On Thu, Jun 28, 2018 at 12:28 PM, <jane....@oracle.com> wrote: > Hi, Dan, > > I'm concerned about this check > > + if (!ndctl_test_attempt(test, KERNEL_VERSION(4, 19, 0))) > > I read it as: this ndctl test only works for upstream kernel 4.19 and above. > > For Linux distributors, such as us, we pick a major upstream release as the > base OS, then painstakingly backport desired upstream patches etc. > Our latest OS version string will be 4.14.x.y.z for a long while, > that means the ndctl test won't work even if we backport the 4.19 nvdimm > patches over.
You can run: make KVER=4.19.0 check ...and it will override the kernel version detection. > Is it possible to come up with a more accommodating method? We're trying to scale back our usage of kernel version gates, but if it crashes the system we want to protect unsuspecting users. > One way might be is to 'dd' a block of 0s to the injected block > immediately after a successful injection? For now, doing so leaves the > 'badblocks' and the poison bit in place. I'll notte that this is one of the tests marked in the "ENABLE_DESTRUCTIVE" section of test/Makefile.am in the ndctl source. Those tests may destroy data or otherwise adversely affect the system. They are only meant for developer workstations and you must pass "--enable-destructive" to 'configure' when building ndctl. The intent is "don't enabled these tests unless you know what you are doing". _______________________________________________ Linux-nvdimm mailing list Linux-nvdimm@lists.01.org https://lists.01.org/mailman/listinfo/linux-nvdimm
