Linux : Security Vulnerabilities, CVEs, Published In June 2018

Linux Kernel version 3.18 to 4.16 incorrectly handles an SG_IO ioctl on /dev/sg0 with dxfer_direction=SG_DXFER_FROM_DEV and an empty 6-byte cmdp. This may lead to copying up to 1000 kernel heap pages to the userspace. This has been fixed upstream in https://github.com/torvalds/linux/commit/a45b599ad808c3c982fdcdc12b0b8611c2f92824 already. The problem has limited scope, as users don't usually have permissions to access SCSI devices. On the other hand, e.g. the Nero user manual suggests doing `chmod o+r+w /dev/sg*` to make the devices accessible. NOTE: third parties dispute the relevance of this report, noting that the requirement for an attacker to have both the CAP_SYS_ADMIN and CAP_SYS_RAWIO capabilities makes it "virtually impossible to exploit.

The Linux Kernel versions 4.14, 4.15, and 4.16 has a null pointer dereference which can result in an out of memory (OOM) killing of large mlocked processes. The issue arises from an oom killed process's final thread calling exit_mmap(), which calls munlock_vma_pages_all() for mlocked vmas.This can happen synchronously with the oom reaper's unmap_page_range() since the vma's VM_LOCKED bit is cleared before munlocking (to determine if any other vmas share the memory and are mlocked).

ntfs_attr_find in the ntfs.ko filesystem driver in the Linux kernel 4.15.0 allows attackers to trigger a stack-based out-of-bounds write and cause a denial of service (kernel oops or panic) or possibly have unspecified other impact via a crafted ntfs filesystem.

ntfs_end_buffer_async_read in the ntfs.ko filesystem driver in the Linux kernel 4.15.0 allows attackers to trigger a stack-based out-of-bounds write and cause a denial of service (kernel oops or panic) or possibly have unspecified other impact via a crafted ntfs filesystem.

ntfs_read_locked_inode in the ntfs.ko filesystem driver in the Linux kernel 4.15.0 allows attackers to trigger a use-after-free read and possibly cause a denial of service (kernel oops or panic) via a crafted ntfs filesystem.

In the Linux kernel 4.15.0, a NULL pointer dereference was discovered in hfs_ext_read_extent in hfs.ko. This can occur during a mount of a crafted hfs filesystem.

In arch/x86/kvm/vmx.c in the Linux kernel before 4.17.2, when nested virtualization is used, local attackers could cause L1 KVM guests to VMEXIT, potentially allowing privilege escalations and denial of service attacks due to lack of checking of CPL.

An issue was discovered in the Linux kernel through 4.17.2. The filter parsing in kernel/trace/trace_events_filter.c could be called with no filter, which is an N=0 case when it expected at least one line to have been read, thus making the N-1 index invalid. This allows attackers to cause a denial of service (slab out-of-bounds write) or possibly have unspecified other impact via crafted perf_event_open and mmap system calls.

An issue was discovered in the Linux kernel through 4.17.2. vbg_misc_device_ioctl() in drivers/virt/vboxguest/vboxguest_linux.c reads the same user data twice with copy_from_user. The header part of the user data is double-fetched, and a malicious user thread can tamper with the critical variables (hdr.size_in and hdr.size_out) in the header between the two fetches because of a race condition, leading to severe kernel errors, such as buffer over-accesses. This bug can cause a local denial of service and information leakage.

In the ea_get function in fs/jfs/xattr.c in the Linux kernel through 4.17.1, a memory corruption bug in JFS can be triggered by calling setxattr twice with two different extended attribute names on the same file. This vulnerability can be triggered by an unprivileged user with the ability to create files and execute programs. A kmalloc call is incorrect, leading to slab-out-of-bounds in jfs_xattr.

In net/socket.c in the Linux kernel through 4.17.1, there is a race condition between fchownat and close in cases where they target the same socket file descriptor, related to the sock_close and sockfs_setattr functions. fchownat does not increment the file descriptor reference count, which allows close to set the socket to NULL during fchownat's execution, leading to a NULL pointer dereference and system crash.

In the Linux Kernel before version 4.16.11, 4.14.43, 4.9.102, and 4.4.133, multiple race condition errors when handling probe, disconnect, and rebind operations can be exploited to trigger a use-after-free condition or a NULL pointer dereference by sending multiple USB over IP packets.

In the Linux Kernel before version 4.15.8, 4.14.25, 4.9.87, 4.4.121, 4.1.51, and 3.2.102, an error in the "_sctp_make_chunk()" function (net/sctp/sm_make_chunk.c) when handling SCTP packets length can be exploited to cause a kernel crash.

A flaw was found affecting the Linux kernel before version 4.17. By mmap()ing a FUSE-backed file onto a process's memory containing command line arguments (or environment strings), an attacker can cause utilities from psutils or procps (such as ps, w) or any other program which makes a read() call to the /proc/<pid>/cmdline (or /proc/<pid>/environ) files to block indefinitely (denial of service) or for some controlled time (as a synchronization primitive for other attacks).

An issue was discovered in the Linux kernel through 4.17.2. Since the page allocator does not yield CPU resources to the owner of the oom_lock mutex, a local unprivileged user can trivially lock up the system forever by wasting CPU resources from the page allocator (e.g., via concurrent page fault events) when the global OOM killer is invoked. NOTE: the software maintainer has not accepted certain proposed patches, in part because of a viewpoint that "the underlying problem is non-trivial to handle.