I wrote:
> The ARM runtime reports the major device type associated with
> /proc/self/mountinfo as MMC_BLOCK_MAJOR, causing on_ssd() in
> readahead-common.c to return false. on_ssd() should return true, as
> MMC like SSD is not rotational.
Lennart writes:
> Not following here. fs_on_ssd() will actually check for the high-level
> ID_SSD property, as well as the "queue/rotational" sysfs attribute for
> the block device.
Those other checks are not reached, as the beginning of fs_on_ssd() has
if (major(st.st_dev) == 0) {}
encapsulating them, and major(st.st_dev)=MMC_BLOCK_MAJOR, not 0. My first
attempt to fix the problem was
+ if (streq(opts, "mmc") || startswith(opts, "mmc,") || endswith(opts,
",mmc") || strstr(opts, ",mmc,"))
but that statement is also not reached. Likely the best practice is also to
compare MTD_BLOCK_MAJOR, but since I have no simple way to test that case, I
didn't include it. That's why I put the change inside an #ifdef. As Auke
points out in systemd-commits, where I erroneously attempted to post this
patch, any arch using an MMC would consider it non-rotational.
-- Alison Chaiken
Mentor Graphics, from rainy Hildesheim
_______________________________________________
systemd-devel mailing list
systemd-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/systemd-devel
