linux-kernel - Re: [PATCH v2 3/4] crypto: ccp: Allow PSP driver to load without SEV/TEE support

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: <d1586b34-1ed3-99fe-ce03-90aa54e88eae@amd.com>
Date:   Thu, 31 Mar 2022 15:10:29 -0500
From:   Tom Lendacky <thomas.lendacky@....com>
To:     Mario Limonciello <mario.limonciello@....com>,
        John Allen <john.allen@....com>,
        Herbert Xu <herbert@...dor.apana.org.au>,
        "David S . Miller" <davem@...emloft.net>
Cc:     open list <linux-kernel@...r.kernel.org>,
        "open list:AMD CRYPTOGRAPHIC COPROCESSOR (CCP) DRIVER" 
        <linux-crypto@...r.kernel.org>,
        Kerneis Gabriel <Gabriel.Kerneis@....gouv.fr>,
        Richard Hughes <hughsient@...il.com>
Subject: Re: [PATCH v2 3/4] crypto: ccp: Allow PSP driver to load without
 SEV/TEE support

On 3/29/22 11:41, Mario Limonciello wrote:
> Previously the PSP probe routine would fail if both SEV and TEE were
> missing.  This is possibly the case for some client parts.
> 
> As capabilities can now be accessed from userspace, it may still be
> useful to have the PSP driver finish loading so that those capabilities
> can be read.
> 
> Signed-off-by: Mario Limonciello <mario.limonciello@....com>
> ---
> v1->v2:
>   * Whitespace fixes
> ---
>   drivers/crypto/ccp/psp-dev.c | 2 +-
>   drivers/crypto/ccp/sp-dev.c  | 6 ++++++
>   2 files changed, 7 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/crypto/ccp/psp-dev.c b/drivers/crypto/ccp/psp-dev.c
> index 8cd404121cd5..3f47b2d81e3c 100644
> --- a/drivers/crypto/ccp/psp-dev.c
> +++ b/drivers/crypto/ccp/psp-dev.c
> @@ -158,7 +158,7 @@ int psp_dev_init(struct sp_device *sp)
>   
>   	ret = psp_check_support(psp);
>   	if (ret)
> -		goto e_disable;
> +		return 0;
>   
>   	/* Disable and clear interrupts until ready */
>   	iowrite32(0, psp->io_regs + psp->vdata->inten_reg);
> diff --git a/drivers/crypto/ccp/sp-dev.c b/drivers/crypto/ccp/sp-dev.c
> index 7eb3e4668286..3486ab2a8982 100644
> --- a/drivers/crypto/ccp/sp-dev.c
> +++ b/drivers/crypto/ccp/sp-dev.c
> @@ -132,6 +132,9 @@ int sp_request_psp_irq(struct sp_device *sp, irq_handler_t handler,
>   
>   void sp_free_ccp_irq(struct sp_device *sp, void *data)
>   {
> +	if (!sp->irq_registered)
> +		return;
> +

I just noticed that irq_registered is only set when a single interrupt 
ends up being used. When both the CCP and PSP get their own interrupts 
(the common case), this will result in free_irq() not being called.

So this needs to be fixed. I would think just allowing the IRQs to be 
requested and freed whenever the driver is loaded (regardless of whether 
they would be used) might be the best way forward.

Thanks,
Tom

>   	if ((sp->psp_irq == sp->ccp_irq) && sp->dev_vdata->psp_vdata) {
>   		/* Using common routine to manage all interrupts */
>   		if (!sp->psp_irq_handler) {
> @@ -151,6 +154,9 @@ void sp_free_ccp_irq(struct sp_device *sp, void *data)
>   
>   void sp_free_psp_irq(struct sp_device *sp, void *data)
>   {
> +	if (!sp->irq_registered)
> +		return;
> +
>   	if ((sp->psp_irq == sp->ccp_irq) && sp->dev_vdata->ccp_vdata) {
>   		/* Using common routine to manage all interrupts */
>   		if (!sp->ccp_irq_handler) {