Re: EXT4 Oops (Re: [PATCH V15 06/22] mmc: block: Add blk-mq support)

From: Andreas Dilger <adilger@dilger.ca>
To: "Theodore Ts'o" <tytso@mit.edu>
Cc: Adrian Hunter <adrian.hunter@intel.com>,
	Dmitry Osipenko <digetx@gmail.com>,
	Ulf Hansson <ulf.hansson@linaro.org>,
	linux-mmc <linux-mmc@vger.kernel.org>,
	linux-block <linux-block@vger.kernel.org>,
	linux-kernel <linux-kernel@vger.kernel.org>,
	Bough Chen <haibo.chen@nxp.com>,
	Alex Lemberg <alex.lemberg@sandisk.com>,
	Mateusz Nowak <mateusz.nowak@intel.com>,
	Yuliy Izrailov <Yuliy.Izrailov@sandisk.com>,
	Jaehoon Chung <jh80.chung@samsung.com>,
	Dong Aisheng <dongas86@gmail.com>,
	Das Asutosh <asutoshd@codeaurora.org>,
	Zhangfei Gao <zhangfei.gao@gmail.com>,
	Sahitya Tummala <stummala@codeaurora.org>,
	Harjani Ritesh <riteshh@codeaurora.org>,
	Venu Byravarasu <vbyravarasu@nvidia.com>,
	Linus Walleij <linus.walleij@linaro.org>,
	Shawn Lin <shawn.lin@rock-chips.com>,
	Bartlomiej Zolnierkiewicz <b.zolnierkie@samsung.com>,
	Christoph Hellwig <hch@lst.de>,
	Thierry Reding <treding@nvidia.com>,
	Krishna Reddy <vdumpa@nvidia.com>,
	linux-ext4 <linux-ext4@vger.kernel.org>
Subject: Re: EXT4 Oops (Re: [PATCH V15 06/22] mmc: block: Add blk-mq support)
Date: Thu, 1 Mar 2018 13:20:16 -0700	[thread overview]
Message-ID: <AF3403B9-9791-4A77-A6FB-3281B8EA7A31@dilger.ca> (raw)
In-Reply-To: <20180301160418.GA2490@thunk.org>

[-- Attachment #1: Type: text/plain, Size: 1895 bytes --]

On Mar 1, 2018, at 9:04 AM, Theodore Ts'o <tytso@mit.edu> wrote:
> This doesn't seem to make sense; the PC is where we are currently
> executing, and LR is the "Link Register" where the flow of control
> will be returning after the current function returns, right?  Well,
> dx_probe should *not* be returning to __wait_on_bit().  So this just
> seems.... weird.
> 
> Ignoring the LR register, this stack trace looks sane...  I can't see
> which pointer could be NULL and getting dereferenced, though.  How
> easily can you reproduce the problem?  Can you either (a) translate
> the PC into a line number, or better yet, if you can reproduce, add a
> series of BUG_ON's so we can see what's going on?
> 
> +	BUG_ON(frame);

I think you mean:
	BUG_ON(frame == NULL);
or
	BUG_ON(!frame);

> 	memset(frame_in, 0, EXT4_HTREE_LEVEL * sizeof(frame_in[0]));
> 	frame->bh = ext4_read_dirblock(dir, 0, INDEX);
> 	if (IS_ERR(frame->bh))
> 		return (struct dx_frame *) frame->bh;
> 
> +	BUG_ON(frame->bh);
> +	BUG_ON(frame->bh->b_data);

Same here.

	BUG_ON(frame->bh == NULL);
	BUG_ON(frame->bh->b_data == NULL);

This is why I don't like implicit "is NULL" or "is non-zero" usage.  Lustre
used to require "== NULL" or "!= NULL" to avoid bugs like this, but had to
abandon that because of upstream code style.

> 	root = (struct dx_root *) frame->bh->b_data;
> 	if (root->info.hash_version != DX_HASH_TEA &&
> 	    root->info.hash_version != DX_HASH_HALF_MD4 &&
> 	    root->info.hash_version != DX_HASH_LEGACY) {
> 
> These are "could never" happen scenarios from looking at the code, but
> that will help explain what is going on.
> 
> If this is reliably only happening with mq, the only way I could see
> that if is something is returning an error when it previously wasn't.
> This isn't a problem we're seeing with any of our testing, though.
> 
> Cheers,
> 
> 						- Ted
> 

Cheers, Andreas

[-- Attachment #2: Message signed with OpenPGP --]
[-- Type: application/pgp-signature, Size: 873 bytes --]