The 7.0.4 kernel finally stops AMD RDNA4 graphics cards from tripping over empty memory tables during boot, so you can actually get past a black screen and into your desktop without staring at a kernel oops. Memory management got patched to keep the slab allocator from corrupting itself when non-maskable interrupts sneak in on single-processor systems, while the networking stack finally rejects malformed packet rules that used to trigger undefined behavior and silent crashes. Nested virtualization logic now properly syncs guest state before resuming execution, and filesystem drivers like NTFS3 and ext2 got stricter bounds checking to stop crafted disk images from reading past allocated buffers or triggering panic conditions. It is a solid stability bump that quietly patches the race conditions and buffer overflows most people never notice until their server decides to reboot itself, so just run your updates and get back to actually using your machine.

Linux Kernel 7.0.4 Released: Fixes RDNA4 Crashes, Memory Leaks, and Network Race Conditions

The Linux Kernel 7.0.4 stable release arrives with exactly what everyone needs right now: patches that stop AMD graphics cards from crashing on boot, fix memory deadlocks that trip up single-processor systems, and patch networking race conditions that break virtualization setups. Desktop users will appreciate the immediate stability improvements for RDNA4 hardware, while server admins can finally sleep better knowing that network stack overflows are no longer sitting idle in the codebase waiting to trigger a kernel panic. The update also tightens up filesystem validation and virtualization state tracking to prevent silent data corruption from creeping into production environments.

Linux Kernel 7.0.4 Fixes for AMD RDNA4 and GPU Stability

The most noticeable change for desktop enthusiasts involves the AMDGPU driver, which previously crashed the system when initializing GDS resources on RDNA4 hardware like the RX 9070 XT. The kernel attempted to register memory manager resources even when the size was zero, triggering a fatal assertion error during module loading. This update adds a simple guard that skips registration for absent hardware resources, allowing the graphics stack to initialize correctly without tripping over empty allocation tables. Users with the latest AMD silicon can now boot into their desktop environments instead of staring at a black screen or a kernel oops trace every time they load the driver.

Memory Management and Slab Allocator Hardening

Memory management subsystems received several patches that prevent deadlocks and buffer overflows during high-load scenarios. The slab allocator previously allowed re-entry from Non-Maskable Interrupt contexts on single-processor kernels, which corrupted internal lock states and triggered BUG reports when debugging spinlocks were enabled. The update forces an early return in these specific NMI paths to keep the memory allocator stable. Additionally, a buffer overflow in the vmalloc subsystem was corrected where shrinking allocations could write past the new buffer boundaries. These changes stop silent data corruption that often manifests as random application crashes or filesystem errors weeks after installation.

Network Stack Race Conditions and Security Patches

The networking layer saw significant work to prevent undefined behavior and out-of-bounds memory access in packet processing paths. Netfilter now rejects zero shift operands in bitwise expressions, stopping malformed rules from triggering undefined behavior that could crash the kernel when packets hit the filter path. IPv6 routing logic for segment routing and RPL tunnels received fixes for race conditions involving destination cache entries on PREEMPT_RT kernels. These changes prevent use-after-free errors that occur when concurrent tasks release shared nexthop objects while caching routines are still running. The updates ensure that network stacks remain stable even under heavy virtualization or container workloads where packet processing happens at line speed.

Virtualization and Filesystem Reliability

Virtualization enthusiasts will appreciate the extensive consistency checks added to nested SVM support, which now properly handles VMRUN failures and ensures guest state is correctly synchronized before execution resumes. Several filesystem drivers also received patches to prevent out-of-bounds reads and integer overflows when processing crafted images. The NTFS3 driver now validates buffer boundaries before accessing journal metadata, while the ext2 driver rejects inodes with zero link counts that indicate corruption. These safeguards stop malicious or malformed disk images from triggering kernel panics or leaking sensitive memory information during standard mount operations.

Linux kernel 7.0.4 released

Linux kernel version 7.0.4 is now available:

Full source: https://cdn.kernel.org/pub/linux/kernel/v7.x/linux-7.0.4.tar.xz
Patch: https://cdn.kernel.org/pub/linux/kernel/v7.x/patch-7.0.4.xz
PGP Signature: https://cdn.kernel.org/pub/linux/kernel/v7.x/linux-7.0.4.tar.sign

You can view the summary of the changes at the following URL:
https://git.kernel.org/stable/ds/v7.0.4/v7.0.3

The update is available through your distribution's package manager or by compiling the source directly. Applying this release keeps systems running smoothly and prevents the kind of silent data corruption that usually shows up weeks later as mysterious application failures. Keep an eye on your system logs after updating to ensure all hardware drivers initialize correctly, especially if you are using newer AMD graphics cards or heavy virtualization setups. Happy patching, and may your uptime stay high while your panic traces stay low.