|
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] Re: [PATCH] PCI: deviate MISRA R11.8 for the one-time SBDF initialization
On 17.11.2026 22:33, Andrew Cooper wrote: > On 17/11/2026 8:55 pm, Dmytro Prokopchuk1 wrote: >> MISRA C Rule 11.8 states: "A cast shall not remove any const or volatile >> qualification from the type pointed to by a pointer". >> >> alloc_pdev() writes the device identity once. seg, bus and devfn are >> members of a const union, so that write goes through a cast that removes >> const. Three such casts performed the same initialization. >> >> Store the whole sbdf in one assignment and document that cast with >> one-line SAF-15-safe comment. >> >> No functional changes. >> >> Signed-off-by: Dmytro Prokopchuk <dmytro_prokopchuk1@xxxxxxxx> >> --- >> Test CI pipeline: >> https://gitlab.com/xen-project/people/dimaprkp4k/xen/-/pipelines/3081067362 >> --- >> docs/misra/safe.json | 8 ++++++++ >> xen/drivers/passthrough/pci.c | 5 ++--- >> 2 files changed, 10 insertions(+), 3 deletions(-) >> >> diff --git a/docs/misra/safe.json b/docs/misra/safe.json >> index 3584cb90c6..3503d2e808 100644 >> --- a/docs/misra/safe.json >> +++ b/docs/misra/safe.json >> @@ -124,6 +124,14 @@ >> }, >> { >> "id": "SAF-15-safe", >> + "analyser": { >> + "eclair": "MC3A2.R11.8" >> + }, >> + "name": "Rule 11.8: one-time init of a const PCI identity", >> + "text": "The cast removes const only to initialize the const >> pci_dev identity fields (seg, bus, devfn). The const qualification prevents >> any later assignment, and this is the only write, performed before the >> device is published on alldevs_list." >> + }, >> + { >> + "id": "SAF-16-safe", >> "analyser": {}, >> "name": "Sentinel", >> "text": "Next ID to be used" >> diff --git a/xen/drivers/passthrough/pci.c b/xen/drivers/passthrough/pci.c >> index 578472d068..14d59b1293 100644 >> --- a/xen/drivers/passthrough/pci.c >> +++ b/xen/drivers/passthrough/pci.c >> @@ -329,9 +329,8 @@ static struct pci_dev *alloc_pdev(struct pci_seg *pseg, >> u8 bus, u8 devfn) >> if ( !pdev ) >> return NULL; >> >> - *(u16*) &pdev->seg = pseg->nr; >> - *((u8*) &pdev->bus) = bus; >> - *((u8*) &pdev->devfn) = devfn; >> + /* SAF-15-safe R11.8 one-time init of the const PCI identity */ >> + *(pci_sbdf_t *)&pdev->sbdf = PCI_SBDF(pseg->nr, bus, devfn); >> pdev->domain = NULL; >> >> INIT_LIST_HEAD(&pdev->vf_list); > > I'm sorry you're in the middle of a fight between the maintainers, but > this too is absurd. > > There is nothing about pdev->sbdf worthy of an exception like this. We > have loads of other structures with identifiers that **if and only if we > were writing C++** would be const members. From a practical standpoint > d->domain_id or v->vcpu_id would be far more worthy than pdev->sbdf of > special casing in this way. As there doesn't look to be any interest in widening the pattern (as you say, there are many more fields where such protection would help), I agree ... > But, Xen is written in C. > > I accept that this isn't creating a class of undefined behaviour, but > it's bogus-looking code try to try and use a C++-ism. In the modern > world of AI's, we're at much greater risk of an LLM seeing this and > deciding it's an acceptable code pattern in Xen. > > > Please just delete the const and make a plain SBDF assignment to drop > the current casts. ... there's no point having such special treatment just here. Hence ... > If I need to call a formal governance vote to put an end to the > maintainer infighting, I will. It has gone on far too long, and is > continuing to waste time we simply don't have. ... I then think no vote is going to be necessary (since aiui it's just me you would want to outvote). Jan
|
![]() |
Lists.xenproject.org is hosted with RackSpace, monitoring our |