{"id":"CVE-2026-98826","published":"2026-10-18T16:17:36.823","lastModified":"2026-10-19T10:19:46.643","description":"In the Linux kernel, the following vulnerability has been resolved:\n\nrtc: ac100: Fix clock provider use-after-free on probe failure\n\nac100_rtc_register_clks() registers the RTC-32k clock with\nclk_hw_register_fixed_rate() and the clock provider with\nof_clk_add_hw_provider(). Neither is device managed, and both are\nreleased only from ac100_rtc_remove(). The driver core does not call\nremove() when probe() fails: really_probe() reaches probe_failed below\nthe device_remove() call and goes straight to releasing the device's\nmanaged resources.\n\nac100_rtc_probe() ends with\n\n\treturn devm_rtc_register_device(chip->rtc);\n\nso if that fails after the clocks have been registered, the provider is\nleft on the global of_clk_providers list holding chip->clk_data, which\nwas allocated with devm_kzalloc() and is freed while probe() unwinds. A\nlater lookup on this device tree node then reads freed memory in\nof_clk_hw_onecell_get(). Consumers that do exactly that exist in tree:\nthe wifi power sequence nodes on sun8i-a83t-bananapi-m3 and\nsun8i-a83t-cubietruck-plus take <&ac100_rtc 1>, and the sun9i-a80 boards\nroute osc32k through <&ac100_rtc 0>. The window is narrow, as\ndevm_rtc_register_device() can only fail with -ENOMEM here, but the\nprovider is left dangling whenever it does.\n\nThe RTC-32k clock is leaked on the same path. Since clk_core_lookup()\nsearches a global list that is not scoped per device, __clk_register()\nrejects the duplicate name with -EEXIST, so the leak also makes a later\nprobe of the same device fail.\n\nThe error path that returns -EINVAL when the ADDA 4M parent clock cannot\nbe found leaks the RTC-32k clock in the same way. No provider has been\nregistered at that point, so that one is a leak rather than a\nuse-after-free.\n\nRegister both with devm_clk_hw_register_fixed_rate() and\ndevm_of_clk_add_hw_provider() so that they are released whenever the\ndevice goes away, on a failed probe as well as on unbind. Devres\nreleases in reverse order of acquisition, so the provider is removed\nbefore chip->clk_data is freed, and the clkout children are now\nunregistered before their parent rather than after it.\nac100_rtc_unregister_clks() and the remove callback then have nothing\nleft to do and are removed.","cvssScore":null,"cvssSeverity":null,"cvssVector":null,"cwes":[],"vendors":[],"products":[],"references":[{"url":"https://git.kernel.org/stable/c/12bcd20d76703ee2baf52224d287b7fea096b12f","tags":[]},{"url":"https://git.kernel.org/stable/c/55ca60ba336632349f7c4585d22ba74ae355b7ce","tags":[]},{"url":"https://git.kernel.org/stable/c/9cff19ddd7f90c1f572a0f83cfc41de8f8dcbcd8","tags":[]},{"url":"https://git.kernel.org/stable/c/fcf0b58e976e83adf246b014518e49c662b96f5e","tags":[]}],"exploitRefs":[],"hasPoc":false,"ai":null}