linux-kernel - Re: [PATCH 08/17] Add main.c

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 for Android: free password hash cracker in your pocket

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

Message-ID: <2e2ec1ba-0c24-4173-af60-ea51004f2e10@ti.com>
Date: Fri, 31 May 2024 16:50:47 +0300
From: "Nemanov, Michael" <michael.nemanov@...com>
To: Krzysztof Kozlowski <krzk@...nel.org>, Kalle Valo <kvalo@...nel.org>,
        Johannes Berg <johannes.berg@...el.com>,
        Breno Leitao <leitao@...ian.org>,
        Justin Stitt <justinstitt@...gle.com>,
        Kees Cook <keescook@...omium.org>, <linux-wireless@...r.kernel.org>,
        <linux-kernel@...r.kernel.org>
CC: Sabeeh Khan <sabeeh-khan@...com>
Subject: Re: [PATCH 08/17] Add main.c

On 5/31/2024 10:35 AM, Krzysztof Kozlowski wrote:
 > It's impossible to read your email.
Sorry for messing the formatting, sending again:


On 5/22/2024 12:46 PM, Krzysztof Kozlowski wrote:>
 >> +}
 >> +
 >> +static int cc33xx_remove(struct platform_device *pdev)
 >
 > Why remove callback is before probe? Please follow standard driver
 > convention. This goes immediately after probe.
Will fix


 >> +{
 >> +	struct cc33xx_platdev_data *pdev_data = dev_get_platdata(&pdev->dev);
 >> +	struct cc33xx *cc = platform_get_drvdata(pdev);
 >> +
 >> +	set_bit(CC33XX_FLAG_DRIVER_REMOVED, &cc->flags);
 >
 > ?!?!
 >
 > Your code is seriously buggy if you depend on setting bit in remove
 > callback.
If removal of the CC33xx driver was caused by the removal of its parent 
SDIO device then all I/O operations will fail as SDIO transport is no 
longer available. This will eventually trigger the recovery mechanism 
which in this case is futile. If this flag is set, no recovery is 
attempted.


 >> +		return -EINVAL;
 >> +	}
 >> +
 >> +	hw = cc33xx_alloc_hw(CC33XX_AGGR_BUFFER_SIZE);
 >> +	if (IS_ERR(hw)) {
 >> +		cc33xx_error("can't allocate hw");
 >
 > Heh? Since when do we print memory allocation failures? Since when
 > memory allocation returns ERR ptr?

Will fix.


 >> +};
 >> +MODULE_DEVICE_TABLE(platform, cc33xx_id_table);
 >> +
 >> +static struct platform_driver cc33xx_driver = {
 >> +	.probe		= cc33xx_probe,
 >> +	.remove		= cc33xx_remove,
 >> +	.id_table	= cc33xx_id_table,
 >> +	.driver = {
 >> +		.name	= "cc33xx_driver",
 >> +	}
 >> +};
 >> +
 >> +u32 cc33xx_debug_level = DEBUG_NO_DATAPATH;
 >
 > Why this is global? Why u32? Why global variable is defined at the end
 > of the file?!?!

cc33xx_debug_level together with cc33xx_debug/info/error() macros is how 
all traces were done in drivers/net/wireless/ti/wlcore/ (originally was 
wl1271_debug/info etc.)
It enables / disables traces without rebuilding or even reloading which 
is very helpful for remote support. These macros map to dynamic_pr_debug 
/ pr_debug. I saw similar wrappers in other wireless drivers (ath12k). 
This is also why there are plenty of cc33xx_debug() all over the code, 
most are silent by default.

 >> +
 >> +module_platform_driver(cc33xx_driver);
 >> +
 >> +module_param_named(debug_level, cc33xx_debug_level, uint, 0600);
 >> +MODULE_PARM_DESC(debug_level, "cc33xx debugging level");
 >> +
 >> +MODULE_PARM_DESC(secure_boot_enable, "Enables secure boot and FW 
downlaod");
 >
 > Eh? why secure boot is module param?
 >
 >> +
 >> +module_param_named(fwlog, fwlog_param, charp, 0);
 >> +MODULE_PARM_DESC(fwlog, "FW logger options: continuous, dbgpins or 
disable");
 >> +
 >> +module_param(no_recovery, int, 0600);
 >> +MODULE_PARM_DESC(no_recovery, "Prevent HW recovery. FW will remain 
stuck.");
 >> +
 >> +module_param_named(ht_mode, ht_mode_param, charp, 0400);
 >> +MODULE_PARM_DESC(ht_mode, "Force HT mode: wide or siso20");
 >
 > Does not look like suitable for module params.
Was useful during development, can be removed.


On 5/31/2024 10:35 AM, Krzysztof Kozlowski wrote:
 > Was never allowed. There is only one version: the kernel version. There
 > were many comments already explaining why "driver version" is
 > wrong/meaningless.

OK. HW / FW versions are OK?


Michael.