CVE-2026-93264

*** Pendiente de traducción *** In the Linux kernel, the following vulnerability has been resolved:<br /> <br /> RDMA/efa: Fix PBL chunk length computation<br /> <br /> On register MR, when creating the PBL, if it&amp;#39;s an indirect PBL we create<br /> a chunk list to hold the PBL pages pointers. Each chunk is 4KB in size<br /> and can hold 510 addresses (EFA_PTRS_PER_CHUNK) and has a 12-byte<br /> control buffer at the end of it holding the next chunk&amp;#39;s pointer and its<br /> length.<br /> <br /> If the PBL number of pages is a multiple of EFA_PTRS_PER_CHUNK, the<br /> calculated last chunk length is wrongly computed as 0, even though that<br /> chunk is fully populated with 510 real page pointers. This wrong length<br /> is used both to DMA map the chunk and is propagated to the device,<br /> causing the device to see the chunk as empty and reject the memory<br /> registration.<br /> <br /> Fix the calculation so it will be performed only if the number of pages<br /> isn&amp;#39;t a multiple of EFA_PTRS_PER_CHUNK, if it is, its already handled in<br /> the above loop correctly.<br /> Also prevent out-of-bounds reach in the chunks array in such scenario.