- 22 Mar, 2017 40 commits
-
-
Mauricio Faria de Oliveira authored
[ Upstream commit 25cdb645 ] The WRITE_SAME commands are not present in the blk_default_cmd_filter write_ok list, and thus are failed with -EPERM when the SG_IO ioctl() is executed without CAP_SYS_RAWIO capability (e.g., unprivileged users). [ sg_io() -> blk_fill_sghdr_rq() > blk_verify_command() -> -EPERM ] The problem can be reproduced with the sg_write_same command # sg_write_same --num 1 --xferlen 512 /dev/sda # # capsh --drop=cap_sys_rawio -- -c \ 'sg_write_same --num 1 --xferlen 512 /dev/sda' Write same: pass through os error: Operation not permitted # For comparison, the WRITE_VERIFY command does not observe this problem, since it is in that list: # capsh --drop=cap_sys_rawio -- -c \ 'sg_write_verify --num 1 --ilen 512 --lba 0 /dev/sda' # So, this patch adds the WRITE_SAME commands to the list, in order for the SG_IO ioctl to finish successfully: # capsh --drop=cap_sys_rawio -- -c \ 'sg_write_same --num 1 --xferlen 512 /dev/sda' # That case happens to be exercised by QEMU KVM guests with 'scsi-block' devices (qemu "-device scsi-block" [1], libvirt "<disk type='block' device='lun'>" [2]), which employs the SG_IO ioctl() and runs as an unprivileged user (libvirt-qemu). In that scenario, when a filesystem (e.g., ext4) performs its zero-out calls, which are translated to write-same calls in the guest kernel, and then into SG_IO ioctls to the host kernel, SCSI I/O errors may be observed in the guest: [...] sd 0:0:0:0: [sda] tag#0 FAILED Result: hostbyte=DID_OK driverbyte=DRIVER_SENSE [...] sd 0:0:0:0: [sda] tag#0 Sense Key : Aborted Command [current] [...] sd 0:0:0:0: [sda] tag#0 Add. Sense: I/O process terminated [...] sd 0:0:0:0: [sda] tag#0 CDB: Write Same(10) 41 00 01 04 e0 78 00 00 08 00 [...] blk_update_request: I/O error, dev sda, sector 17096824 Links: [1] http://git.qemu.org/?p=qemu.git;a=commit;h=336a6915bc7089fb20fea4ba99972ad9a97c5f52 [2] https://libvirt.org/formatdomain.html#elementsDisks (see 'disk' -> 'device') Signed-off-by:
Mauricio Faria de Oliveira <mauricfo@linux.vnet.ibm.com> Signed-off-by:
Brahadambal Srinivasan <latha@linux.vnet.ibm.com> Reported-by:
Manjunatha H R <manjuhr1@in.ibm.com> Reviewed-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
Jens Axboe <axboe@fb.com> Signed-off-by:
Sasha Levin <alexander.levin@verizon.com> Signed-off-by:
Greg Kroah-Hartman <gregkh@linuxfoundation.org>
-
Ben Skeggs authored
[ Upstream commit 2a32b9b1 ] Signed-off-by:
Ben Skeggs <bskeggs@redhat.com> Signed-off-by:
-
Ben Skeggs authored
[ Upstream commit 4391d7f5 ] GP102/GP104 make life difficult by redefining the channel indices for some registers, but not others. Signed-off-by:
-
Ben Skeggs authored
[ Upstream commit e50fcff1 ] Signed-off-by:
-
Alexey Kardashevskiy authored
[ Upstream commit d9c72894 ] We are going to allow the userspace to configure container in one memory context and pass container fd to another so we are postponing memory allocations accounted against the locked memory limit. One of previous patches took care of it_userspace. At the moment we create the default DMA window when the first group is attached to a container; this is done for the userspace which is not DDW-aware but familiar with the SPAPR TCE IOMMU v2 in the part of memory pre-registration - such client expects the default DMA window to exist. This postpones the default DMA window allocation till one of the folliwing happens: 1. first map/unmap request arrives; 2. new window is requested; This adds noop for the case when the userspace requested removal of the default window which has not been created yet. Signed-off-by:
Alexey Kardashevskiy <aik@ozlabs.ru> Reviewed-by:
David Gibson <david@gibson.dropbear.id.au> Acked-by:
Alex Williamson <alex.williamson@redhat.com> Signed-off-by:
Michael Ellerman <mpe@ellerman.id.au> Signed-off-by:
-
Alexey Kardashevskiy authored
[ Upstream commit 6f01cc69 ] There is already a helper to create a DMA window which does allocate a table and programs it to the IOMMU group. However tce_iommu_take_ownership_ddw() did not use it and did these 2 calls itself to simplify error path. Since we are going to delay the default window creation till the default window is accessed/removed or new window is added, we need a helper to create a default window from all these cases. This adds tce_iommu_create_default_window(). Since it relies on a VFIO container to have at least one IOMMU group (for future use), this changes tce_iommu_attach_group() to add a group to the container first and then call the new helper. Signed-off-by:
-
Alexey Kardashevskiy authored
[ Upstream commit 4b6fad70 ] At the moment the userspace tool is expected to request pinning of the entire guest RAM when VFIO IOMMU SPAPR v2 driver is present. When the userspace process finishes, all the pinned pages need to be put; this is done as a part of the userspace memory context (MM) destruction which happens on the very last mmdrop(). This approach has a problem that a MM of the userspace process may live longer than the userspace process itself as kernel threads use userspace process MMs which was runnning on a CPU where the kernel thread was scheduled to. If this happened, the MM remains referenced until this exact kernel thread wakes up again and releases the very last reference to the MM, on an idle system this can take even hours. This moves preregistered regions tracking from MM to VFIO; insteads of using mm_iommu_table_group_mem_t::used, tce_container::prereg_list is added so each container releases regions which it has pre-registered. This changes the userspace interface to return EBUSY if a memory region is already registered in a container. However it should not have any practical effect as the only userspace tool available now does register memory region once per container anyway. As tce_iommu_register_pages/tce_iommu_unregister_pages are called under container->lock, this does not need additional locking. Signed-off-by:
Nicholas Piggin <npiggin@gmail.com> Acked-by:
-
Alexey Kardashevskiy authored
[ Upstream commit bc82d122 ] In some situations the userspace memory context may live longer than the userspace process itself so if we need to do proper memory context cleanup, we better have tce_container take a reference to mm_struct and use it later when the process is gone (@current or @current->mm is NULL). This references mm and stores the pointer in the container; this is done in a new helper - tce_iommu_mm_set() - when one of the following happens: - a container is enabled (IOMMU v1); - a first attempt to pre-register memory is made (IOMMU v2); - a DMA window is created (IOMMU v2). The @mm stays referenced till the container is destroyed. This replaces current->mm with container->mm everywhere except debug prints. This adds a check that current->mm is the same as the one stored in the container to prevent userspace from making changes to a memory context of other processes. DMA map/unmap ioctls() do not check for @mm as they already check for @enabled which is set after tce_iommu_mm_set() is called. This does not reference a task as multiple threads within the same mm are allowed to ioctl() to vfio and supposedly they will have same limits and capabilities and if they do not, we'll just fail with no harm made. Signed-off-by:
-
Alexey Kardashevskiy authored
[ Upstream commit d7baee69 ] This changes mm_iommu_xxx helpers to take mm_struct as a parameter instead of getting it from @current which in some situations may not have a valid reference to mm. This changes helpers to receive @mm and moves all references to @current to the caller, including checks for !current and !current->mm; checks in mm_iommu_preregistered() are removed as there is no caller yet. This moves the mm_iommu_adjust_locked_vm() call to the caller as it receives mm_iommu_table_group_mem_t but it needs mm. This should cause no behavioral change. Signed-off-by:
-
Alexey Kardashevskiy authored
[ Upstream commit 88f54a35 ] We are going to get rid of @current references in mmu_context_boos3s64.c and cache mm_struct in the VFIO container. Since mm_context_t does not have reference counting, we will be using mm_struct which does have the reference counter. This changes mm_iommu_init/mm_iommu_cleanup to receive mm_struct rather than mm_context_t (which is embedded into mm). This should not cause any behavioral change. Signed-off-by:
-
Alexey Kardashevskiy authored
[ Upstream commit 39701e56 ] The iommu_table struct manages a hardware TCE table and a vmalloc'd table with corresponding userspace addresses. Both are allocated when the default DMA window is created and this happens when the very first group is attached to a container. As we are going to allow the userspace to configure container in one memory context and pas container fd to another, we have to postpones such allocations till a container fd is passed to the destination user process so we would account locked memory limit against the actual container user constrainsts. This postpones the it_userspace array allocation till it is used first time for mapping. The unmapping patch already checks if the array is allocated. Signed-off-by:
-
Vitaly Kuznetsov authored
[ Upstream commit fa32ff65 ] With wrap around mappings in place we can always provide drivers with direct links to packets on the ring buffer, even when they wrap around. Do the required updates to get_next_pkt_raw()/put_pkt_raw() The first version of this commit was reverted (65a532f3) to deal with cross-tree merge issues which are (hopefully) resolved now. Signed-off-by:
Vitaly Kuznetsov <vkuznets@redhat.com> Signed-off-by:
K. Y. Srinivasan <kys@microsoft.com> Tested-by:
Dexuan Cui <decui@microsoft.com> Signed-off-by:
-
Thomas Falcon authored
[ Upstream commit 94acf164 ] Include calculations to compute the number of segments that comprise an aggregated large packet. Signed-off-by:
Thomas Falcon <tlfalcon@linux.vnet.ibm.com> Reviewed-by:
Marcelo Ricardo Leitner <marcelo.leitner@gmail.com> Reviewed-by:
Jonathan Maxwell <jmaxwell37@gmail.com> Signed-off-by:
David S. Miller <davem@davemloft.net> Signed-off-by:
-
Gavin Shan authored
[ Upstream commit f40ec3c7 ] Previously we enabled VFs and enable their memory space before calling pcibios_sriov_enable(). But pcibios_sriov_enable() may update the VF BARs: for example, on PPC PowerNV we may change them to manage the association of VFs to PEs. Because 64-bit BARs cannot be updated atomically, it's unsafe to update them while they're enabled. The half-updated state may conflict with other devices in the system. Call pcibios_sriov_enable() before enabling the VFs so any BAR updates happen while the VF BARs are disabled. [bhelgaas: changelog] Tested-by:
Carol Soto <clsoto@us.ibm.com> Signed-off-by:
Gavin Shan <gwshan@linux.vnet.ibm.com> Signed-off-by:
Bjorn Helgaas <bhelgaas@google.com> Signed-off-by:
-
Bjorn Helgaas authored
[ Upstream commit 63880b23 ] VF BARs are read-only zero, so updating VF BARs will not have any effect. See the SR-IOV spec r1.1, sec 3.4.1.11. We already ignore these updates because of 70675e0b ("PCI: Don't try to restore VF BARs"); this merely restructures it slightly to make it easier to split updates for standard and SR-IOV BARs. Signed-off-by:
-
Bjorn Helgaas authored
[ Upstream commit 45d004f4 ] The BAR property bits (0-3 for memory BARs, 0-1 for I/O BARs) are supposed to be read-only, but we do save them in res->flags and include them when updating the BAR. Mask the I/O property bits with ~PCI_BASE_ADDRESS_IO_MASK (0x3) instead of PCI_REGION_FLAG_MASK (0xf) to make it obvious that we can't corrupt bits 2-3 of I/O addresses. Use PCI_ROM_ADDRESS_MASK for ROM BARs. This means we'll only check the top 21 bits (instead of the 28 bits we used to check) of a ROM BAR to see if the update was successful. Signed-off-by:
-
Bjorn Helgaas authored
[ Upstream commit 546ba9f8 ] If we update a VF BAR while it's enabled, there are two potential problems: 1) Any driver that's using the VF has a cached BAR value that is stale after the update, and 2) We can't update 64-bit BARs atomically, so the intermediate state (new lower dword with old upper dword) may conflict with another device, and an access by a driver unrelated to the VF may cause a bus error. Warn about attempts to update VF BARs while they are enabled. This is a programming error, so use dev_WARN() to get a backtrace. Signed-off-by:
-
Bjorn Helgaas authored
[ Upstream commit 7a6d312b ] Remove the assumption that IORESOURCE_ROM_ENABLE == PCI_ROM_ADDRESS_ENABLE. PCI_ROM_ADDRESS_ENABLE is the ROM enable bit defined by the PCI spec, so if we're reading or writing a BAR register value, that's what we should use. IORESOURCE_ROM_ENABLE is a corresponding bit in struct resource flags. Signed-off-by:
-
Bjorn Helgaas authored
[ Upstream commit 0b457dde ] pci_update_resource() updates a hardware BAR so its address matches the kernel's struct resource UNLESS it's a disabled ROM BAR. We only update those when we enable the ROM. It's not obvious from the code why ROM BARs should be handled specially. Apparently there are Matrox devices with defective ROM BARs that read as zero when disabled. That means that if pci_enable_rom() reads the disabled BAR, sets PCI_ROM_ADDRESS_ENABLE (without re-inserting the address), and writes it back, it would enable the ROM at address zero. Add comments and references to explain why we can't make the code look more rational. The code changes are from 755528c8 ("Ignore disabled ROM resources at setup") and 8085ce08 ("[PATCH] Fix PCI ROM mapping"). Link: https://lkml.org/lkml/2005/8/30/138Signed-off-by:
-
Bjorn Helgaas authored
[ Upstream commit 286c2378 ] pci_std_update_resource() only deals with standard BARs, so we don't have to worry about the complications of VF BARs in an SR-IOV capability. Compute the BAR address inline and remove pci_resource_bar(). That makes pci_iov_resource_bar() unused, so remove that as well. Signed-off-by:
-
Bjorn Helgaas authored
[ Upstream commit 6ffa2489 ] Previously pci_update_resource() used the same code path for updating standard BARs and VF BARs in SR-IOV capabilities. Split the VF BAR update into a new pci_iov_update_resource() internal interface, which makes it simpler to compute the BAR address (we can get rid of pci_resource_bar() and pci_iov_resource_bar()). This patch: - Renames pci_update_resource() to pci_std_update_resource(), - Adds pci_iov_update_resource(), - Makes pci_update_resource() a wrapper that calls the appropriate one, No functional change intended. Signed-off-by:
-
Vitaly Kuznetsov authored
[ Upstream commit 59107e2f ] There is a feature in Hyper-V ('Debug-VM --InjectNonMaskableInterrupt') which injects NMI to the guest. We may want to crash the guest and do kdump on this NMI by enabling unknown_nmi_panic. To make kdump succeed we need to allow the kdump kernel to re-establish VMBus connection so it will see VMBus devices (storage, network,..). To properly unload VMBus making it possible to start over during kdump we need to do the following: - Send an 'unload' message to the hypervisor. This can be done on any CPU so we do this the crashing CPU. - Receive the 'unload finished' reply message. WS2012R2 delivers this message to the CPU which was used to establish VMBus connection during module load and this CPU may differ from the CPU sending 'unload'. Receiving a VMBus message means the following: - There is a per-CPU slot in memory for one message. This slot can in theory be accessed by any CPU. - We get an interrupt on the CPU when a message was placed into the slot. - When we read the message we need to clear the slot and signal the fact to the hypervisor. In case there are more messages to this CPU pending the hypervisor will deliver the next message. The signaling is done by writing to an MSR so this can only be done on the appropriate CPU. To avoid doing cross-CPU work on crash we have vmbus_wait_for_unload() function which checks message slots for all CPUs in a loop waiting for the 'unload finished' messages. However, there is an issue which arises when these conditions are met: - We're crashing on a CPU which is different from the one which was used to initially contact the hypervisor. - The CPU which was used for the initial contact is blocked with interrupts disabled and there is a message pending in the message slot. In this case we won't be able to read the 'unload finished' message on the crashing CPU. This is reproducible when we receive unknown NMIs on all CPUs simultaneously: the first CPU entering panic() will proceed to crash and all other CPUs will stop themselves with interrupts disabled. The suggested solution is to handle unknown NMIs for Hyper-V guests on the first CPU which gets them only. This will allow us to rely on VMBus interrupt handler being able to receive the 'unload finish' message in case it is delivered to a different CPU. The issue is not reproducible on WS2016 as Debug-VM delivers NMI to the boot CPU only, WS2012R2 and earlier Hyper-V versions are affected. Signed-off-by:
Thomas Gleixner <tglx@linutronix.de> Signed-off-by:
Ingo Molnar <mingo@kernel.org> Signed-off-by:
-
Michael Cyr authored
[ Upstream commit 8bf11557 ] This patch adds code to disconnect from the client, which will make sure any outstanding commands have been completed, before continuing on with the remove operation. Signed-off-by:
Michael Cyr <mikecyr@us.ibm.com> Signed-off-by:
Bryant G. Ly <bryantly@linux.vnet.ibm.com> Tested-by:
Steven Royer <seroyer@linux.vnet.ibm.com> Signed-off-by:
Martin K. Petersen <martin.petersen@oracle.com> Signed-off-by:
-
Michael Cyr authored
[ Upstream commit c9b3379f ] This patch changes the way the IBM vSCSI server driver manages its Command/Response Queue (CRQ). We used to register the CRQ with phyp at probe time. Now we wait until tpg_enable_store. Similarly, when tpg_enable_store is called to "disable" (i.e. the stored value is 0), we unregister the queue with phyp. One consquence to this is that we have no need for the PART_UP_WAIT_ENAB state, since we can't get an Init Message from the client in our CRQ if we're waiting to be enabled, since we haven't registered the queue yet. Signed-off-by:
-
Michael Cyr authored
[ Upstream commit 79fac9c9 ] This patch reorders functions in a manner necessary for a follow-on patch. It also makes some minor styling changes (mostly removing extra spaces) and fixes some typos. There are no code changes in this patch, with one exception: due to the reordering of the functions, I needed to explicitly declare a function at the top of the file. However, this will be removed in the next patch, since the code requiring the predeclaration will be removed. Signed-off-by:
-
Michael Cyr authored
[ Upstream commit 7435b32e ] Signed-off-by:
-
Michael Cyr authored
[ Upstream commit 9c93cf03 ] Signed-off-by:
-
Michael Cyr authored
[ Upstream commit 11950d70 ] Signed-off-by:
-
Todd Fujinaka authored
[ Upstream commit 5bc8c230 ] i210 and i211 share the same PHY but have different PCI IDs. Don't forget i211 for any i210 workarounds. Signed-off-by:
Todd Fujinaka <todd.fujinaka@intel.com> Tested-by:
Aaron Brown <aaron.f.brown@intel.com> Signed-off-by:
Jeff Kirsher <jeffrey.t.kirsher@intel.com> Signed-off-by:
-
Chris J Arges authored
[ Upstream commit 4e684f59 ] Sometimes firmware may not properly initialize I347AT4_PAGE_SELECT causing the probe of an igb i210 NIC to fail. This patch adds an addition zeroing of this register during igb_get_phy_id to workaround this issue. Thanks for Jochen Henneberg for the idea and original patch. Signed-off-by:
Chris J Arges <christopherarges@gmail.com> Tested-by:
-
Dan Streetman authored
[ Upstream commit c74fd80f ] Revert the main part of commit: af42b8d1 ("xen: fix MSI setup and teardown for PV on HVM guests") That commit introduced reading the pci device's msi message data to see if a pirq was previously configured for the device's msi/msix, and re-use that pirq. At the time, that was the correct behavior. However, a later change to Qemu caused it to call into the Xen hypervisor to unmap all pirqs for a pci device, when the pci device disables its MSI/MSIX vectors; specifically the Qemu commit: c976437c7dba9c7444fb41df45468968aaa326ad ("qemu-xen: free all the pirqs for msi/msix when driver unload") Once Qemu added this pirq unmapping, it was no longer correct for the kernel to re-use the pirq number cached in the pci device msi message data. All Qemu releases since 2.1.0 contain the patch that unmaps the pirqs when the pci device disables its MSI/MSIX vectors. This bug is causing failures to initialize multiple NVMe controllers under Xen, because the NVMe driver sets up a single MSIX vector for each controller (concurrently), and then after using that to talk to the controller for some configuration data, it disables the single MSIX vector and re-configures all the MSIX vectors it needs. So the MSIX setup code tries to re-use the cached pirq from the first vector for each controller, but the hypervisor has already given away that pirq to another controller, and its initialization fails. This is discussed in more detail at: https://lists.xen.org/archives/html/xen-devel/2017-01/msg00447.html Fixes: af42b8d1 ("xen: fix MSI setup and teardown for PV on HVM guests") Signed-off-by:
Dan Streetman <dan.streetman@canonical.com> Reviewed-by:
Stefano Stabellini <sstabellini@kernel.org> Acked-by:
Konrad Rzeszutek Wilk <konrad.wilk@oracle.com> Signed-off-by:
Boris Ostrovsky <boris.ostrovsky@oracle.com> Signed-off-by:
-
Krister Johansen authored
commit 21d25f6a upstream. On a kernel with DEBUG_LOCKS, ioat_free_chan_resources triggers an in_interrupt() warning. With PROVE_LOCKING, it reports detecting a SOFTIRQ-safe to SOFTIRQ-unsafe lock ordering in the same code path. This is because dma_generic_alloc_coherent() checks if the GFP flags permit blocking. It allocates from different subsystems if blocking is permitted. The free path knows how to return the memory to the correct allocator. If GFP_KERNEL is specified then the alloc and free end up going through cma_alloc(), which uses mutexes. Given that ioat_free_chan_resources() can be called in interrupt context, ioat_alloc_chan_resources() must specify GFP_NOWAIT so that the allocations do not block and instead use an allocator that uses spinlocks. Signed-off-by:
Krister Johansen <kjlx@templeofstupid.com> Acked-by:
Dave Jiang <dave.jiang@intel.com> Signed-off-by:
Vinod Koul <vinod.koul@intel.com> Signed-off-by:
-
Daniel Borkmann authored
[ Upstream commit 6760bf2d ] Martin reported a verifier issue that hit the BUG_ON() for his test case in the mark_reg_unknown_value() function: [ 202.861380] kernel BUG at kernel/bpf/verifier.c:467! [...] [ 203.291109] Call Trace: [ 203.296501] [<ffffffff811364d5>] mark_map_reg+0x45/0x50 [ 203.308225] [<ffffffff81136558>] mark_map_regs+0x78/0x90 [ 203.320140] [<ffffffff8113938d>] do_check+0x226d/0x2c90 [ 203.331865] [<ffffffff8113a6ab>] bpf_check+0x48b/0x780 [ 203.343403] [<ffffffff81134c8e>] bpf_prog_load+0x27e/0x440 [ 203.355705] [<ffffffff8118a38f>] ? handle_mm_fault+0x11af/0x1230 [ 203.369158] [<ffffffff812d8188>] ? security_capable+0x48/0x60 [ 203.382035] [<ffffffff811351a4>] SyS_bpf+0x124/0x960 [ 203.393185] [<ffffffff810515f6>] ? __do_page_fault+0x276/0x490 [ 203.406258] [<ffffffff816db320>] entry_SYSCALL_64_fastpath+0x13/0x94 This issue got uncovered after the fix in a08dd0da ("bpf: fix regression on verifier pruning wrt map lookups"). The reason why it wasn't noticed before was, because as mentioned in a08dd0da, mark_map_regs() was doing the id matching incorrectly based on the uncached regs[regno].id. So, in the first loop, we walked all regs and as soon as we found regno == i, then this reg's id was cleared when calling mark_reg_unknown_value() thus that every subsequent register was probed against id of 0 (which, in combination with the PTR_TO_MAP_VALUE_OR_NULL type is an invalid condition that no other register state can hold), and therefore wasn't type transitioned such as in the spilled register case for the second loop. Now since that got fixed, it turned out that 57a09bf0 ("bpf: Detect identical PTR_TO_MAP_VALUE_OR_NULL registers") used mark_reg_unknown_value() incorrectly for the spilled regs, and thus hitting the BUG_ON() in some cases due to regno >= MAX_BPF_REG. Although spilled regs have the same type as the non-spilled regs for the verifier state, that is, struct bpf_reg_state, they are semantically different from the non-spilled regs. In other words, there can be up to 64 (MAX_BPF_STACK / BPF_REG_SIZE) spilled regs in the stack, for example, register R<x> could have been spilled by the program to stack location X, Y, Z, and in mark_map_regs() we need to scan these stack slots of type STACK_SPILL for potential registers that we have to transition from PTR_TO_MAP_VALUE_OR_NULL. Therefore, depending on the location, the spilled_regs regno can be a lot higher than just MAX_BPF_REG's value since we operate on stack instead. The reset in mark_reg_unknown_value() itself is just fine, only that the BUG_ON() was inappropriate for this. Fix it by making a __mark_reg_unknown_value() version that can be called from mark_map_reg() generically; we know for the non-spilled case that the regno is always < MAX_BPF_REG anyway. Fixes: 57a09bf0 ("bpf: Detect identical PTR_TO_MAP_VALUE_OR_NULL registers") Reported-by:
Martin KaFai Lau <kafai@fb.com> Signed-off-by:
Daniel Borkmann <daniel@iogearbox.net> Acked-by:
Alexei Starovoitov <ast@kernel.org> Signed-off-by:
-
Daniel Borkmann authored
[ Upstream commit a08dd0da ] Commit 57a09bf0 ("bpf: Detect identical PTR_TO_MAP_VALUE_OR_NULL registers") introduced a regression where existing programs stopped loading due to reaching the verifier's maximum complexity limit, whereas prior to this commit they were loading just fine; the affected program has roughly 2k instructions. What was found is that state pruning couldn't be performed effectively anymore due to mismatches of the verifier's register state, in particular in the id tracking. It doesn't mean that 57a09bf0 is incorrect per se, but rather that verifier needs to perform a lot more work for the same program with regards to involved map lookups. Since commit 57a09bf0 is only about tracking registers with type PTR_TO_MAP_VALUE_OR_NULL, the id is only needed to follow registers until they are promoted through pattern matching with a NULL check to either PTR_TO_MAP_VALUE or UNKNOWN_VALUE type. After that point, the id becomes irrelevant for the transitioned types. For UNKNOWN_VALUE, id is already reset to 0 via mark_reg_unknown_value(), but not so for PTR_TO_MAP_VALUE where id is becoming stale. It's even transferred further into other types that don't make use of it. Among others, one example is where UNKNOWN_VALUE is set on function call return with RET_INTEGER return type. states_equal() will then fall through the memcmp() on register state; note that the second memcmp() uses offsetofend(), so the id is part of that since d2a4dd37 ("bpf: fix state equivalence"). But the bisect pointed already to 57a09bf0, where we really reach beyond complexity limit. What I found was that states_equal() often failed in this case due to id mismatches in spilled regs with registers in type PTR_TO_MAP_VALUE. Unlike non-spilled regs, spilled regs just perform a memcmp() on their reg state and don't have any other optimizations in place, therefore also id was relevant in this case for making a pruning decision. We can safely reset id to 0 as well when converting to PTR_TO_MAP_VALUE. For the affected program, it resulted in a ~17 fold reduction of complexity and let the program load fine again. Selftest suite also runs fine. The only other place where env->id_gen is used currently is through direct packet access, but for these cases id is long living, thus a different scenario. Also, the current logic in mark_map_regs() is not fully correct when marking NULL branch with UNKNOWN_VALUE. We need to cache the destination reg's id in any case. Otherwise, once we marked that reg as UNKNOWN_VALUE, it's id is reset and any subsequent registers that hold the original id and are of type PTR_TO_MAP_VALUE_OR_NULL won't be marked UNKNOWN_VALUE anymore, since mark_map_reg() reuses the uncached regs[regno].id that was just overridden. Note, we don't need to cache it outside of mark_map_regs(), since it's called once on this_branch and the other time on other_branch, which are both two independent verifier states. A test case for this is added here, too. Fixes: 57a09bf0 ("bpf: Detect identical PTR_TO_MAP_VALUE_OR_NULL registers") Signed-off-by:
Thomas Graf <tgraf@suug.ch> Acked-by:
-
Alexei Starovoitov authored
[ Upstream commit d2a4dd37 ] Commmits 57a09bf0 ("bpf: Detect identical PTR_TO_MAP_VALUE_OR_NULL registers") and 48461135 ("bpf: allow access into map value arrays") by themselves are correct, but in combination they make state equivalence ignore 'id' field of the register state which can lead to accepting invalid program. Fixes: 57a09bf0 ("bpf: Detect identical PTR_TO_MAP_VALUE_OR_NULL registers") Fixes: 48461135 ("bpf: allow access into map value arrays") Signed-off-by:
-
Thomas Graf authored
[ Upstream commit 57a09bf0 ] A BPF program is required to check the return register of a map_elem_lookup() call before accessing memory. The verifier keeps track of this by converting the type of the result register from PTR_TO_MAP_VALUE_OR_NULL to PTR_TO_MAP_VALUE after a conditional jump ensures safety. This check is currently exclusively performed for the result register 0. In the event the compiler reorders instructions, BPF_MOV64_REG instructions may be moved before the conditional jump which causes them to keep their type PTR_TO_MAP_VALUE_OR_NULL to which the verifier objects when the register is accessed: 0: (b7) r1 = 10 1: (7b) *(u64 *)(r10 -8) = r1 2: (bf) r2 = r10 3: (07) r2 += -8 4: (18) r1 = 0x59c00000 6: (85) call 1 7: (bf) r4 = r0 8: (15) if r0 == 0x0 goto pc+1 R0=map_value(ks=8,vs=8) R4=map_value_or_null(ks=8,vs=8) R10=fp 9: (7a) *(u64 *)(r4 +0) = 0 R4 invalid mem access 'map_value_or_null' This commit extends the verifier to keep track of all identical PTR_TO_MAP_VALUE_OR_NULL registers after a map_elem_lookup() by assigning them an ID and then marking them all when the conditional jump is observed. Signed-off-by:
Josef Bacik <jbacik@fb.com> Acked-by:
-
Hannes Frederic Sowa authored
[ Upstream commit 72ef9c41 ] This patch fixes a memory leak, which happens if the connection request is not fulfilled between parsing the DCCP options and handling the SYN (because e.g. the backlog is full), because we forgot to free the list of ack vectors. Reported-by:
Jianwen Ji <jiji@redhat.com> Signed-off-by:
Hannes Frederic Sowa <hannes@stressinduktion.org> Signed-off-by:
-
Hannes Frederic Sowa authored
[ Upstream commit b20e2d54 ] aszlig observed failing ssh tunnels (-w) during initialization since commit cc9da6cc ("ipv6: addrconf: use stable address generator for ARPHRD_NONE"). We already had reports that the mentioned commit breaks Juniper VPN connections. I can't clearly say that the Juniper VPN client has the same problem, but it is worth a try to hint to this patch. Because of the early generation of link local addresses, the kernel now can start asking for routers on the local subnet much earlier than usual. Those router solicitation packets arrive inside the ssh channels and should be transmitted to the tun fd before the configuration scripts might have upped the interface and made it ready for transmission. ssh polls on the interface and receives back a POLL_OUT. It tries to send the earily router solicitation packet to the tun interface. Unfortunately it hasn't been up'ed yet by config scripts, thus failing with -EIO. ssh doesn't retry again and considers the tun interface broken forever. Link: https://bugzilla.kernel.org/show_bug.cgi?id=121131 Fixes: cc9da6cc ("ipv6: addrconf: use stable address generator for ARPHRD_NONE") Cc: Bjørn Mork <bjorn@mork.no> Reported-by:
Valdis Kletnieks <Valdis.Kletnieks@vt.edu> Cc: Valdis Kletnieks <Valdis.Kletnieks@vt.edu> Reported-by:
Jonas Lippuner <jonas@lippuner.ca> Cc: Jonas Lippuner <jonas@lippuner.ca> Reported-by:
aszlig <aszlig@redmoonstudios.org> Cc: aszlig <aszlig@redmoonstudios.org> Signed-off-by:
-
Jon Maxwell authored
[ Upstream commit 45caeaa5 ] As Eric Dumazet pointed out this also needs to be fixed in IPv6. v2: Contains the IPv6 tcp/Ipv6 dccp patches as well. We have seen a few incidents lately where a dst_enty has been freed with a dangling TCP socket reference (sk->sk_dst_cache) pointing to that dst_entry. If the conditions/timings are right a crash then ensues when the freed dst_entry is referenced later on. A Common crashing back trace is: #8 [] page_fault at ffffffff8163e648 [exception RIP: __tcp_ack_snd_check+74] . . #9 [] tcp_rcv_established at ffffffff81580b64 #10 [] tcp_v4_do_rcv at ffffffff8158b54a #11 [] tcp_v4_rcv at ffffffff8158cd02 #12 [] ip_local_deliver_finish at ffffffff815668f4 #13 [] ip_local_deliver at ffffffff81566bd9 #14 [] ip_rcv_finish at ffffffff8156656d #15 [] ip_rcv at ffffffff81566f06 #16 [] __netif_receive_skb_core at ffffffff8152b3a2 #17 [] __netif_receive_skb at ffffffff8152b608 #18 [] netif_receive_skb at ffffffff8152b690 #19 [] vmxnet3_rq_rx_complete at ffffffffa015eeaf [vmxnet3] #20 [] vmxnet3_poll_rx_only at ffffffffa015f32a [vmxnet3] #21 [] net_rx_action at ffffffff8152bac2 #22 [] __do_softirq at ffffffff81084b4f #23 [] call_softirq at ffffffff8164845c #24 [] do_softirq at ffffffff81016fc5 #25 [] irq_exit at ffffffff81084ee5 #26 [] do_IRQ at ffffffff81648ff8 Of course it may happen with other NIC drivers as well. It's found the freed dst_entry here: 224 static bool tcp_in_quickack_mode(struct sock *sk)
↩ 225 {↩ 226 ▹ const struct inet_connection_sock *icsk = inet_csk(sk);↩ 227 ▹ const struct dst_entry *dst = __sk_dst_get(sk);↩ 228↩ 229 ▹ return (dst && dst_metric(dst, RTAX_QUICKACK)) ||↩ 230 ▹ ▹ (icsk->icsk_ack.quick && !icsk->icsk_ack.pingpong);↩ 231 }↩ But there are other backtraces attributed to the same freed dst_entry in netfilter code as well. All the vmcores showed 2 significant clues: - Remote hosts behind the default gateway had always been redirected to a different gateway. A rtable/dst_entry will be added for that host. Making more dst_entrys with lower reference counts. Making this more probable. - All vmcores showed a postitive LockDroppedIcmps value, e.g: LockDroppedIcmps 267 A closer look at the tcp_v4_err() handler revealed that do_redirect() will run regardless of whether user space has the socket locked. This can result in a race condition where the same dst_entry cached in sk->sk_dst_entry can be decremented twice for the same socket via: do_redirect()->__sk_dst_check()-> dst_release(). Which leads to the dst_entry being prematurely freed with another socket pointing to it via sk->sk_dst_cache and a subsequent crash. To fix this skip do_redirect() if usespace has the socket locked. Instead let the redirect take place later when user space does not have the socket locked. The dccp/IPv6 code is very similar in this respect, so fixing it there too. As Eric Garver pointed out the following commit now invalidates routes. Which can set the dst->obsolete flag so that ipv4_dst_check() returns null and triggers the dst_release(). Fixes: ceb33206 ("ipv4: Kill routes during PMTU/redirect updates.") Cc: Eric Garver <egarver@redhat.com> Cc: Hannes Sowa <hsowa@redhat.com> Signed-off-by: -
Florian Westphal authored
[ Upstream commit a13b2082 ] Andreas reports kernel oops during rmmod of the br_netfilter module. Hannes debugged the oops down to a NULL rt6info->rt6i_indev. Problem is that br_netfilter has the nasty concept of adding a fake rtable to skb->dst; this happens in a br_netfilter prerouting hook. A second hook (in bridge LOCAL_IN) is supposed to remove these again before the skb is handed up the stack. However, on module unload hooks get unregistered which means an skb could traverse the prerouting hook that attaches the fake_rtable, while the 'fake rtable remove' hook gets removed from the hooklist immediately after. Fixes: 34666d46 ("netfilter: bridge: move br_netfilter out of the core") Reported-by:
Andreas Karis <akaris@redhat.com> Debugged-by:
Florian Westphal <fw@strlen.de> Acked-by:
Pablo Neira Ayuso <pablo@netfilter.org> Signed-off-by:
-
