[ale] SSD stuck in R/O.

[Ed Cashin]

It would be interesting to know what the drive has to say for itself.

Vendors often provide highly annoying but ultimately useful software for
diagnostics.

Also, you can sometimes use SMART to find out what the disk thinks is going
on.

Regarding the, "Where does the data go?" question, the answer depends on
whether the page cache is being used.  Usually it is.  First written data
goes to RAM, and the kernel notes that the page of memory is dirty.

Dirty pages are later written out to the underlying block device.  You can
configure the virtual memory settings to force the writing process to wait
for that sync to occur, in order to throttle the process, or you can allow
the dirty pages to accumulate, and all the processes on the system pay when
they have to wait for the dirty pages to get written out.  (That is why you
sometimes get people complaining that Linux is behaving in a silly way when
stuff seems to become unresponsive.)

If the block device then says that there's been an error, the filesystem
handles the error in some pre-configured way, e.g., by making the
filesystem read only.  Subsequent I/O fails.  But if the device doesn't say
there's an error, the kernel doesn't know any better, and it simply marks
the pages as clean.  Then you will never know that your data is being
written out by the kernel but silently ignored by the misbehaving disk.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.ale.org/pipermail/ale/attachments/20191011/a84ad025/attachment.html>