For Azure, SGX cgroup support feature is very useful. It is needed to enforce the EPC resource limitation of Kubernetes pods on SGX nodes. Today, in Azure Kubernetes Service, each pod on SGX node claims a nominal EPC memory requirement. K8s will track the unclaimed EPC memories on SGX nodes to schedule pods. However, there's no enforcement on the node whether the pod uses more EPC memory than what it claims. If EPC is running out on the node, the kernel will do EPC paging, which will cause all pods suffering performance degradation. Cgroup support for EPC will enforce EPC resource limitation on pod level, so that when a pod tries to use more EPC than what it claims, it will be EPC paged while other pods are not affected. -Anand From: Anand Krishnamoorthi <anakrish@xxxxxxxxxxxxx> Sent: Monday, April 3, 2023 2:26 PM To: Kristen Carlson Accardi <kristen@xxxxxxxxxxxxxxx>; jarkko@xxxxxxxxxx <jarkko@xxxxxxxxxx>; dave.hansen@xxxxxxxxxxxxxxx <dave.hansen@xxxxxxxxxxxxxxx>; tj@xxxxxxxxxx <tj@xxxxxxxxxx>; linux-kernel@xxxxxxxxxxxxxxx <linux-kernel@xxxxxxxxxxxxxxx>; linux-sgx@xxxxxxxxxxxxxxx <linux-sgx@xxxxxxxxxxxxxxx>; cgroups@xxxxxxxxxxxxxxx <cgroups@xxxxxxxxxxxxxxx>; Bo Zhang (ACC) <zhanb@xxxxxxxxxxxxx> Cc: zhiquan1.li@xxxxxxxxx <zhiquan1.li@xxxxxxxxx> Subject: Re: [EXTERNAL] [PATCH v2 00/18] Add Cgroup support for SGX EPC memory Adding Bo Zhang to thread. -Anand From: Kristen Carlson Accardi <kristen@xxxxxxxxxxxxxxx> Sent: Friday, December 2, 2022 10:36 AM To: jarkko@xxxxxxxxxx <jarkko@xxxxxxxxxx>; dave.hansen@xxxxxxxxxxxxxxx <dave.hansen@xxxxxxxxxxxxxxx>; tj@xxxxxxxxxx <tj@xxxxxxxxxx>; linux-kernel@xxxxxxxxxxxxxxx <linux-kernel@xxxxxxxxxxxxxxx>; linux-sgx@xxxxxxxxxxxxxxx <linux-sgx@xxxxxxxxxxxxxxx>; cgroups@xxxxxxxxxxxxxxx <cgroups@xxxxxxxxxxxxxxx> Cc: zhiquan1.li@xxxxxxxxx <zhiquan1.li@xxxxxxxxx> Subject: [EXTERNAL] [PATCH v2 00/18] Add Cgroup support for SGX EPC memory Utilize the Miscellaneous cgroup controller to regulate the distribution of SGX EPC memory, which is a subset of system RAM that is used to provide SGX-enabled applications with protected memory, and is otherwise inaccessible. SGX EPC memory allocations are separate from normal RAM allocations, and is managed solely by the SGX subsystem. The existing cgroup memory controller cannot be used to limit or account for SGX EPC memory. This patchset implements the support for sgx_epc memory within the misc cgroup controller, and then utilizes the misc cgroup controller to provide support for setting the total system capacity, max limit per cgroup, and events. This work was originally authored by Sean Christopherson a few years ago, and was modified to work with more recent kernels, and to utilize the misc cgroup controller rather than a custom controller. It is currently based on top of the MCA patches. Here's the MCA patchset for reference. https://lore.kernel.org/linux-sgx/2d52c8c4-8ed0-6df2-2911-da5b9fcc9ae4@xxxxxxxxx/T/#t The patchset adds support for multiple LRUs to track both reclaimable EPC pages (i.e. pages the reclaimer knows about), as well as unreclaimable EPC pages (i.e. pages which the reclaimer isn't aware of, such as va pages). These pages are assigned to an LRU, as well as an enclave, so that an enclave's full EPC usage can be tracked, and limited to a max value. During OOM events, an enclave can be have its memory zapped, and all the EPC pages not tracked by the reclaimer can be freed. I appreciate your comments and feedback. Changelog: v2: * rename struct sgx_epc_lru to sgx_epc_lru_lists to be more clear that this struct contains 2 lists. * use inline functions rather than macros for sgx_epc_page_list* wrappers. * Remove flags macros and open code all flags. * Improve the commit message for RECLAIM_IN_PROGRESS patch to make it more clear what the patch does. * remove notifier_block from misc cgroup changes and use a set of ops for callbacks instead. * rename root_misc to misc_cg_root and parent_misc to misc_cg_parent * consolidate misc cgroup changes to 2 patches and remove most of the previous helper functions. Kristen Carlson Accardi (7): x86/sgx: Add 'struct sgx_epc_lru_lists' to encapsulate lru list(s) x86/sgx: Use sgx_epc_lru_lists for existing active page list x86/sgx: Track epc pages on reclaimable or unreclaimable lists cgroup/misc: Add per resource callbacks for css events cgroup/misc: Prepare for SGX usage x86/sgx: Add support for misc cgroup controller Docs/x86/sgx: Add description for cgroup support Sean Christopherson (11): x86/sgx: Call cond_resched() at the end of sgx_reclaim_pages() x86/sgx: Store struct sgx_encl when allocating new VA pages x86/sgx: Introduce RECLAIM_IN_PROGRESS flag for EPC pages x86/sgx: Use a list to track to-be-reclaimed pages during reclaim x86/sgx: Allow reclaiming up to 32 pages, but scan 16 by default x86/sgx: Return the number of EPC pages that were successfully reclaimed x86/sgx: Add option to ignore age of page during EPC reclaim x86/sgx: Prepare for multiple LRUs x86/sgx: Expose sgx_reclaim_pages() for use by EPC cgroup x86/sgx: Add helper to grab pages from an arbitrary EPC LRU x86/sgx: Add EPC OOM path to forcefully reclaim EPC Documentation/x86/sgx.rst | 77 ++++ arch/x86/Kconfig | 13 + arch/x86/kernel/cpu/sgx/Makefile | 1 + arch/x86/kernel/cpu/sgx/encl.c | 90 ++++- arch/x86/kernel/cpu/sgx/encl.h | 4 +- arch/x86/kernel/cpu/sgx/epc_cgroup.c | 539 +++++++++++++++++++++++++++ arch/x86/kernel/cpu/sgx/epc_cgroup.h | 59 +++ arch/x86/kernel/cpu/sgx/ioctl.c | 14 +- arch/x86/kernel/cpu/sgx/main.c | 412 ++++++++++++++++---- arch/x86/kernel/cpu/sgx/sgx.h | 122 +++++- arch/x86/kernel/cpu/sgx/virt.c | 28 +- include/linux/misc_cgroup.h | 35 ++ kernel/cgroup/misc.c | 76 +++- 13 files changed, 1341 insertions(+), 129 deletions(-) create mode 100644 arch/x86/kernel/cpu/sgx/epc_cgroup.c create mode 100644 arch/x86/kernel/cpu/sgx/epc_cgroup.h -- 2.38.1
