On Thu, 9 Jan 2025 at 16:50, James Bottomley <James.Bottomley@xxxxxxxxxxxxxxxxxxxxx> wrote: > > On Thu, 2025-01-09 at 10:50 +0100, Ard Biesheuvel wrote: > > On Tue, 7 Jan 2025 at 03:36, James Bottomley > > <James.Bottomley@xxxxxxxxxxxxxxxxxxxxx> wrote: > [...] > > > James Bottomley (6): > > > efivarfs: remove unused efi_varaible.Attributes and .kobj > > > efivarfs: add helper to convert from UC16 name and GUID to utf8 > > > name > > > efivarfs: make variable_is_present use dcache lookup > > > efivarfs: move freeing of variable entry into evict_inode > > > efivarfs: remove unused efivarfs_list > > > efivarfs: fix error on write to new variable leaving remnants > > > > > > > Thanks James, > > > > I've tentatively queued up this series, as well as the hibernate one, > > to get some coverage from the robots while I run some tests myself. > > > > Are there any existing test suites that cover efivarfs that you could > > recommend? > > I'm afraid I couldn't find any. I finally wrote a few shell scripts to > try out multiple threads updating the same variable. I think I can > probably work out how to add these to the kselftest infrastructure. > Hibernation was a real pain because it doesn't work with secure boot, > but I finally wrote a UEFI shell script to modify variables and reset. > Unfortunately I don't think we have a testing framework I can add these > to. > I tested the hibernation changes using QEMU/arm64, using a non-persistent varstore image, and checked whether adding and/or deleting variables via efivarfs resulted in the expected behavior after a resume from hibernate. I also checked the behavior regarding zero sized files, and that looks sound to me. But I didn't go as far as torture test it with concurrent updates as you have. All in all, I think it is reasonable to queue this up so it gets some wider exposure.
