Query: Y2038 patch series for 3.18 Kernel
Hi, We have an arm32 based embedded product which is based on the 3.18 kernel and a simple busybox. We wanted to support the Y2038 issue on this older kernel. Is this feasible Do we have the Y2038 separate patches available for both kernel and userspace ? Or upgrading the kernel is the only option ? Thanks, Pintu
Mon, 03 Nov 2025 09:57:18 +0530 को Pintu Kumar Agarwal ने लिखा :
We have an arm32 based embedded product which is based on the 3.18 kernel and a simple busybox. We wanted to support the Y2038 issue on this older kernel. Is this feasible Do we have the Y2038 separate patches available for both kernel and userspace ? Or upgrading the kernel is the only option ?
Upgrading is a much much better option. 3.18 is extremely ancient and extremely insecure. (Sending again as I didn't press "reply all") Thanks, Siddh
On Mon, 3 Nov 2025 at 10:35, Siddh Raman Pant <sanganaka@siddh.me> wrote:
Mon, 03 Nov 2025 09:57:18 +0530 को Pintu Kumar Agarwal ने लिखा :
We have an arm32 based embedded product which is based on the 3.18 kernel and a simple busybox. We wanted to support the Y2038 issue on this older kernel. Is this feasible Do we have the Y2038 separate patches available for both kernel and userspace ? Or upgrading the kernel is the only option ?
Upgrading is a much much better option.
Yes, I understand, but upgrading the kernel on this older SoC brings more complexities, challenges and time taking. Customers also do not agree for the upgrade at this stage and they are looking for alternatives. So, we are exploring both the options right now. Are there any patchset maintained separately for this Y2038 issue for each kernel version to have a look ? Thanks, Pintu
On Tue, Nov 04, 2025 at 10:32:28AM +0530, Pintu Kumar Agarwal wrote:
On Mon, 3 Nov 2025 at 10:35, Siddh Raman Pant <sanganaka@siddh.me> wrote:
Mon, 03 Nov 2025 09:57:18 +0530 को Pintu Kumar Agarwal ने लिखा :
We have an arm32 based embedded product which is based on the 3.18 kernel and a simple busybox. We wanted to support the Y2038 issue on this older kernel. Is this feasible Do we have the Y2038 separate patches available for both kernel and userspace ? Or upgrading the kernel is the only option ?
Upgrading is a much much better option.
Yes, I understand, but upgrading the kernel on this older SoC brings more complexities, challenges and time taking. Customers also do not agree for the upgrade at this stage and they are looking for alternatives. So, we are exploring both the options right now.
Are there any patchset maintained separately for this Y2038 issue for each kernel version to have a look ?
Nope, sorry. Please just update to a newer kernel version, that is the only way forward (and as a bonus, you actually will have tens of thousands of bugs fixed on your system.) good luck! greg k-h
On Tue, Nov 4, 2025, at 06:02, Pintu Kumar Agarwal wrote:
On Mon, 3 Nov 2025 at 10:35, Siddh Raman Pant <sanganaka@siddh.me> wrote:
Mon, 03 Nov 2025 09:57:18 +0530 को Pintu Kumar Agarwal ने लिखा :
We have an arm32 based embedded product which is based on the 3.18 kernel and a simple busybox. We wanted to support the Y2038 issue on this older kernel. Is this feasible Do we have the Y2038 separate patches available for both kernel and userspace ? Or upgrading the kernel is the only option ?
Upgrading is a much much better option.
Yes, I understand, but upgrading the kernel on this older SoC brings more complexities, challenges and time taking. Customers also do not agree for the upgrade at this stage and they are looking for alternatives. So, we are exploring both the options right now.
Are there any patchset maintained separately for this Y2038 issue for each kernel version to have a look ?
I used to maintain a 5.4 kernel with backports from the 5.6 patches, but I would not recommend using that. Anything earlier than 5.4 is completely hopeless for 2038 because one would have to essentially redo the changes from scratch. Which chip exactly are you using? Arnd
On Tue, 4 Nov 2025 at 15:18, Arnd Bergmann <arnd@arndb.de> wrote:
On Tue, Nov 4, 2025, at 06:02, Pintu Kumar Agarwal wrote:
On Mon, 3 Nov 2025 at 10:35, Siddh Raman Pant <sanganaka@siddh.me> wrote:
Mon, 03 Nov 2025 09:57:18 +0530 को Pintu Kumar Agarwal ने लिखा :
We have an arm32 based embedded product which is based on the 3.18 kernel and a simple busybox. We wanted to support the Y2038 issue on this older kernel. Is this feasible Do we have the Y2038 separate patches available for both kernel and userspace ? Or upgrading the kernel is the only option ?
Upgrading is a much much better option.
Yes, I understand, but upgrading the kernel on this older SoC brings more complexities, challenges and time taking. Customers also do not agree for the upgrade at this stage and they are looking for alternatives. So, we are exploring both the options right now.
Are there any patchset maintained separately for this Y2038 issue for each kernel version to have a look ?
I used to maintain a 5.4 kernel with backports from the 5.6 patches, but I would not recommend using that. Anything earlier than 5.4 is completely hopeless for 2038 because one would have to essentially redo the changes from scratch.
OK thanks.
Which chip exactly are you using? This request is for one of the legacy QC MDM 9K chipset.
On Wed, Nov 5, 2025, at 10:47, Pintu Kumar Agarwal wrote:
On Tue, 4 Nov 2025 at 15:18, Arnd Bergmann <arnd@arndb.de> wrote:
Which chip exactly are you using? This request is for one of the legacy QC MDM 9K chipset.
Ah, I see. Support for those was never fully merged, so it's much harder to upgrade than other platforms, especially if you need the modem driver to work. We do have some support for the newer sdx55/sdxx65 chips, and much more rudimentary support for mdm9607 and mdm9615 (without modem) in mainline, but since we've never had a fully supported board use the older chips I would not expect them to work without additional bugfixes and a rebase of the missing drivers. There is of course a chance that you may get them to work in a way that does not rely on knowing the time. If you don't have (or use) and RTC and the system time is always set to Jan 1 1970 on boot, this would technically work indefinitely. Running a 3.18 kernel is obviously a bad idea for a number of other reasons. Arnd
On Thu, 6 Nov 2025 at 13:48, Arnd Bergmann <arnd@arndb.de> wrote:
On Wed, Nov 5, 2025, at 10:47, Pintu Kumar Agarwal wrote:
On Tue, 4 Nov 2025 at 15:18, Arnd Bergmann <arnd@arndb.de> wrote:
Which chip exactly are you using? This request is for one of the legacy QC MDM 9K chipset.
Ah, I see. Support for those was never fully merged, so it's much harder to upgrade than other platforms, especially if you need the modem driver to work.
Sorry for my late reply. Yes, I see some supports are present but not fully functional.
We do have some support for the newer sdx55/sdxx65 chips, and much more rudimentary support for mdm9607 and mdm9615 (without modem) in mainline, but since we've never had a fully supported board use the older chips I would not expect them to work without additional bugfixes and a rebase of the missing drivers.
Yes, last year and this year as well we already did kernel upgrades for some of them but few are still remaining.
There is of course a chance that you may get them to work in a way that does not rely on knowing the time. If you don't have (or use) and RTC and the system time is always set to Jan 1 1970 on boot, this would technically work indefinitely. Running a 3.18 kernel is obviously a bad idea for a number of other reasons.
Yes, we would also like to propose for the kernel upgrade as the last option if there are no valid alternatives. Thank you so much Arnd for your help and guidance. Thank you all for the discussion! Regards, Pintu
participants (4)
-
Arnd Bergmann -
Greg KH -
Pintu Kumar Agarwal -
Siddh Raman Pant