On Tue, Feb 25, 2020 at 10:42:40AM -0800, Christoph Hellwig wrote: > On Tue, Feb 25, 2020 at 10:40:23AM -0800, Darrick J. Wong wrote: > > Prior to this patch, the "!(flags & DEBUGGER)" expressions in the call > > sites evaluate to 0 or 1, and this effectively results in libxfs_mount > > passing EXIT_ON_FAILURE to the buffer read functions as the flag value. > > The flag value is passed all the way down to __read_buf, and when it > > sees an IO failure, it exits. > > > > After this patch, libxfs_mount passes flags==0, which means that we get > > a buffer back, possibly with b_error set. If b_error is set, we log a > > warning about the screwed up filesystem and return a null mount if the > > libxfs_mount caller didn't indicate that it is a debugger. Presumably > > the libxfs_mount caller will exit with error if we return a null mount. > > > > IOWs, I'm doing exactly what the commit message says, but in a rather > > subtle way. I'll clarify that, if you'd like. > > Ok, with a proper commit message this looks good to me: > > Reviewed-by: Christoph Hellwig <hch@xxxxxx> The new commit message reads as follows: "libxfs: open-code "exit on buffer read failure" in upper level callers "Make all functions that use LIBXFS_EXIT_ON_FAILURE to abort on buffer read errors implement that logic themselves. This also removes places where libxfs can abort the program with no warning. "Note that in libxfs_mount, the "!(flags & DEBUGGER)" code would indirectly select LIBXFS_EXIT_ON_FAILURE, so we're replacing the hidden library exit(1) with a null xfs_mount return, which should cause the utilities to exit with an error." --D
