[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20230622144042.GA21415@lst.de>
Date: Thu, 22 Jun 2023 16:40:42 +0200
From: Christoph Hellwig <hch@....de>
To: Guenter Roeck <linux@...ck-us.net>
Cc: Christoph Hellwig <hch@....de>, Jens Axboe <axboe@...nel.dk>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
"Rafael J. Wysocki" <rafael@...nel.org>,
Mike Snitzer <snitzer@...nel.org>,
Joern Engel <joern@...ybastard.org>,
Miquel Raynal <miquel.raynal@...tlin.com>,
Richard Weinberger <richard@....at>,
Vignesh Raghavendra <vigneshr@...com>,
Pavel Machek <pavel@....cz>, dm-devel@...hat.com,
linux-kernel@...r.kernel.org, linux-block@...r.kernel.org,
linux-mtd@...ts.infradead.org, linux-pm@...r.kernel.org
Subject: Re: [PATCH 14/24] init: clear root_wait on all invalid root=
strings
On Thu, Jun 22, 2023 at 06:54:41AM -0700, Guenter Roeck wrote:
> On 6/21/23 23:00, Christoph Hellwig wrote:
>> Hi Guenter,
>>
>> can you try this patch?
>>
>> diff --git a/block/early-lookup.c b/block/early-lookup.c
>> index a5be3c68ed079c..66e4514d671179 100644
>> --- a/block/early-lookup.c
>> +++ b/block/early-lookup.c
>> @@ -174,7 +174,7 @@ static int __init devt_from_devname(const char *name, dev_t *devt)
>> while (p > s && isdigit(p[-1]))
>> p--;
>> if (p == s || !*p || *p == '0')
>> - return -EINVAL;
>> + return -ENODEV;
>> /* try disk name without <part number> */
>> part = simple_strtoul(p, NULL, 10);
>
> Not completely. Tests with root=/dev/sda still fail.
>
> "name" passed to devt_from_devname() is "sda".
>
> for (p = s; *p; p++) {
> if (*p == '/')
> *p = '!';
> }
>
> advances 'p' to the end of the string.
>
> while (p > s && isdigit(p[-1]))
> p--;
>
> moves it back to point to the first digit (if there is one).
>
> if (p == s || !*p || *p == '0')
> return -EINVAL;
>
> then fails because *p is 0. In other words, the function only accepts
> drive names with digits at the end (and the first digit must not be '0').
>
> I don't recall how I hit the other condition earlier. I have various
> "/dev/mmcblkX" in my tests, where X can be any number including 0.
> Maybe those fail randomly as well.
>
> Overall I am not sure though what an "invalid" devicename is supposed
> to be in this context. I have "sda", "sr0", "vda", "mtdblkX",
> "nvme0n1", "mmcblkX", and "hda". Why would any of those not be eligible
> for "rootwait" ?
>
> In practice, everything not ending with a digit, or ending with
> '0', fails the first test. Everything ending with a digit > 0
> fails the second test. But "humptydump3p4" passes all those tests.
Yeah. I guess I should give up on the idea of error out in this
particular parser. The idea sounded good, but I guess it doesn't
work. So we'll probably want his fix:
diff --git a/block/early-lookup.c b/block/early-lookup.c
index a5be3c68ed079c..9e2d5a19de1b3b 100644
--- a/block/early-lookup.c
+++ b/block/early-lookup.c
@@ -174,7 +174,7 @@ static int __init devt_from_devname(const char *name, dev_t *devt)
while (p > s && isdigit(p[-1]))
p--;
if (p == s || !*p || *p == '0')
- return -EINVAL;
+ return -ENODEV;
/* try disk name without <part number> */
part = simple_strtoul(p, NULL, 10);
@@ -185,7 +185,7 @@ static int __init devt_from_devname(const char *name, dev_t *devt)
/* try disk name without p<part number> */
if (p < s + 2 || !isdigit(p[-2]) || p[-1] != 'p')
- return -EINVAL;
+ return -ENODEV;
p[-1] = '\0';
*devt = blk_lookup_devt(s, part);
if (*devt)
Powered by blists - more mailing lists