linux-kernel - Re: [BUG] mtd: rawnand: denali_pci: page fault when probing fails

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20220411093715.4cecc280@xps13>
Date:   Mon, 11 Apr 2022 09:37:15 +0200
From:   Miquel Raynal <miquel.raynal@...tlin.com>
To:     Zheyu Ma <zheyuma97@...il.com>
Cc:     richard@....at, vigneshr@...com, linux-mtd@...ts.infradead.org,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [BUG] mtd: rawnand: denali_pci: page fault when probing fails

Hi Zheyu,

zheyuma97@...il.com wrote on Sun, 10 Apr 2022 22:17:35 +0800:

> Hello,
> 
> I found a bug in the denali_pci module.
> When the driver fails to probe, we will get the following splat:
> 
> [    4.472703] denali-nand-pci 0000:00:05.0: timeout while waiting for
> irq 0x1000
> [    4.474071] denali-nand-pci: probe of 0000:00:05.0 failed with error -5
> [    4.473538] nand: No NAND device found
> [    4.474068] BUG: unable to handle page fault for address:
> ffffc90005000410
> [    4.475169] #PF: supervisor write access in kernel mode
> [    4.475579] #PF: error_code(0x0002) - not-present page
> [    4.478362] RIP: 0010:iowrite32+0x9/0x50
> [    4.486068] Call Trace:
> [    4.486269]  <IRQ>
> [    4.486443]  denali_isr+0x15b/0x300 [denali]
> [    4.486788]  ? denali_direct_write+0x50/0x50 [denali]
> [    4.487189]  __handle_irq_event_percpu+0x161/0x3b0
> [    4.487571]  handle_irq_event+0x7d/0x1b0
> [    4.487884]  handle_fasteoi_irq+0x2b0/0x770
> [    4.488219]  __common_interrupt+0xc8/0x1b0
> [    4.488549]  common_interrupt+0x9a/0xc0
> 
> It seems that the driver unmap the memory region before disabling the irq.

Thanks for the report! The mapping is done with devm_ helpers and so is
the IRQ registration, so it's slightly more complicated that just
moving a function call in the remove path, apparently. Would you mind
investigating and proposing a patch?

Thanks,
Miquèl