with external notifier call:\n" - "=2D We still need a ehea_bmap (whatever structure it has)\n" + "- We still need a ehea_bmap (whatever structure it has)\n" "\n" "Memory remove with notifier:\n" - "=2D We have to rebuild the ehea_bmap instantly to remove the pages that are\n" + "- We have to rebuild the ehea_bmap instantly to remove the pages that are\n" " no longer available. Without doing that, the firmware (pHYP) cannot remove\n" - " that memory from the LPAR. As we don't know if or how many additional=20\n" - " sections are to be removed before the DLPAR user space tool tells the=20\n" + " that memory from the LPAR. As we don't know if or how many additional \n" + " sections are to be removed before the DLPAR user space tool tells the \n" " firmware to remove the memory, we can't wait with the rebuild.\n" "\n" "\n" @@ -85,37 +79,36 @@ "(please correct me\n" "if I'm wrong):\n" "\n" - "=2D depends on sparse mem\n" - "=2D only whole memory sections are added / removed\n" - "=2D for each section a memory resource is registered\n" + "- depends on sparse mem\n" + "- only whole memory sections are added / removed\n" + "- for each section a memory resource is registered\n" "\n" "\n" - "=46rom the driver side we need:\n" - "=2D some kind of memory notification mechanism.\n" - "=A0 For memory add we can live without any external memory notification\n" + ">From the driver side we need:\n" + "- some kind of memory notification mechanism.\n" + "\302\240 For memory add we can live without any external memory notification\n" " event. For memory remove we do need an external trigger (see explanation\n" " above).\n" - "=2D a way to iterate over all kernel pages and a way to detect holes in the\n" + "- a way to iterate over all kernel pages and a way to detect holes in the\n" " kernel memory layout in order to build up our own ehea_bmap.\n" "\n" "\n" "Memory notification trigger:\n" - "=2D These triggers exist, an exported \"register_memory_notifier\" /\n" - "=A0 \"unregister_memory_notifier\" would work in this scheme\n" + "- These triggers exist, an exported \"register_memory_notifier\" /\n" + "\302\240 \"unregister_memory_notifier\" would work in this scheme\n" "\n" - "=46unctions to use while building ehea_bmap + MRs:\n" - "=2D Use either the functions that are used by the memory hotplug system as\n" + "Functions to use while building ehea_bmap + MRs:\n" + "- Use either the functions that are used by the memory hotplug system as\n" " well, that means using the section defines + functions (section_nr_to_pfn,\n" - "=A0 pfn_valid)\n" - "=2D Use currently other not exported functions in kernel/resource.c, like\n" - " walk_memory_resource (where we would still need the maximum possible numb=\n" - "er\n" + "\302\240 pfn_valid)\n" + "- Use currently other not exported functions in kernel/resource.c, like\n" + " walk_memory_resource (where we would still need the maximum possible number\n" " of pages NR_MEM_SECTIONS)\n" - "=2D Maybe some kind of new interface?\n" + "- Maybe some kind of new interface?\n" "\n" "What would you suggest?\n" "\n" "Regards,\n" Jan-Bernd & Christoph -0e3dd35328cab4f06248746293b4eb3629329b7c3c1f70c4a339131e2e7b359e +a6ef93a19b1660ad4dd2b7de7b1faed562d34d4fb333fbcfdee896851d43b196
diff --git a/a/1.txt b/N2/1.txt index be97cbb..02afc45 100644 --- a/a/1.txt +++ b/N2/1.txt @@ -15,50 +15,44 @@ to form a base for the eHEA memory add / remove concept discussion: Explanation of the current eHEA memory add / remove concept: Constraints imposed by HW / FW: -=2D eHEA has own MMU -=2D eHEA =A0Memory Regions (MRs) are used by the eHEA MMU =A0to translate v= -irtual -=A0 addresses to absolute addresses (like DMA mapped memory on a PCI bus) -=2D The number of MRs is limited (not enough to have one MR per packet) -=2D Registration of MRs is comparativley slow as done via slow firmware call +- eHEA has own MMU +- eHEA Memory Regions (MRs) are used by the eHEA MMU to translate virtual + addresses to absolute addresses (like DMA mapped memory on a PCI bus) +- The number of MRs is limited (not enough to have one MR per packet) +- Registration of MRs is comparativley slow as done via slow firmware call (H_CALL) -=2D MRs can have a maximum size of the memory available under linux -=2D MRs cover a contiguous virtual memory block (no holes) +- MRs can have a maximum size of the memory available under linux +- MRs cover a contiguous virtual memory block (no holes) Because of this there is just one big MR that covers entire kernel memory. We also need a mapping table from kernel addresses to this contiguous "virtual memory IO space" (here called ehea_bmap). -=2D When memory is added / removed to LPAR (and linux), the MR has to be up= -dated. -=A0 This can only be done by destroying and recreating the MR. There is no = -H_CALL -=A0 to modify MR size. To find holes in the linux kernel memory layout we h= -ave to -=A0 iterate over the memory sections for recreating a ehea_bmap +- When memory is added / removed to LPAR (and linux), the MR has to be updated. + This can only be done by destroying and recreating the MR. There is no H_CALL + to modify MR size. To find holes in the linux kernel memory layout we have to + iterate over the memory sections for recreating a ehea_bmap (otherwise MR would be bigger then available memory causing the registration to fail) -=2D DLPAR userspace tools, kernel, driver, firmware and HMC are involved in= - that -=A0 process on System p +- DLPAR userspace tools, kernel, driver, firmware and HMC are involved in that + process on System p Memory add: version without a external memory notifier call -=2D new memory used in a transfer_xmit will result in a "ehea_bmap +- new memory used in a transfer_xmit will result in a "ehea_bmap translation miss", which triggers a rebuild and reregistration -=A0 of the ehea_bmap based on the current kernel memory setup. -=2D advantage: the number of MR rebuilds is reduced significantly compared = -to + of the ehea_bmap based on the current kernel memory setup. +- advantage: the number of MR rebuilds is reduced significantly compared to a rebuild for each 16MB chunk of memory added. Memory add: version with external notifier call: -=2D We still need a ehea_bmap (whatever structure it has) +- We still need a ehea_bmap (whatever structure it has) Memory remove with notifier: -=2D We have to rebuild the ehea_bmap instantly to remove the pages that are +- We have to rebuild the ehea_bmap instantly to remove the pages that are no longer available. Without doing that, the firmware (pHYP) cannot remove - that memory from the LPAR. As we don't know if or how many additional=20 - sections are to be removed before the DLPAR user space tool tells the=20 + that memory from the LPAR. As we don't know if or how many additional + sections are to be removed before the DLPAR user space tool tells the firmware to remove the memory, we can't wait with the rebuild. @@ -66,33 +60,32 @@ Our current understanding about the current Memory Hotplug System are (please correct me if I'm wrong): -=2D depends on sparse mem -=2D only whole memory sections are added / removed -=2D for each section a memory resource is registered +- depends on sparse mem +- only whole memory sections are added / removed +- for each section a memory resource is registered -=46rom the driver side we need: -=2D some kind of memory notification mechanism. -=A0 For memory add we can live without any external memory notification +>From the driver side we need: +- some kind of memory notification mechanism. + For memory add we can live without any external memory notification event. For memory remove we do need an external trigger (see explanation above). -=2D a way to iterate over all kernel pages and a way to detect holes in the +- a way to iterate over all kernel pages and a way to detect holes in the kernel memory layout in order to build up our own ehea_bmap. Memory notification trigger: -=2D These triggers exist, an exported "register_memory_notifier" / -=A0 "unregister_memory_notifier" would work in this scheme +- These triggers exist, an exported "register_memory_notifier" / + "unregister_memory_notifier" would work in this scheme -=46unctions to use while building ehea_bmap + MRs: -=2D Use either the functions that are used by the memory hotplug system as +Functions to use while building ehea_bmap + MRs: +- Use either the functions that are used by the memory hotplug system as well, that means using the section defines + functions (section_nr_to_pfn, -=A0 pfn_valid) -=2D Use currently other not exported functions in kernel/resource.c, like - walk_memory_resource (where we would still need the maximum possible numb= -er + pfn_valid) +- Use currently other not exported functions in kernel/resource.c, like + walk_memory_resource (where we would still need the maximum possible number of pages NR_MEM_SECTIONS) -=2D Maybe some kind of new interface? +- Maybe some kind of new interface? What would you suggest? diff --git a/a/content_digest b/N2/content_digest index 0ebf518..8cd82b6 100644 --- a/a/content_digest +++ b/N2/content_digest @@ -34,50 +34,44 @@ "Explanation of the current eHEA memory add / remove concept:\n" "\n" "Constraints imposed by HW / FW:\n" - "=2D eHEA has own MMU\n" - "=2D eHEA =A0Memory Regions (MRs) are used by the eHEA MMU =A0to translate v=\n" - "irtual\n" - "=A0 addresses to absolute addresses (like DMA mapped memory on a PCI bus)\n" - "=2D The number of MRs is limited (not enough to have one MR per packet)\n" - "=2D Registration of MRs is comparativley slow as done via slow firmware call\n" + "- eHEA has own MMU\n" + "- eHEA \302\240Memory Regions (MRs) are used by the eHEA MMU \302\240to translate virtual\n" + "\302\240 addresses to absolute addresses (like DMA mapped memory on a PCI bus)\n" + "- The number of MRs is limited (not enough to have one MR per packet)\n" + "- Registration of MRs is comparativley slow as done via slow firmware call\n" "(H_CALL)\n" - "=2D MRs can have a maximum size of the memory available under linux\n" - "=2D MRs cover a contiguous virtual memory block (no holes)\n" + "- MRs can have a maximum size of the memory available under linux\n" + "- MRs cover a contiguous virtual memory block (no holes)\n" "\n" "Because of this there is just one big MR that covers entire kernel memory.\n" "We also need a mapping table from kernel addresses to this\n" "contiguous \"virtual memory IO space\" (here called ehea_bmap).\n" "\n" - "=2D When memory is added / removed to LPAR (and linux), the MR has to be up=\n" - "dated.\n" - "=A0 This can only be done by destroying and recreating the MR. There is no =\n" - "H_CALL\n" - "=A0 to modify MR size. To find holes in the linux kernel memory layout we h=\n" - "ave to\n" - "=A0 iterate over the memory sections for recreating a ehea_bmap\n" + "- When memory is added / removed to LPAR (and linux), the MR has to be updated.\n" + "\302\240 This can only be done by destroying and recreating the MR. There is no H_CALL\n" + "\302\240 to modify MR size. To find holes in the linux kernel memory layout we have to\n" + "\302\240 iterate over the memory sections for recreating a ehea_bmap\n" " (otherwise MR would be bigger then available memory causing the\n" " registration to fail)\n" "\n" - "=2D DLPAR userspace tools, kernel, driver, firmware and HMC are involved in=\n" - " that\n" - "=A0 process on System p\n" + "- DLPAR userspace tools, kernel, driver, firmware and HMC are involved in that\n" + "\302\240 process on System p\n" "\n" "Memory add: version without a external memory notifier call\n" - "=2D new memory used in a transfer_xmit will result in a \"ehea_bmap\n" + "- new memory used in a transfer_xmit will result in a \"ehea_bmap\n" " translation miss\", which triggers a rebuild and reregistration\n" - "=A0 of the ehea_bmap based on the current kernel memory setup.\n" - "=2D advantage: the number of MR rebuilds is reduced significantly compared =\n" - "to\n" + "\302\240 of the ehea_bmap based on the current kernel memory setup.\n" + "- advantage: the number of MR rebuilds is reduced significantly compared to\n" " a rebuild for each 16MB chunk of memory added.\n" "\n" "Memory add: version with external notifier call:\n" - "=2D We still need a ehea_bmap (whatever structure it has)\n" + "- We still need a ehea_bmap (whatever structure it has)\n" "\n" "Memory remove with notifier:\n" - "=2D We have to rebuild the ehea_bmap instantly to remove the pages that are\n" + "- We have to rebuild the ehea_bmap instantly to remove the pages that are\n" " no longer available. Without doing that, the firmware (pHYP) cannot remove\n" - " that memory from the LPAR. As we don't know if or how many additional=20\n" - " sections are to be removed before the DLPAR user space tool tells the=20\n" + " that memory from the LPAR. As we don't know if or how many additional \n" + " sections are to be removed before the DLPAR user space tool tells the \n" " firmware to remove the memory, we can't wait with the rebuild.\n" "\n" "\n" @@ -85,37 +79,36 @@ "(please correct me\n" "if I'm wrong):\n" "\n" - "=2D depends on sparse mem\n" - "=2D only whole memory sections are added / removed\n" - "=2D for each section a memory resource is registered\n" + "- depends on sparse mem\n" + "- only whole memory sections are added / removed\n" + "- for each section a memory resource is registered\n" "\n" "\n" - "=46rom the driver side we need:\n" - "=2D some kind of memory notification mechanism.\n" - "=A0 For memory add we can live without any external memory notification\n" + ">From the driver side we need:\n" + "- some kind of memory notification mechanism.\n" + "\302\240 For memory add we can live without any external memory notification\n" " event. For memory remove we do need an external trigger (see explanation\n" " above).\n" - "=2D a way to iterate over all kernel pages and a way to detect holes in the\n" + "- a way to iterate over all kernel pages and a way to detect holes in the\n" " kernel memory layout in order to build up our own ehea_bmap.\n" "\n" "\n" "Memory notification trigger:\n" - "=2D These triggers exist, an exported \"register_memory_notifier\" /\n" - "=A0 \"unregister_memory_notifier\" would work in this scheme\n" + "- These triggers exist, an exported \"register_memory_notifier\" /\n" + "\302\240 \"unregister_memory_notifier\" would work in this scheme\n" "\n" - "=46unctions to use while building ehea_bmap + MRs:\n" - "=2D Use either the functions that are used by the memory hotplug system as\n" + "Functions to use while building ehea_bmap + MRs:\n" + "- Use either the functions that are used by the memory hotplug system as\n" " well, that means using the section defines + functions (section_nr_to_pfn,\n" - "=A0 pfn_valid)\n" - "=2D Use currently other not exported functions in kernel/resource.c, like\n" - " walk_memory_resource (where we would still need the maximum possible numb=\n" - "er\n" + "\302\240 pfn_valid)\n" + "- Use currently other not exported functions in kernel/resource.c, like\n" + " walk_memory_resource (where we would still need the maximum possible number\n" " of pages NR_MEM_SECTIONS)\n" - "=2D Maybe some kind of new interface?\n" + "- Maybe some kind of new interface?\n" "\n" "What would you suggest?\n" "\n" "Regards,\n" Jan-Bernd & Christoph -0e3dd35328cab4f06248746293b4eb3629329b7c3c1f70c4a339131e2e7b359e +08cc3808333eb42af36d10c8ddf444b2924e9b84c87bb6c4cb5ab7f1173908f8
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.